The DOI matches its Crossref record; 33 of the 33 references it cites were found in a registry or an index; 3 of the 21 sources its claims cite were checked for support.
The Moral Machine experiment
Open accessAbstract
With the rapid development of artificial intelligence have come concerns about how machines will make moral decisions, and the major challenge of quantifying societal expectations about the ethical principles that should guide machine behaviour. To address this challenge, we deployed the Moral Machine, an online experimental platform designed to explore the moral dilemmas faced by autonomous vehicles. This platform gathered 40 million decisions in ten languages from millions of people in 233 countries and territories. Here we describe the results of this experiment. First, we summarize global moral preferences. Second, we document individual variations in preferences, based on respondents' demographics. Third, we report cross-cultural ethical variation, and uncover three major clusters of countries. Fourth, we show that these differences correlate with modern institutions and deep cultural traits. We discuss how these preferences can contribute to developing global, socially acceptable principles for machine ethics. All data used in this article are publicly available.
Overview
AI snapshot
Written from the paper’s full text.
TL;DR
The Moral Machine used a multilingual online game to gather millions of judgments about unavoidable autonomous-vehicle crash dilemmas, revealing strong global preferences for sparing humans, more lives, and younger lives. It also found demographic variation that did not reverse preferences, three broad cultural clusters, and country-level correlations with cultural and economic indicators.
Key findings
- The platform collected 39.61 million decisions from 233 countries and territories; the abstract describes roughly 40 million decisions from over 2.3 million people in ten languages.
- Worldwide, the strongest preferences were for sparing humans over animals, sparing more characters, and sparing younger characters; weaker preferences included sparing women and sparing pedestrians.
- Among 492,921 respondents who completed the optional demographic survey, individual characteristics such as age, gender, income, education, politics, and religion did not change the direction of any preference, with effects all below 0.1.
- Country-level clustering identified three groups—Western, Eastern, and Southern—and preferences correlated with individualism/collectivism, GDP per capita, rule of law, economic inequality, and gender gaps in health and survival.
Methods
- The Moral Machine was an online “serious game” launched in June 2016 and later translated into nine additional languages; users chose between two outcomes in unavoidable accident scenarios.
- In Judge mode, sessions contained 13 dilemmas generated around nine factors: species, number of lives, age, gender, physical fitness, social status, pedestrian legality, passenger versus pedestrian status, and whether the car swerves.
- Analyses used conjoint analysis to estimate average marginal component effects, hierarchical clustering of 130 countries with at least 100 respondents, and country-level correlations with socio-economic and cultural indicators.
Limitations
- The paper states that the sample was self-selected and not guaranteed to be representative, so policymakers should not treat the results as the final word on societal preferences.
- The paper states that scenarios assumed certainty about character classification and life-and-death outcomes, did not manipulate respondent-character relationships, and captured preferences before autonomous vehicles are deployed.
- Evident from the text: the findings are based on hypothetical choices from an internet-connected, tech-interested population, and the cross-sectional design cannot establish causal mechanisms behind country-level correlations.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
38 Own result · 16 Own interpretation · 7 Own method · 7 No basis given · 6 No full text of the sources · 5 Sources not found · 3 Not checked yet · 2 Partly supported · 1 Supported · accuracy not yet measured
Every source the paper cites for this claim was checked and supports it.
In the paper
“Our previous work did not find a strong impact of this variable on moral preferences 12”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > DISCUSSION
What the sources say
- Supportedcites [12] Bonnefon, J.-F., Shariff, A. & Rahwan, I. The social dilemma of autonomous vehicles. Science 352, 1573–1576 (2016). source
The Figure 3 caption (p. 11) says the moral preference was robust when respondents imagined riding with a coworker, family member, or their own child. Study 3 (p. 4) found that imagining a family member significantly lowered the morality rating, but it remained above the scale midpoint; thus the relationship affected ratings without overturning the preference. (LLM, conf=0.87)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“These codes seldom recognize, though, that humans experience inner conflict, interpersonal disagreements, and cultural dissimilarities in the moral domain 31 , 32 , 33 .”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > DISCUSSION
What the sources say
- Not checked — no full text of the sourcecites [31] Gastil, J., Braman, D., Kahan, D. & Slovic, P. The cultural orientation of mass political opinion. PS Polit. Sci. Polit. 44, 711–714 (2011). source
- Partly supportedcites [32] Nishi, A., Christakis, N. A. & Rand, D. G. Cooperation, decision time, and culture: online experiments with American and Indian participants. PLoS ONE 12, e0171252 (2017). sourceExperimental — accuracy not yet measured
The Introduction describes conflicted choices between cooperation and defection, and the Discussion reports cross-cultural differences in cooperation between American and Indian participants. The paper does not address interpersonal disagreements or whether the codes in the citing sentence recognize these experiences; its findings concern cooperation in economic games rather than the moral domain generally. (LLM, conf=0.85)
- Not matchedcites [33]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
In the paper
“are broadly consistent with both geographical and cultural proximity according to the Inglehart-Welzel Cultural Map 2010-2014 21 .”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > CULTURAL CLUSTERS
What the sources say
- Not checked — no full text of the sourcecites [21] Inglehart, R. & Welzel, C. Modernization, Cultural Change, and Democracy: The Human Development Sequence (Cambridge Univ. Press, Cambridge, 2005).
plaincorrelationaln = 100- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption: The 130 countries with at least 100 respondents are representative of their countries' moral preferences.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Spain appears in the Western cluster, adjacent to Portugal and other European countries, rather than in the Spanish-speaking Latin American sub-cluster.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > CULTURAL CLUSTERS
hedged- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption: The 130 countries with at least 100 respondents are representative of their countries' moral preferences.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“clusters largely differ in the weight they give to some preferences.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > CULTURAL CLUSTERS
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption: The 130 countries with at least 100 respondents are representative of their countries' moral preferences.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“uncover three major clusters of countries exhibiting substantial differences along key moral preferences about autonomous vehicles”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > Abstract With the rapid development of Artificial Intelligence technology come widespread concerns about how machines will behave in morally charged situations.Addressing these concerns raises the major challenge of quantifying societal expectations about the ethical principles that should guide machine behavior.To address this challenge, we deployed the Moral Machine, an Internet-based experimental platform that is designed to explore the multi-dimensional moral dilemmas faced by autonomous vehicles.This platform enabled us to gather 40 million decisions in ten languages from over 2.3 million people in 233 countries and territories, and thus to assess the paths and obstacles to machine ethics.First, we summarize global moral preferences, some of them strong (e.g., sparing more lives or younger individuals), and some of them weak (e.g., sparing women, sparing pedestrians).Second, we document individual variations of these preferences, based on respondents' demographics (e.g., age, gender, income), and find that demographics do not change the direction of any preference.Third, we report cross-cultural ethical variation, and uncover three major clusters of countries exhibiting substantial differences along key moral preferences about autonomous vehicles.Fourth, we show that these differences correlate with modern institutions, but also with deep cultural traits.We discuss how these three layers of preferences can contribute to developing global, harmonious, and socially acceptable principles for machine ethics.All data used in this article can be accessed and downloaded at https://goo.gl/JXRrBP .We are entering an age in which machines are not only tasked to promote well-being and minimize harm, but also to distribute the well-being they create, and the harm they cannot eliminate.Distributing well-being and harm inevitably creates tradeoffs, whose resolution falls in the moral domain 1 , 2 , 3 .Think of an autonomous vehicle (AV) that is about to crash, and cannot find a trajectory that would save everyone.Should it swerve onto one jaywalking teenager to spare its three elderly passengers?Should it swerve away from a group of children and kill a single adult passenger instead?Even in the more common instances in which harm is not inevitable, but just possible, AVs will need to decide how to divvy up the risk of harm between the different stakeholders on the road.Car manufacturers and policymakers are currently struggling with these moral dilemmas, in large part because they cannot be solved by any simple normative ethical principles like Asimov's laws of robotics 4 . Asimov's laws were not designed to solve the problem of universal machine ethics, and they were not even designed to let machines distribute harm between humans.They were a narrative device whose goal was to generate good stories, by showcasing how challenging it is to create moral machines with a dozen lines of code.And yet, we do not have the luxury to give up on creating moral machines 5 , 6 , 7 , 8 .AVs will cruise our roads soon, necessitating agreement on the principles that should apply when, inevitably, life-threatening dilemmas emerge.The frequency at which these dilemma will emerge is extremely hard to estimate, just as it is extremely hard to estimate the rate at which human drivers find themselves in comparable situations.Human drivers who die in crashes cannot report whether they were faced with a dilemma; and human drivers who survive a crash may not have realized that they were in a dilemma situation.Note though that ethical guidelines for AV choices in dilemma situations do not depend on the frequency of these situations.Whether these cases are rare, very rare, or extremely rare, we need to agree beforehand on how they should be solved. The keyword here is "we".As emphasized by former U.S. president Barack Obama 9 , consensus in this matter is going to be important.Decisions about the ethical principles that will guide AVs cannot be left to solely to either the engineers or the ethicists.For consumers to switch from traditional human-driven cars to AVs, and for the wider public to accept the proliferation of AI-driven vehicles on their roads, both groups will need to understand the origins of the ethical principles programmed into these vehicles 10 .In other words, even if ethicists were to agree on how AVs should solve moral dilemmas, their work would be useless if citizens were to disagree with their solution, and thus opt out of the future that AVs promise in lieu of the status quo.Any attempt to devise AI ethics must be at least cognizant of public morality. Accordingly, we need to gauge social expectations about the way AVs should solve moral dilemmas.This enterprise, however, is not without challenges 11 .The first challenge comes from the high-dimensionality of the problem.In a typical survey, one may test whether people prefer to spare many lives rather than few 9,12 , 13 ; or whether people prefer to spare the young rather than the elderly 14 , 15 ; or whether people prefer to spare pedestrians who cross legally, rather than pedestrians who jaywalk; or yet some other preference, or a simple combination of two or three of these preferences.But combining a dozen of such preferences leads to millions of possible scenarios, requiring a sample size that defies any conventional method of data collection.
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption: The 130 countries with at least 100 respondents are representative of their countries' moral preferences.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The same is true about the preference for sparing higher status characters.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > CULTURAL CLUSTERS
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption: The 130 countries with at least 100 respondents are representative of their countries' moral preferences.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“we find that individual variations have no sizable impact on any of the nine attributes (all below 0.1; see Figure S6 in SI).”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > INDIVIDUAL VARIATIONS
plainn = 492,921- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“find that demographics do not change the direction of any preference”
- Also stated in: § INDIVIDUAL VARIATIONS · plain
“none of the six characteristics splits its subpopulations into opposing directions of effect”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > Abstract With the rapid development of Artificial Intelligence technology come widespread concerns about how machines will behave in morally charged situations.Addressing these concerns raises the major challenge of quantifying societal expectations about the ethical principles that should guide machine behavior.To address this challenge, we deployed the Moral Machine, an Internet-based experimental platform that is designed to explore the multi-dimensional moral dilemmas faced by autonomous vehicles.This platform enabled us to gather 40 million decisions in ten languages from over 2.3 million people in 233 countries and territories, and thus to assess the paths and obstacles to machine ethics.First, we summarize global moral preferences, some of them strong (e.g., sparing more lives or younger individuals), and some of them weak (e.g., sparing women, sparing pedestrians).Second, we document individual variations of these preferences, based on respondents' demographics (e.g., age, gender, income), and find that demographics do not change the direction of any preference.Third, we report cross-cultural ethical variation, and uncover three major clusters of countries exhibiting substantial differences along key moral preferences about autonomous vehicles.Fourth, we show that these differences correlate with modern institutions, but also with deep cultural traits.We discuss how these three layers of preferences can contribute to developing global, harmonious, and socially acceptable principles for machine ethics.All data used in this article can be accessed and downloaded at https://goo.gl/JXRrBP .We are entering an age in which machines are not only tasked to promote well-being and minimize harm, but also to distribute the well-being they create, and the harm they cannot eliminate.Distributing well-being and harm inevitably creates tradeoffs, whose resolution falls in the moral domain 1 , 2 , 3 .Think of an autonomous vehicle (AV) that is about to crash, and cannot find a trajectory that would save everyone.Should it swerve onto one jaywalking teenager to spare its three elderly passengers?Should it swerve away from a group of children and kill a single adult passenger instead?Even in the more common instances in which harm is not inevitable, but just possible, AVs will need to decide how to divvy up the risk of harm between the different stakeholders on the road.Car manufacturers and policymakers are currently struggling with these moral dilemmas, in large part because they cannot be solved by any simple normative ethical principles like Asimov's laws of robotics 4 . Asimov's laws were not designed to solve the problem of universal machine ethics, and they were not even designed to let machines distribute harm between humans.They were a narrative device whose goal was to generate good stories, by showcasing how challenging it is to create moral machines with a dozen lines of code.And yet, we do not have the luxury to give up on creating moral machines 5 , 6 , 7 , 8 .AVs will cruise our roads soon, necessitating agreement on the principles that should apply when, inevitably, life-threatening dilemmas emerge.The frequency at which these dilemma will emerge is extremely hard to estimate, just as it is extremely hard to estimate the rate at which human drivers find themselves in comparable situations.Human drivers who die in crashes cannot report whether they were faced with a dilemma; and human drivers who survive a crash may not have realized that they were in a dilemma situation.Note though that ethical guidelines for AV choices in dilemma situations do not depend on the frequency of these situations.Whether these cases are rare, very rare, or extremely rare, we need to agree beforehand on how they should be solved. The keyword here is "we".As emphasized by former U.S. president Barack Obama 9 , consensus in this matter is going to be important.Decisions about the ethical principles that will guide AVs cannot be left to solely to either the engineers or the ethicists.For consumers to switch from traditional human-driven cars to AVs, and for the wider public to accept the proliferation of AI-driven vehicles on their roads, both groups will need to understand the origins of the ethical principles programmed into these vehicles 10 .In other words, even if ethicists were to agree on how AVs should solve moral dilemmas, their work would be useless if citizens were to disagree with their solution, and thus opt out of the future that AVs promise in lieu of the status quo.Any attempt to devise AI ethics must be at least cognizant of public morality. Accordingly, we need to gauge social expectations about the way AVs should solve moral dilemmas.This enterprise, however, is not without challenges 11 .The first challenge comes from the high-dimensionality of the problem.In a typical survey, one may test whether people prefer to spare many lives rather than few 9,12 , 13 ; or whether people prefer to spare the young rather than the elderly 14 , 15 ; or whether people prefer to spare pedestrians who cross legally, rather than pedestrians who jaywalk; or yet some other preference, or a simple combination of two or three of these preferences.But combining a dozen of such preferences leads to millions of possible scenarios, requiring a sample size that defies any conventional method of data collection.
plainn = 492,921- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § INDIVIDUAL VARIATIONS · plain
In the paper
“Figure 4c shows a substantial correlation ( ) between this MM distance and the cultural distance from the US based on the World Values .49ρ = 0 Survey 22 .”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > COUNTRY-LEVEL PREDICTORS
What the sources say
- Not checked — no full text of the sourcecites [22] Muthukrishna, M. Beyond WEIRD psychology: measuring and mapping scales of cultural and psychological distance. Preprint at https://ssrn.com/abstract=3259613 (2018). source
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption: The 130 countries with at least 100 respondents are representative of their countries' moral preferences.
- It may state two separate claims as one. 28
- The quote only partly states this claim. 1
- Probably not a substantial claim. 16
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“participants from countries which are poorer and suffer from weaker institutions are more tolerant of pedestrians who cross illegally, presumably because of their experience of lower rule compliance and weaker punishment of rule deviation 26 .”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > COUNTRY-LEVEL PREDICTORS
What the sources say
- Partly supportedcites [26] Gächter, S. & Schulz, J. F. Intrinsic honesty and the prevalence of rule violations across societies. Nature 531, 496–499 (2016). sourceExperimental — accuracy not yet measured
The paper reports in its main text (labelled Abstract in our copy; page unavailable) that participants from countries with more prevalent rule violations showed less intrinsic honesty, and it suggests that unpunished cheating may make dishonesty seem justifiable. It does not report that participants from poorer countries were more tolerant of pedestrians crossing illegally, or establish weaker punishment as the reason for such tolerance; its experiment measured honesty using a die-rolling task. (LLM, conf=0.83)
plainconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Partly supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“some of them strong (e.g., sparing more lives or younger individuals), and some of them weak (e.g., sparing women, sparing pedestrians)”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > Abstract With the rapid development of Artificial Intelligence technology come widespread concerns about how machines will behave in morally charged situations.Addressing these concerns raises the major challenge of quantifying societal expectations about the ethical principles that should guide machine behavior.To address this challenge, we deployed the Moral Machine, an Internet-based experimental platform that is designed to explore the multi-dimensional moral dilemmas faced by autonomous vehicles.This platform enabled us to gather 40 million decisions in ten languages from over 2.3 million people in 233 countries and territories, and thus to assess the paths and obstacles to machine ethics.First, we summarize global moral preferences, some of them strong (e.g., sparing more lives or younger individuals), and some of them weak (e.g., sparing women, sparing pedestrians).Second, we document individual variations of these preferences, based on respondents' demographics (e.g., age, gender, income), and find that demographics do not change the direction of any preference.Third, we report cross-cultural ethical variation, and uncover three major clusters of countries exhibiting substantial differences along key moral preferences about autonomous vehicles.Fourth, we show that these differences correlate with modern institutions, but also with deep cultural traits.We discuss how these three layers of preferences can contribute to developing global, harmonious, and socially acceptable principles for machine ethics.All data used in this article can be accessed and downloaded at https://goo.gl/JXRrBP .We are entering an age in which machines are not only tasked to promote well-being and minimize harm, but also to distribute the well-being they create, and the harm they cannot eliminate.Distributing well-being and harm inevitably creates tradeoffs, whose resolution falls in the moral domain 1 , 2 , 3 .Think of an autonomous vehicle (AV) that is about to crash, and cannot find a trajectory that would save everyone.Should it swerve onto one jaywalking teenager to spare its three elderly passengers?Should it swerve away from a group of children and kill a single adult passenger instead?Even in the more common instances in which harm is not inevitable, but just possible, AVs will need to decide how to divvy up the risk of harm between the different stakeholders on the road.Car manufacturers and policymakers are currently struggling with these moral dilemmas, in large part because they cannot be solved by any simple normative ethical principles like Asimov's laws of robotics 4 . Asimov's laws were not designed to solve the problem of universal machine ethics, and they were not even designed to let machines distribute harm between humans.They were a narrative device whose goal was to generate good stories, by showcasing how challenging it is to create moral machines with a dozen lines of code.And yet, we do not have the luxury to give up on creating moral machines 5 , 6 , 7 , 8 .AVs will cruise our roads soon, necessitating agreement on the principles that should apply when, inevitably, life-threatening dilemmas emerge.The frequency at which these dilemma will emerge is extremely hard to estimate, just as it is extremely hard to estimate the rate at which human drivers find themselves in comparable situations.Human drivers who die in crashes cannot report whether they were faced with a dilemma; and human drivers who survive a crash may not have realized that they were in a dilemma situation.Note though that ethical guidelines for AV choices in dilemma situations do not depend on the frequency of these situations.Whether these cases are rare, very rare, or extremely rare, we need to agree beforehand on how they should be solved. The keyword here is "we".As emphasized by former U.S. president Barack Obama 9 , consensus in this matter is going to be important.Decisions about the ethical principles that will guide AVs cannot be left to solely to either the engineers or the ethicists.For consumers to switch from traditional human-driven cars to AVs, and for the wider public to accept the proliferation of AI-driven vehicles on their roads, both groups will need to understand the origins of the ethical principles programmed into these vehicles 10 .In other words, even if ethicists were to agree on how AVs should solve moral dilemmas, their work would be useless if citizens were to disagree with their solution, and thus opt out of the future that AVs promise in lieu of the status quo.Any attempt to devise AI ethics must be at least cognizant of public morality. Accordingly, we need to gauge social expectations about the way AVs should solve moral dilemmas.This enterprise, however, is not without challenges 11 .The first challenge comes from the high-dimensionality of the problem.In a typical survey, one may test whether people prefer to spare many lives rather than few 9,12 , 13 ; or whether people prefer to spare the young rather than the elderly 14 , 15 ; or whether people prefer to spare pedestrians who cross legally, rather than pedestrians who jaywalk; or yet some other preference, or a simple combination of two or three of these preferences.But combining a dozen of such preferences leads to millions of possible scenarios, requiring a sample size that defies any conventional method of data collection.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“the preference to spare younger characters rather than older characters is much less pronounced for countries in the Eastern cluster, and much higher for countries in the Southern cluster.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > CULTURAL CLUSTERS
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Rests on an assumption: The 130 countries with at least 100 respondents are representative of their countries' moral preferences.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The third cluster (a broadly Southern cluster) consists of the Latin American countries of Central and South America, in addition to some countries that are characterized in part by French influence”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > CULTURAL CLUSTERS
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Rests on an assumption: The 130 countries with at least 100 respondents are representative of their countries' moral preferences.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Only the (weak) preference for sparing pedestrians over passengers and the (moderate) preference for sparing the lawful over the unlawful appear to be shared to the same extent in all clusters.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > CULTURAL CLUSTERS
hedged- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Rests on an assumption: The 130 countries with at least 100 respondents are representative of their countries' moral preferences.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“both male and female respondents indicated Δ preference for Sparing Females, but the latter group showed stronger preference”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > INDIVIDUAL VARIATIONS
plaincorrelationaln = 492,921- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“male respondents are 0.06 percentage point less inclined to spare females and 0.09 percentage point more inclined to spare humans compared to female respondents”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > INDIVIDUAL VARIATIONS
plaincorrelationaln = 492,921- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The second cluster (which we call the Eastern cluster) contains many far eastern countries such as Japan and Taiwan, belonging to the Confucianist cultural group, and Islamic countries such as Indonesia, Pakistan and Saudi Arabia.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > CULTURAL CLUSTERS
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Rests on an assumption: The 130 countries with at least 100 respondents are representative of their countries' moral preferences.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“these differences correlate with modern institutions, but also with deep cultural traits”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > Abstract With the rapid development of Artificial Intelligence technology come widespread concerns about how machines will behave in morally charged situations.Addressing these concerns raises the major challenge of quantifying societal expectations about the ethical principles that should guide machine behavior.To address this challenge, we deployed the Moral Machine, an Internet-based experimental platform that is designed to explore the multi-dimensional moral dilemmas faced by autonomous vehicles.This platform enabled us to gather 40 million decisions in ten languages from over 2.3 million people in 233 countries and territories, and thus to assess the paths and obstacles to machine ethics.First, we summarize global moral preferences, some of them strong (e.g., sparing more lives or younger individuals), and some of them weak (e.g., sparing women, sparing pedestrians).Second, we document individual variations of these preferences, based on respondents' demographics (e.g., age, gender, income), and find that demographics do not change the direction of any preference.Third, we report cross-cultural ethical variation, and uncover three major clusters of countries exhibiting substantial differences along key moral preferences about autonomous vehicles.Fourth, we show that these differences correlate with modern institutions, but also with deep cultural traits.We discuss how these three layers of preferences can contribute to developing global, harmonious, and socially acceptable principles for machine ethics.All data used in this article can be accessed and downloaded at https://goo.gl/JXRrBP .We are entering an age in which machines are not only tasked to promote well-being and minimize harm, but also to distribute the well-being they create, and the harm they cannot eliminate.Distributing well-being and harm inevitably creates tradeoffs, whose resolution falls in the moral domain 1 , 2 , 3 .Think of an autonomous vehicle (AV) that is about to crash, and cannot find a trajectory that would save everyone.Should it swerve onto one jaywalking teenager to spare its three elderly passengers?Should it swerve away from a group of children and kill a single adult passenger instead?Even in the more common instances in which harm is not inevitable, but just possible, AVs will need to decide how to divvy up the risk of harm between the different stakeholders on the road.Car manufacturers and policymakers are currently struggling with these moral dilemmas, in large part because they cannot be solved by any simple normative ethical principles like Asimov's laws of robotics 4 . Asimov's laws were not designed to solve the problem of universal machine ethics, and they were not even designed to let machines distribute harm between humans.They were a narrative device whose goal was to generate good stories, by showcasing how challenging it is to create moral machines with a dozen lines of code.And yet, we do not have the luxury to give up on creating moral machines 5 , 6 , 7 , 8 .AVs will cruise our roads soon, necessitating agreement on the principles that should apply when, inevitably, life-threatening dilemmas emerge.The frequency at which these dilemma will emerge is extremely hard to estimate, just as it is extremely hard to estimate the rate at which human drivers find themselves in comparable situations.Human drivers who die in crashes cannot report whether they were faced with a dilemma; and human drivers who survive a crash may not have realized that they were in a dilemma situation.Note though that ethical guidelines for AV choices in dilemma situations do not depend on the frequency of these situations.Whether these cases are rare, very rare, or extremely rare, we need to agree beforehand on how they should be solved. The keyword here is "we".As emphasized by former U.S. president Barack Obama 9 , consensus in this matter is going to be important.Decisions about the ethical principles that will guide AVs cannot be left to solely to either the engineers or the ethicists.For consumers to switch from traditional human-driven cars to AVs, and for the wider public to accept the proliferation of AI-driven vehicles on their roads, both groups will need to understand the origins of the ethical principles programmed into these vehicles 10 .In other words, even if ethicists were to agree on how AVs should solve moral dilemmas, their work would be useless if citizens were to disagree with their solution, and thus opt out of the future that AVs promise in lieu of the status quo.Any attempt to devise AI ethics must be at least cognizant of public morality. Accordingly, we need to gauge social expectations about the way AVs should solve moral dilemmas.This enterprise, however, is not without challenges 11 .The first challenge comes from the high-dimensionality of the problem.In a typical survey, one may test whether people prefer to spare many lives rather than few 9,12 , 13 ; or whether people prefer to spare the young rather than the elderly 14 , 15 ; or whether people prefer to spare pedestrians who cross legally, rather than pedestrians who jaywalk; or yet some other preference, or a simple combination of two or three of these preferences.But combining a dozen of such preferences leads to millions of possible scenarios, requiring a sample size that defies any conventional method of data collection.
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Rests on an assumption: The 130 countries with at least 100 respondents are representative of their countries' moral preferences.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“the four most spared characters are the baby, the little girl, the little boy, and the pregnant woman”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > GLOBAL PREFERENCES
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“or for that matter, women over men, overweight persons over athletes, or executives over homeless persons--all of which we see weaker but clear effects for.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > GLOBAL PREFERENCES
boostedcorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“This platform enabled us to gather 40 million decisions in ten languages from over 2.3 million people in 233 countries and territories”
- Also stated in: § Abstract With the rapid development of Artificial Intelligence technology come widespread concerns about how machines will behave in morally charged situations.Addressing these concerns raises the major challenge of quantifying societal expectations about the ethical principles that should guide machine behavior.To address this challenge, we deployed the Moral Machine, an Internet-based experimental platform that is designed to explore the multi-dimensional moral dilemmas faced by autonomous vehicles.This platform enabled us to gather 40 million decisions in ten languages from over 2.3 million people in 233 countries and territories, and thus to assess the paths and obstacles to machine ethics.First, we summarize global moral preferences, some of them strong (e.g., sparing more lives or younger individuals), and some of them weak (e.g., sparing women, sparing pedestrians).Second, we document individual variations of these preferences, based on respondents' demographics (e.g., age, gender, income), and find that demographics do not change the direction of any preference.Third, we report cross-cultural ethical variation, and uncover three major clusters of countries exhibiting substantial differences along key moral preferences about autonomous vehicles.Fourth, we show that these differences correlate with modern institutions, but also with deep cultural traits.We discuss how these three layers of preferences can contribute to developing global, harmonious, and socially acceptable principles for machine ethics.All data used in this article can be accessed and downloaded at https://goo.gl/JXRrBP .We are entering an age in which machines are not only tasked to promote well-being and minimize harm, but also to distribute the well-being they create, and the harm they cannot eliminate.Distributing well-being and harm inevitably creates tradeoffs, whose resolution falls in the moral domain 1 , 2 , 3 .Think of an autonomous vehicle (AV) that is about to crash, and cannot find a trajectory that would save everyone.Should it swerve onto one jaywalking teenager to spare its three elderly passengers?Should it swerve away from a group of children and kill a single adult passenger instead?Even in the more common instances in which harm is not inevitable, but just possible, AVs will need to decide how to divvy up the risk of harm between the different stakeholders on the road.Car manufacturers and policymakers are currently struggling with these moral dilemmas, in large part because they cannot be solved by any simple normative ethical principles like Asimov's laws of robotics 4 . Asimov's laws were not designed to solve the problem of universal machine ethics, and they were not even designed to let machines distribute harm between humans.They were a narrative device whose goal was to generate good stories, by showcasing how challenging it is to create moral machines with a dozen lines of code.And yet, we do not have the luxury to give up on creating moral machines 5 , 6 , 7 , 8 .AVs will cruise our roads soon, necessitating agreement on the principles that should apply when, inevitably, life-threatening dilemmas emerge.The frequency at which these dilemma will emerge is extremely hard to estimate, just as it is extremely hard to estimate the rate at which human drivers find themselves in comparable situations.Human drivers who die in crashes cannot report whether they were faced with a dilemma; and human drivers who survive a crash may not have realized that they were in a dilemma situation.Note though that ethical guidelines for AV choices in dilemma situations do not depend on the frequency of these situations.Whether these cases are rare, very rare, or extremely rare, we need to agree beforehand on how they should be solved. The keyword here is "we".As emphasized by former U.S. president Barack Obama 9 , consensus in this matter is going to be important.Decisions about the ethical principles that will guide AVs cannot be left to solely to either the engineers or the ethicists.For consumers to switch from traditional human-driven cars to AVs, and for the wider public to accept the proliferation of AI-driven vehicles on their roads, both groups will need to understand the origins of the ethical principles programmed into these vehicles 10 .In other words, even if ethicists were to agree on how AVs should solve moral dilemmas, their work would be useless if citizens were to disagree with their solution, and thus opt out of the future that AVs promise in lieu of the status quo.Any attempt to devise AI ethics must be at least cognizant of public morality. Accordingly, we need to gauge social expectations about the way AVs should solve moral dilemmas.This enterprise, however, is not without challenges 11 .The first challenge comes from the high-dimensionality of the problem.In a typical survey, one may test whether people prefer to spare many lives rather than few 9,12 , 13 ; or whether people prefer to spare the young rather than the elderly 14 , 15 ; or whether people prefer to spare pedestrians who cross legally, rather than pedestrians who jaywalk; or yet some other preference, or a simple combination of two or three of these preferences.But combining a dozen of such preferences leads to millions of possible scenarios, requiring a sample size that defies any conventional method of data collection. · plain
“allowed us to collect 39.61 million decisions in 233 countries, dependencies, or territories”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > Abstract With the rapid development of Artificial Intelligence technology come widespread concerns about how machines will behave in morally charged situations.Addressing these concerns raises the major challenge of quantifying societal expectations about the ethical principles that should guide machine behavior.To address this challenge, we deployed the Moral Machine, an Internet-based experimental platform that is designed to explore the multi-dimensional moral dilemmas faced by autonomous vehicles.This platform enabled us to gather 40 million decisions in ten languages from over 2.3 million people in 233 countries and territories, and thus to assess the paths and obstacles to machine ethics.First, we summarize global moral preferences, some of them strong (e.g., sparing more lives or younger individuals), and some of them weak (e.g., sparing women, sparing pedestrians).Second, we document individual variations of these preferences, based on respondents' demographics (e.g., age, gender, income), and find that demographics do not change the direction of any preference.Third, we report cross-cultural ethical variation, and uncover three major clusters of countries exhibiting substantial differences along key moral preferences about autonomous vehicles.Fourth, we show that these differences correlate with modern institutions, but also with deep cultural traits.We discuss how these three layers of preferences can contribute to developing global, harmonious, and socially acceptable principles for machine ethics.All data used in this article can be accessed and downloaded at https://goo.gl/JXRrBP .We are entering an age in which machines are not only tasked to promote well-being and minimize harm, but also to distribute the well-being they create, and the harm they cannot eliminate.Distributing well-being and harm inevitably creates tradeoffs, whose resolution falls in the moral domain 1 , 2 , 3 .Think of an autonomous vehicle (AV) that is about to crash, and cannot find a trajectory that would save everyone.Should it swerve onto one jaywalking teenager to spare its three elderly passengers?Should it swerve away from a group of children and kill a single adult passenger instead?Even in the more common instances in which harm is not inevitable, but just possible, AVs will need to decide how to divvy up the risk of harm between the different stakeholders on the road.Car manufacturers and policymakers are currently struggling with these moral dilemmas, in large part because they cannot be solved by any simple normative ethical principles like Asimov's laws of robotics 4 . Asimov's laws were not designed to solve the problem of universal machine ethics, and they were not even designed to let machines distribute harm between humans.They were a narrative device whose goal was to generate good stories, by showcasing how challenging it is to create moral machines with a dozen lines of code.And yet, we do not have the luxury to give up on creating moral machines 5 , 6 , 7 , 8 .AVs will cruise our roads soon, necessitating agreement on the principles that should apply when, inevitably, life-threatening dilemmas emerge.The frequency at which these dilemma will emerge is extremely hard to estimate, just as it is extremely hard to estimate the rate at which human drivers find themselves in comparable situations.Human drivers who die in crashes cannot report whether they were faced with a dilemma; and human drivers who survive a crash may not have realized that they were in a dilemma situation.Note though that ethical guidelines for AV choices in dilemma situations do not depend on the frequency of these situations.Whether these cases are rare, very rare, or extremely rare, we need to agree beforehand on how they should be solved. The keyword here is "we".As emphasized by former U.S. president Barack Obama 9 , consensus in this matter is going to be important.Decisions about the ethical principles that will guide AVs cannot be left to solely to either the engineers or the ethicists.For consumers to switch from traditional human-driven cars to AVs, and for the wider public to accept the proliferation of AI-driven vehicles on their roads, both groups will need to understand the origins of the ethical principles programmed into these vehicles 10 .In other words, even if ethicists were to agree on how AVs should solve moral dilemmas, their work would be useless if citizens were to disagree with their solution, and thus opt out of the future that AVs promise in lieu of the status quo.Any attempt to devise AI ethics must be at least cognizant of public morality. Accordingly, we need to gauge social expectations about the way AVs should solve moral dilemmas.This enterprise, however, is not without challenges 11 .The first challenge comes from the high-dimensionality of the problem.In a typical survey, one may test whether people prefer to spare many lives rather than few 9,12 , 13 ; or whether people prefer to spare the young rather than the elderly 14 , 15 ; or whether people prefer to spare pedestrians who cross legally, rather than pedestrians who jaywalk; or yet some other preference, or a simple combination of two or three of these preferences.But combining a dozen of such preferences leads to millions of possible scenarios, requiring a sample size that defies any conventional method of data collection.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one. The quote only partly states this claim.
- Also stated in: § Abstract With the rapid development of Artificial Intelligence technology come widespread concerns about how machines will behave in morally charged situations.Addressing these concerns raises the major challenge of quantifying societal expectations about the ethical principles that should guide machine behavior.To address this challenge, we deployed the Moral Machine, an Internet-based experimental platform that is designed to explore the multi-dimensional moral dilemmas faced by autonomous vehicles.This platform enabled us to gather 40 million decisions in ten languages from over 2.3 million people in 233 countries and territories, and thus to assess the paths and obstacles to machine ethics.First, we summarize global moral preferences, some of them strong (e.g., sparing more lives or younger individuals), and some of them weak (e.g., sparing women, sparing pedestrians).Second, we document individual variations of these preferences, based on respondents' demographics (e.g., age, gender, income), and find that demographics do not change the direction of any preference.Third, we report cross-cultural ethical variation, and uncover three major clusters of countries exhibiting substantial differences along key moral preferences about autonomous vehicles.Fourth, we show that these differences correlate with modern institutions, but also with deep cultural traits.We discuss how these three layers of preferences can contribute to developing global, harmonious, and socially acceptable principles for machine ethics.All data used in this article can be accessed and downloaded at https://goo.gl/JXRrBP .We are entering an age in which machines are not only tasked to promote well-being and minimize harm, but also to distribute the well-being they create, and the harm they cannot eliminate.Distributing well-being and harm inevitably creates tradeoffs, whose resolution falls in the moral domain 1 , 2 , 3 .Think of an autonomous vehicle (AV) that is about to crash, and cannot find a trajectory that would save everyone.Should it swerve onto one jaywalking teenager to spare its three elderly passengers?Should it swerve away from a group of children and kill a single adult passenger instead?Even in the more common instances in which harm is not inevitable, but just possible, AVs will need to decide how to divvy up the risk of harm between the different stakeholders on the road.Car manufacturers and policymakers are currently struggling with these moral dilemmas, in large part because they cannot be solved by any simple normative ethical principles like Asimov's laws of robotics 4 . Asimov's laws were not designed to solve the problem of universal machine ethics, and they were not even designed to let machines distribute harm between humans.They were a narrative device whose goal was to generate good stories, by showcasing how challenging it is to create moral machines with a dozen lines of code.And yet, we do not have the luxury to give up on creating moral machines 5 , 6 , 7 , 8 .AVs will cruise our roads soon, necessitating agreement on the principles that should apply when, inevitably, life-threatening dilemmas emerge.The frequency at which these dilemma will emerge is extremely hard to estimate, just as it is extremely hard to estimate the rate at which human drivers find themselves in comparable situations.Human drivers who die in crashes cannot report whether they were faced with a dilemma; and human drivers who survive a crash may not have realized that they were in a dilemma situation.Note though that ethical guidelines for AV choices in dilemma situations do not depend on the frequency of these situations.Whether these cases are rare, very rare, or extremely rare, we need to agree beforehand on how they should be solved. The keyword here is "we".As emphasized by former U.S. president Barack Obama 9 , consensus in this matter is going to be important.Decisions about the ethical principles that will guide AVs cannot be left to solely to either the engineers or the ethicists.For consumers to switch from traditional human-driven cars to AVs, and for the wider public to accept the proliferation of AI-driven vehicles on their roads, both groups will need to understand the origins of the ethical principles programmed into these vehicles 10 .In other words, even if ethicists were to agree on how AVs should solve moral dilemmas, their work would be useless if citizens were to disagree with their solution, and thus opt out of the future that AVs promise in lieu of the status quo.Any attempt to devise AI ethics must be at least cognizant of public morality. Accordingly, we need to gauge social expectations about the way AVs should solve moral dilemmas.This enterprise, however, is not without challenges 11 .The first challenge comes from the high-dimensionality of the problem.In a typical survey, one may test whether people prefer to spare many lives rather than few 9,12 , 13 ; or whether people prefer to spare the young rather than the elderly 14 , 15 ; or whether people prefer to spare pedestrians who cross legally, rather than pedestrians who jaywalk; or yet some other preference, or a simple combination of two or three of these preferences.But combining a dozen of such preferences leads to millions of possible scenarios, requiring a sample size that defies any conventional method of data collection. · plain
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“the strongest preferences are observed for sparing humans over animals, sparing more lives, and sparing young lives.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > GLOBAL PREFERENCES
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“we observe some striking peculiarities, like the strong preference for sparing women and the strong preference for sparing fit characters in the Southern cluster.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > CULTURAL CLUSTERS
boostedcorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Rests on an assumption: The 130 countries with at least 100 respondents are representative of their countries' moral preferences.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“one increase in standard deviation of religiosity of respondent is associated with 0.09 more inclination to spare humans, 0.08 less inclination to spare the lawful, 0.06 more inclination to spare pedestrians, and 0.06 less inclination to spare the fit.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > INDIVIDUAL VARIATIONS
plaincorrelationaln = 492,921- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Association between the preference for sparing the lawful and each of rule of law and log of GDP per capita (inset).”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > METHODS
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Rests on an assumption: The country from which users access the website, geo-localized through IP address, is a valid proxy for the country-level moral preferences of that country.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Association between individualism and each of the preference for sparing more characters, and the preference for sparing the young (inset).”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > METHODS
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Rests on an assumption: The country from which users access the website, geo-localized through IP address, is a valid proxy for the country-level moral preferences of that country.
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“The Moral Machine was deployed to initiate such a conversation, and millions of people weighed in from around the world .”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > DISCUSSION
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“The frequency at which these dilemma will emerge is extremely hard to estimate”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > Abstract With the rapid development of Artificial Intelligence technology come widespread concerns about how machines will behave in morally charged situations.Addressing these concerns raises the major challenge of quantifying societal expectations about the ethical principles that should guide machine behavior.To address this challenge, we deployed the Moral Machine, an Internet-based experimental platform that is designed to explore the multi-dimensional moral dilemmas faced by autonomous vehicles.This platform enabled us to gather 40 million decisions in ten languages from over 2.3 million people in 233 countries and territories, and thus to assess the paths and obstacles to machine ethics.First, we summarize global moral preferences, some of them strong (e.g., sparing more lives or younger individuals), and some of them weak (e.g., sparing women, sparing pedestrians).Second, we document individual variations of these preferences, based on respondents' demographics (e.g., age, gender, income), and find that demographics do not change the direction of any preference.Third, we report cross-cultural ethical variation, and uncover three major clusters of countries exhibiting substantial differences along key moral preferences about autonomous vehicles.Fourth, we show that these differences correlate with modern institutions, but also with deep cultural traits.We discuss how these three layers of preferences can contribute to developing global, harmonious, and socially acceptable principles for machine ethics.All data used in this article can be accessed and downloaded at https://goo.gl/JXRrBP .We are entering an age in which machines are not only tasked to promote well-being and minimize harm, but also to distribute the well-being they create, and the harm they cannot eliminate.Distributing well-being and harm inevitably creates tradeoffs, whose resolution falls in the moral domain 1 , 2 , 3 .Think of an autonomous vehicle (AV) that is about to crash, and cannot find a trajectory that would save everyone.Should it swerve onto one jaywalking teenager to spare its three elderly passengers?Should it swerve away from a group of children and kill a single adult passenger instead?Even in the more common instances in which harm is not inevitable, but just possible, AVs will need to decide how to divvy up the risk of harm between the different stakeholders on the road.Car manufacturers and policymakers are currently struggling with these moral dilemmas, in large part because they cannot be solved by any simple normative ethical principles like Asimov's laws of robotics 4 . Asimov's laws were not designed to solve the problem of universal machine ethics, and they were not even designed to let machines distribute harm between humans.They were a narrative device whose goal was to generate good stories, by showcasing how challenging it is to create moral machines with a dozen lines of code.And yet, we do not have the luxury to give up on creating moral machines 5 , 6 , 7 , 8 .AVs will cruise our roads soon, necessitating agreement on the principles that should apply when, inevitably, life-threatening dilemmas emerge.The frequency at which these dilemma will emerge is extremely hard to estimate, just as it is extremely hard to estimate the rate at which human drivers find themselves in comparable situations.Human drivers who die in crashes cannot report whether they were faced with a dilemma; and human drivers who survive a crash may not have realized that they were in a dilemma situation.Note though that ethical guidelines for AV choices in dilemma situations do not depend on the frequency of these situations.Whether these cases are rare, very rare, or extremely rare, we need to agree beforehand on how they should be solved. The keyword here is "we".As emphasized by former U.S. president Barack Obama 9 , consensus in this matter is going to be important.Decisions about the ethical principles that will guide AVs cannot be left to solely to either the engineers or the ethicists.For consumers to switch from traditional human-driven cars to AVs, and for the wider public to accept the proliferation of AI-driven vehicles on their roads, both groups will need to understand the origins of the ethical principles programmed into these vehicles 10 .In other words, even if ethicists were to agree on how AVs should solve moral dilemmas, their work would be useless if citizens were to disagree with their solution, and thus opt out of the future that AVs promise in lieu of the status quo.Any attempt to devise AI ethics must be at least cognizant of public morality. Accordingly, we need to gauge social expectations about the way AVs should solve moral dilemmas.This enterprise, however, is not without challenges 11 .The first challenge comes from the high-dimensionality of the problem.In a typical survey, one may test whether people prefer to spare many lives rather than few 9,12 , 13 ; or whether people prefer to spare the young rather than the elderly 14 , 15 ; or whether people prefer to spare pedestrians who cross legally, rather than pedestrians who jaywalk; or yet some other preference, or a simple combination of two or three of these preferences.But combining a dozen of such preferences leads to millions of possible scenarios, requiring a sample size that defies any conventional method of data collection.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
In the paper
“German Ethical Rule #9 does not take a clear stance on whether and when AVs should be programmed to sacrifice the few to spare the many, but leaves this possibility open”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > GLOBAL PREFERENCES
What the sources say
- Not checked — no full text of the sourcecites [19] Luetge, C. The German Ethics Code for automated and connected driving. Philos. Technol. 30, 547–558 (2017). source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
In the paper
“German Ethical Rule #7 unambiguously states that in dilemma situations, the protection of human life should enjoy top priority over the protection of other animal life.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > GLOBAL PREFERENCES
What the sources say
- Not checked — no full text of the sourcecites [19] Luetge, C. The German Ethics Code for automated and connected driving. Philos. Technol. 30, 547–558 (2017). source
boosted- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Accordingly, these three preferences may be considered essential building blocks for machine ethics, or at least essential topics to be considered by policymakers.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > GLOBAL PREFERENCES
hedged- Asserted: passes — The authors put it forward as true.
- Checkable: fails — It may be too vague to check against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. It may be too vague to check against evidence. It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“AVs will cruise our roads soon, necessitating agreement on the principles that should apply when, inevitably, life-threatening dilemmas emerge.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > Abstract With the rapid development of Artificial Intelligence technology come widespread concerns about how machines will behave in morally charged situations.Addressing these concerns raises the major challenge of quantifying societal expectations about the ethical principles that should guide machine behavior.To address this challenge, we deployed the Moral Machine, an Internet-based experimental platform that is designed to explore the multi-dimensional moral dilemmas faced by autonomous vehicles.This platform enabled us to gather 40 million decisions in ten languages from over 2.3 million people in 233 countries and territories, and thus to assess the paths and obstacles to machine ethics.First, we summarize global moral preferences, some of them strong (e.g., sparing more lives or younger individuals), and some of them weak (e.g., sparing women, sparing pedestrians).Second, we document individual variations of these preferences, based on respondents' demographics (e.g., age, gender, income), and find that demographics do not change the direction of any preference.Third, we report cross-cultural ethical variation, and uncover three major clusters of countries exhibiting substantial differences along key moral preferences about autonomous vehicles.Fourth, we show that these differences correlate with modern institutions, but also with deep cultural traits.We discuss how these three layers of preferences can contribute to developing global, harmonious, and socially acceptable principles for machine ethics.All data used in this article can be accessed and downloaded at https://goo.gl/JXRrBP .We are entering an age in which machines are not only tasked to promote well-being and minimize harm, but also to distribute the well-being they create, and the harm they cannot eliminate.Distributing well-being and harm inevitably creates tradeoffs, whose resolution falls in the moral domain 1 , 2 , 3 .Think of an autonomous vehicle (AV) that is about to crash, and cannot find a trajectory that would save everyone.Should it swerve onto one jaywalking teenager to spare its three elderly passengers?Should it swerve away from a group of children and kill a single adult passenger instead?Even in the more common instances in which harm is not inevitable, but just possible, AVs will need to decide how to divvy up the risk of harm between the different stakeholders on the road.Car manufacturers and policymakers are currently struggling with these moral dilemmas, in large part because they cannot be solved by any simple normative ethical principles like Asimov's laws of robotics 4 . Asimov's laws were not designed to solve the problem of universal machine ethics, and they were not even designed to let machines distribute harm between humans.They were a narrative device whose goal was to generate good stories, by showcasing how challenging it is to create moral machines with a dozen lines of code.And yet, we do not have the luxury to give up on creating moral machines 5 , 6 , 7 , 8 .AVs will cruise our roads soon, necessitating agreement on the principles that should apply when, inevitably, life-threatening dilemmas emerge.The frequency at which these dilemma will emerge is extremely hard to estimate, just as it is extremely hard to estimate the rate at which human drivers find themselves in comparable situations.Human drivers who die in crashes cannot report whether they were faced with a dilemma; and human drivers who survive a crash may not have realized that they were in a dilemma situation.Note though that ethical guidelines for AV choices in dilemma situations do not depend on the frequency of these situations.Whether these cases are rare, very rare, or extremely rare, we need to agree beforehand on how they should be solved. The keyword here is "we".As emphasized by former U.S. president Barack Obama 9 , consensus in this matter is going to be important.Decisions about the ethical principles that will guide AVs cannot be left to solely to either the engineers or the ethicists.For consumers to switch from traditional human-driven cars to AVs, and for the wider public to accept the proliferation of AI-driven vehicles on their roads, both groups will need to understand the origins of the ethical principles programmed into these vehicles 10 .In other words, even if ethicists were to agree on how AVs should solve moral dilemmas, their work would be useless if citizens were to disagree with their solution, and thus opt out of the future that AVs promise in lieu of the status quo.Any attempt to devise AI ethics must be at least cognizant of public morality. Accordingly, we need to gauge social expectations about the way AVs should solve moral dilemmas.This enterprise, however, is not without challenges 11 .The first challenge comes from the high-dimensionality of the problem.In a typical survey, one may test whether people prefer to spare many lives rather than few 9,12 , 13 ; or whether people prefer to spare the young rather than the elderly 14 , 15 ; or whether people prefer to spare pedestrians who cross legally, rather than pedestrians who jaywalk; or yet some other preference, or a simple combination of two or three of these preferences.But combining a dozen of such preferences leads to millions of possible scenarios, requiring a sample size that defies any conventional method of data collection.
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: fails — It may be too vague to check against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: fails — The paper may give no reason why it holds.
Probably not a substantial claim. It may be too vague to check against evidence. The paper may give no reason why it holds. It may state two separate claims as one.
What the paper did — it cites no source for this, so there is none to check. The quote shows it is in the paper.
In the paper
“The internal structure within this cluster also exhibits notable face validity, with a sub-cluster containing Protestant / Scandinavian countries, and a sub-cluster containing Commonwealth / English-speaking countries.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > CULTURAL CLUSTERS
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Rests on an assumption: The 130 countries with at least 100 respondents are representative of their countries' moral preferences.
In the paper
“consensus in this matter is going to be important.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > Abstract With the rapid development of Artificial Intelligence technology come widespread concerns about how machines will behave in morally charged situations.Addressing these concerns raises the major challenge of quantifying societal expectations about the ethical principles that should guide machine behavior.To address this challenge, we deployed the Moral Machine, an Internet-based experimental platform that is designed to explore the multi-dimensional moral dilemmas faced by autonomous vehicles.This platform enabled us to gather 40 million decisions in ten languages from over 2.3 million people in 233 countries and territories, and thus to assess the paths and obstacles to machine ethics.First, we summarize global moral preferences, some of them strong (e.g., sparing more lives or younger individuals), and some of them weak (e.g., sparing women, sparing pedestrians).Second, we document individual variations of these preferences, based on respondents' demographics (e.g., age, gender, income), and find that demographics do not change the direction of any preference.Third, we report cross-cultural ethical variation, and uncover three major clusters of countries exhibiting substantial differences along key moral preferences about autonomous vehicles.Fourth, we show that these differences correlate with modern institutions, but also with deep cultural traits.We discuss how these three layers of preferences can contribute to developing global, harmonious, and socially acceptable principles for machine ethics.All data used in this article can be accessed and downloaded at https://goo.gl/JXRrBP .We are entering an age in which machines are not only tasked to promote well-being and minimize harm, but also to distribute the well-being they create, and the harm they cannot eliminate.Distributing well-being and harm inevitably creates tradeoffs, whose resolution falls in the moral domain 1 , 2 , 3 .Think of an autonomous vehicle (AV) that is about to crash, and cannot find a trajectory that would save everyone.Should it swerve onto one jaywalking teenager to spare its three elderly passengers?Should it swerve away from a group of children and kill a single adult passenger instead?Even in the more common instances in which harm is not inevitable, but just possible, AVs will need to decide how to divvy up the risk of harm between the different stakeholders on the road.Car manufacturers and policymakers are currently struggling with these moral dilemmas, in large part because they cannot be solved by any simple normative ethical principles like Asimov's laws of robotics 4 . Asimov's laws were not designed to solve the problem of universal machine ethics, and they were not even designed to let machines distribute harm between humans.They were a narrative device whose goal was to generate good stories, by showcasing how challenging it is to create moral machines with a dozen lines of code.And yet, we do not have the luxury to give up on creating moral machines 5 , 6 , 7 , 8 .AVs will cruise our roads soon, necessitating agreement on the principles that should apply when, inevitably, life-threatening dilemmas emerge.The frequency at which these dilemma will emerge is extremely hard to estimate, just as it is extremely hard to estimate the rate at which human drivers find themselves in comparable situations.Human drivers who die in crashes cannot report whether they were faced with a dilemma; and human drivers who survive a crash may not have realized that they were in a dilemma situation.Note though that ethical guidelines for AV choices in dilemma situations do not depend on the frequency of these situations.Whether these cases are rare, very rare, or extremely rare, we need to agree beforehand on how they should be solved. The keyword here is "we".As emphasized by former U.S. president Barack Obama 9 , consensus in this matter is going to be important.Decisions about the ethical principles that will guide AVs cannot be left to solely to either the engineers or the ethicists.For consumers to switch from traditional human-driven cars to AVs, and for the wider public to accept the proliferation of AI-driven vehicles on their roads, both groups will need to understand the origins of the ethical principles programmed into these vehicles 10 .In other words, even if ethicists were to agree on how AVs should solve moral dilemmas, their work would be useless if citizens were to disagree with their solution, and thus opt out of the future that AVs promise in lieu of the status quo.Any attempt to devise AI ethics must be at least cognizant of public morality. Accordingly, we need to gauge social expectations about the way AVs should solve moral dilemmas.This enterprise, however, is not without challenges 11 .The first challenge comes from the high-dimensionality of the problem.In a typical survey, one may test whether people prefer to spare many lives rather than few 9,12 , 13 ; or whether people prefer to spare the young rather than the elderly 14 , 15 ; or whether people prefer to spare pedestrians who cross legally, rather than pedestrians who jaywalk; or yet some other preference, or a simple combination of two or three of these preferences.But combining a dozen of such preferences leads to millions of possible scenarios, requiring a sample size that defies any conventional method of data collection.
What the sources say
- Not matchedcites [9] Dadich, S. Barack Obama, neural nets, self-driving cars, and the future of the world. Wired https://www.wired.com/2016/10/president-obama-mit-joi-ito-interview/ (2016).
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: fails — It may be too vague to check against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. It may be too vague to check against evidence.
- Not matched
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“But combining a dozen of such preferences leads to millions of possible scenarios, requiring a sample size that defies any conventional method of data collection.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > Abstract With the rapid development of Artificial Intelligence technology come widespread concerns about how machines will behave in morally charged situations.Addressing these concerns raises the major challenge of quantifying societal expectations about the ethical principles that should guide machine behavior.To address this challenge, we deployed the Moral Machine, an Internet-based experimental platform that is designed to explore the multi-dimensional moral dilemmas faced by autonomous vehicles.This platform enabled us to gather 40 million decisions in ten languages from over 2.3 million people in 233 countries and territories, and thus to assess the paths and obstacles to machine ethics.First, we summarize global moral preferences, some of them strong (e.g., sparing more lives or younger individuals), and some of them weak (e.g., sparing women, sparing pedestrians).Second, we document individual variations of these preferences, based on respondents' demographics (e.g., age, gender, income), and find that demographics do not change the direction of any preference.Third, we report cross-cultural ethical variation, and uncover three major clusters of countries exhibiting substantial differences along key moral preferences about autonomous vehicles.Fourth, we show that these differences correlate with modern institutions, but also with deep cultural traits.We discuss how these three layers of preferences can contribute to developing global, harmonious, and socially acceptable principles for machine ethics.All data used in this article can be accessed and downloaded at https://goo.gl/JXRrBP .We are entering an age in which machines are not only tasked to promote well-being and minimize harm, but also to distribute the well-being they create, and the harm they cannot eliminate.Distributing well-being and harm inevitably creates tradeoffs, whose resolution falls in the moral domain 1 , 2 , 3 .Think of an autonomous vehicle (AV) that is about to crash, and cannot find a trajectory that would save everyone.Should it swerve onto one jaywalking teenager to spare its three elderly passengers?Should it swerve away from a group of children and kill a single adult passenger instead?Even in the more common instances in which harm is not inevitable, but just possible, AVs will need to decide how to divvy up the risk of harm between the different stakeholders on the road.Car manufacturers and policymakers are currently struggling with these moral dilemmas, in large part because they cannot be solved by any simple normative ethical principles like Asimov's laws of robotics 4 . Asimov's laws were not designed to solve the problem of universal machine ethics, and they were not even designed to let machines distribute harm between humans.They were a narrative device whose goal was to generate good stories, by showcasing how challenging it is to create moral machines with a dozen lines of code.And yet, we do not have the luxury to give up on creating moral machines 5 , 6 , 7 , 8 .AVs will cruise our roads soon, necessitating agreement on the principles that should apply when, inevitably, life-threatening dilemmas emerge.The frequency at which these dilemma will emerge is extremely hard to estimate, just as it is extremely hard to estimate the rate at which human drivers find themselves in comparable situations.Human drivers who die in crashes cannot report whether they were faced with a dilemma; and human drivers who survive a crash may not have realized that they were in a dilemma situation.Note though that ethical guidelines for AV choices in dilemma situations do not depend on the frequency of these situations.Whether these cases are rare, very rare, or extremely rare, we need to agree beforehand on how they should be solved. The keyword here is "we".As emphasized by former U.S. president Barack Obama 9 , consensus in this matter is going to be important.Decisions about the ethical principles that will guide AVs cannot be left to solely to either the engineers or the ethicists.For consumers to switch from traditional human-driven cars to AVs, and for the wider public to accept the proliferation of AI-driven vehicles on their roads, both groups will need to understand the origins of the ethical principles programmed into these vehicles 10 .In other words, even if ethicists were to agree on how AVs should solve moral dilemmas, their work would be useless if citizens were to disagree with their solution, and thus opt out of the future that AVs promise in lieu of the status quo.Any attempt to devise AI ethics must be at least cognizant of public morality. Accordingly, we need to gauge social expectations about the way AVs should solve moral dilemmas.This enterprise, however, is not without challenges 11 .The first challenge comes from the high-dimensionality of the problem.In a typical survey, one may test whether people prefer to spare many lives rather than few 9,12 , 13 ; or whether people prefer to spare the young rather than the elderly 14 , 15 ; or whether people prefer to spare pedestrians who cross legally, rather than pedestrians who jaywalk; or yet some other preference, or a simple combination of two or three of these preferences.But combining a dozen of such preferences leads to millions of possible scenarios, requiring a sample size that defies any conventional method of data collection.
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
In the paper
“German Ethical Rule #9 also states that any distinction based on personal features, such as age, should be prohibited.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > GLOBAL PREFERENCES
What the sources say
- Not checked — no full text of the sourcecites [19] Luetge, C. The German Ethics Code for automated and connected driving. Philos. Technol. 30, 547–558 (2017). source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim.
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“Asimov's laws were not designed to solve the problem of universal machine ethics, and they were not even designed to let machines distribute harm between humans.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > Abstract With the rapid development of Artificial Intelligence technology come widespread concerns about how machines will behave in morally charged situations.Addressing these concerns raises the major challenge of quantifying societal expectations about the ethical principles that should guide machine behavior.To address this challenge, we deployed the Moral Machine, an Internet-based experimental platform that is designed to explore the multi-dimensional moral dilemmas faced by autonomous vehicles.This platform enabled us to gather 40 million decisions in ten languages from over 2.3 million people in 233 countries and territories, and thus to assess the paths and obstacles to machine ethics.First, we summarize global moral preferences, some of them strong (e.g., sparing more lives or younger individuals), and some of them weak (e.g., sparing women, sparing pedestrians).Second, we document individual variations of these preferences, based on respondents' demographics (e.g., age, gender, income), and find that demographics do not change the direction of any preference.Third, we report cross-cultural ethical variation, and uncover three major clusters of countries exhibiting substantial differences along key moral preferences about autonomous vehicles.Fourth, we show that these differences correlate with modern institutions, but also with deep cultural traits.We discuss how these three layers of preferences can contribute to developing global, harmonious, and socially acceptable principles for machine ethics.All data used in this article can be accessed and downloaded at https://goo.gl/JXRrBP .We are entering an age in which machines are not only tasked to promote well-being and minimize harm, but also to distribute the well-being they create, and the harm they cannot eliminate.Distributing well-being and harm inevitably creates tradeoffs, whose resolution falls in the moral domain 1 , 2 , 3 .Think of an autonomous vehicle (AV) that is about to crash, and cannot find a trajectory that would save everyone.Should it swerve onto one jaywalking teenager to spare its three elderly passengers?Should it swerve away from a group of children and kill a single adult passenger instead?Even in the more common instances in which harm is not inevitable, but just possible, AVs will need to decide how to divvy up the risk of harm between the different stakeholders on the road.Car manufacturers and policymakers are currently struggling with these moral dilemmas, in large part because they cannot be solved by any simple normative ethical principles like Asimov's laws of robotics 4 . Asimov's laws were not designed to solve the problem of universal machine ethics, and they were not even designed to let machines distribute harm between humans.They were a narrative device whose goal was to generate good stories, by showcasing how challenging it is to create moral machines with a dozen lines of code.And yet, we do not have the luxury to give up on creating moral machines 5 , 6 , 7 , 8 .AVs will cruise our roads soon, necessitating agreement on the principles that should apply when, inevitably, life-threatening dilemmas emerge.The frequency at which these dilemma will emerge is extremely hard to estimate, just as it is extremely hard to estimate the rate at which human drivers find themselves in comparable situations.Human drivers who die in crashes cannot report whether they were faced with a dilemma; and human drivers who survive a crash may not have realized that they were in a dilemma situation.Note though that ethical guidelines for AV choices in dilemma situations do not depend on the frequency of these situations.Whether these cases are rare, very rare, or extremely rare, we need to agree beforehand on how they should be solved. The keyword here is "we".As emphasized by former U.S. president Barack Obama 9 , consensus in this matter is going to be important.Decisions about the ethical principles that will guide AVs cannot be left to solely to either the engineers or the ethicists.For consumers to switch from traditional human-driven cars to AVs, and for the wider public to accept the proliferation of AI-driven vehicles on their roads, both groups will need to understand the origins of the ethical principles programmed into these vehicles 10 .In other words, even if ethicists were to agree on how AVs should solve moral dilemmas, their work would be useless if citizens were to disagree with their solution, and thus opt out of the future that AVs promise in lieu of the status quo.Any attempt to devise AI ethics must be at least cognizant of public morality. Accordingly, we need to gauge social expectations about the way AVs should solve moral dilemmas.This enterprise, however, is not without challenges 11 .The first challenge comes from the high-dimensionality of the problem.In a typical survey, one may test whether people prefer to spare many lives rather than few 9,12 , 13 ; or whether people prefer to spare the young rather than the elderly 14 , 15 ; or whether people prefer to spare pedestrians who cross legally, rather than pedestrians who jaywalk; or yet some other preference, or a simple combination of two or three of these preferences.But combining a dozen of such preferences leads to millions of possible scenarios, requiring a sample size that defies any conventional method of data collection.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. It may state two separate claims as one.
None of the sources cited for this claim have been checked yet.
In the paper
“Distributing well-being and harm inevitably creates tradeoffs, whose resolution falls in the moral domain 1 , 2 , 3 .”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > Abstract With the rapid development of Artificial Intelligence technology come widespread concerns about how machines will behave in morally charged situations.Addressing these concerns raises the major challenge of quantifying societal expectations about the ethical principles that should guide machine behavior.To address this challenge, we deployed the Moral Machine, an Internet-based experimental platform that is designed to explore the multi-dimensional moral dilemmas faced by autonomous vehicles.This platform enabled us to gather 40 million decisions in ten languages from over 2.3 million people in 233 countries and territories, and thus to assess the paths and obstacles to machine ethics.First, we summarize global moral preferences, some of them strong (e.g., sparing more lives or younger individuals), and some of them weak (e.g., sparing women, sparing pedestrians).Second, we document individual variations of these preferences, based on respondents' demographics (e.g., age, gender, income), and find that demographics do not change the direction of any preference.Third, we report cross-cultural ethical variation, and uncover three major clusters of countries exhibiting substantial differences along key moral preferences about autonomous vehicles.Fourth, we show that these differences correlate with modern institutions, but also with deep cultural traits.We discuss how these three layers of preferences can contribute to developing global, harmonious, and socially acceptable principles for machine ethics.All data used in this article can be accessed and downloaded at https://goo.gl/JXRrBP .We are entering an age in which machines are not only tasked to promote well-being and minimize harm, but also to distribute the well-being they create, and the harm they cannot eliminate.Distributing well-being and harm inevitably creates tradeoffs, whose resolution falls in the moral domain 1 , 2 , 3 .Think of an autonomous vehicle (AV) that is about to crash, and cannot find a trajectory that would save everyone.Should it swerve onto one jaywalking teenager to spare its three elderly passengers?Should it swerve away from a group of children and kill a single adult passenger instead?Even in the more common instances in which harm is not inevitable, but just possible, AVs will need to decide how to divvy up the risk of harm between the different stakeholders on the road.Car manufacturers and policymakers are currently struggling with these moral dilemmas, in large part because they cannot be solved by any simple normative ethical principles like Asimov's laws of robotics 4 . Asimov's laws were not designed to solve the problem of universal machine ethics, and they were not even designed to let machines distribute harm between humans.They were a narrative device whose goal was to generate good stories, by showcasing how challenging it is to create moral machines with a dozen lines of code.And yet, we do not have the luxury to give up on creating moral machines 5 , 6 , 7 , 8 .AVs will cruise our roads soon, necessitating agreement on the principles that should apply when, inevitably, life-threatening dilemmas emerge.The frequency at which these dilemma will emerge is extremely hard to estimate, just as it is extremely hard to estimate the rate at which human drivers find themselves in comparable situations.Human drivers who die in crashes cannot report whether they were faced with a dilemma; and human drivers who survive a crash may not have realized that they were in a dilemma situation.Note though that ethical guidelines for AV choices in dilemma situations do not depend on the frequency of these situations.Whether these cases are rare, very rare, or extremely rare, we need to agree beforehand on how they should be solved. The keyword here is "we".As emphasized by former U.S. president Barack Obama 9 , consensus in this matter is going to be important.Decisions about the ethical principles that will guide AVs cannot be left to solely to either the engineers or the ethicists.For consumers to switch from traditional human-driven cars to AVs, and for the wider public to accept the proliferation of AI-driven vehicles on their roads, both groups will need to understand the origins of the ethical principles programmed into these vehicles 10 .In other words, even if ethicists were to agree on how AVs should solve moral dilemmas, their work would be useless if citizens were to disagree with their solution, and thus opt out of the future that AVs promise in lieu of the status quo.Any attempt to devise AI ethics must be at least cognizant of public morality. Accordingly, we need to gauge social expectations about the way AVs should solve moral dilemmas.This enterprise, however, is not without challenges 11 .The first challenge comes from the high-dimensionality of the problem.In a typical survey, one may test whether people prefer to spare many lives rather than few 9,12 , 13 ; or whether people prefer to spare the young rather than the elderly 14 , 15 ; or whether people prefer to spare pedestrians who cross legally, rather than pedestrians who jaywalk; or yet some other preference, or a simple combination of two or three of these preferences.But combining a dozen of such preferences leads to millions of possible scenarios, requiring a sample size that defies any conventional method of data collection.
What the sources say
- Not matchedcites [1] Greene, J. Moral Tribes: Emotion, Reason and the Gap Between Us and Them (Atlantic Books, London, 2013).
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Not checked — no full text of the sourcecites [2] Tomasello, M. A Natural History of Human Thinking (Harvard Univ. Press, Cambridge, 2014). source
- Not checked — no full text of the sourcecites [3] Cushman, F. & Young, L. The psychology of dilemmas and the philosophy of morality. Ethical Theory Moral Pract. 12, 9–24 (2009). source
boosteddirect- Asserted: passes — The authors put it forward as true.
- Checkable: fails — It may be too vague to check against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. It may be too vague to check against evidence.
- Not matched
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“Note though that ethical guidelines for AV choices in dilemma situations do not depend on the frequency of these situations.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > Abstract With the rapid development of Artificial Intelligence technology come widespread concerns about how machines will behave in morally charged situations.Addressing these concerns raises the major challenge of quantifying societal expectations about the ethical principles that should guide machine behavior.To address this challenge, we deployed the Moral Machine, an Internet-based experimental platform that is designed to explore the multi-dimensional moral dilemmas faced by autonomous vehicles.This platform enabled us to gather 40 million decisions in ten languages from over 2.3 million people in 233 countries and territories, and thus to assess the paths and obstacles to machine ethics.First, we summarize global moral preferences, some of them strong (e.g., sparing more lives or younger individuals), and some of them weak (e.g., sparing women, sparing pedestrians).Second, we document individual variations of these preferences, based on respondents' demographics (e.g., age, gender, income), and find that demographics do not change the direction of any preference.Third, we report cross-cultural ethical variation, and uncover three major clusters of countries exhibiting substantial differences along key moral preferences about autonomous vehicles.Fourth, we show that these differences correlate with modern institutions, but also with deep cultural traits.We discuss how these three layers of preferences can contribute to developing global, harmonious, and socially acceptable principles for machine ethics.All data used in this article can be accessed and downloaded at https://goo.gl/JXRrBP .We are entering an age in which machines are not only tasked to promote well-being and minimize harm, but also to distribute the well-being they create, and the harm they cannot eliminate.Distributing well-being and harm inevitably creates tradeoffs, whose resolution falls in the moral domain 1 , 2 , 3 .Think of an autonomous vehicle (AV) that is about to crash, and cannot find a trajectory that would save everyone.Should it swerve onto one jaywalking teenager to spare its three elderly passengers?Should it swerve away from a group of children and kill a single adult passenger instead?Even in the more common instances in which harm is not inevitable, but just possible, AVs will need to decide how to divvy up the risk of harm between the different stakeholders on the road.Car manufacturers and policymakers are currently struggling with these moral dilemmas, in large part because they cannot be solved by any simple normative ethical principles like Asimov's laws of robotics 4 . Asimov's laws were not designed to solve the problem of universal machine ethics, and they were not even designed to let machines distribute harm between humans.They were a narrative device whose goal was to generate good stories, by showcasing how challenging it is to create moral machines with a dozen lines of code.And yet, we do not have the luxury to give up on creating moral machines 5 , 6 , 7 , 8 .AVs will cruise our roads soon, necessitating agreement on the principles that should apply when, inevitably, life-threatening dilemmas emerge.The frequency at which these dilemma will emerge is extremely hard to estimate, just as it is extremely hard to estimate the rate at which human drivers find themselves in comparable situations.Human drivers who die in crashes cannot report whether they were faced with a dilemma; and human drivers who survive a crash may not have realized that they were in a dilemma situation.Note though that ethical guidelines for AV choices in dilemma situations do not depend on the frequency of these situations.Whether these cases are rare, very rare, or extremely rare, we need to agree beforehand on how they should be solved. The keyword here is "we".As emphasized by former U.S. president Barack Obama 9 , consensus in this matter is going to be important.Decisions about the ethical principles that will guide AVs cannot be left to solely to either the engineers or the ethicists.For consumers to switch from traditional human-driven cars to AVs, and for the wider public to accept the proliferation of AI-driven vehicles on their roads, both groups will need to understand the origins of the ethical principles programmed into these vehicles 10 .In other words, even if ethicists were to agree on how AVs should solve moral dilemmas, their work would be useless if citizens were to disagree with their solution, and thus opt out of the future that AVs promise in lieu of the status quo.Any attempt to devise AI ethics must be at least cognizant of public morality. Accordingly, we need to gauge social expectations about the way AVs should solve moral dilemmas.This enterprise, however, is not without challenges 11 .The first challenge comes from the high-dimensionality of the problem.In a typical survey, one may test whether people prefer to spare many lives rather than few 9,12 , 13 ; or whether people prefer to spare the young rather than the elderly 14 , 15 ; or whether people prefer to spare pedestrians who cross legally, rather than pedestrians who jaywalk; or yet some other preference, or a simple combination of two or three of these preferences.But combining a dozen of such preferences leads to millions of possible scenarios, requiring a sample size that defies any conventional method of data collection.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: fails — The paper may give no reason why it holds.
Probably not a substantial claim. The paper may give no reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“even if ethicists were to agree on how AVs should solve moral dilemmas, their work would be useless if citizens were to disagree with their solution, and thus opt out of the future that AVs promise in lieu of the status quo.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > Abstract With the rapid development of Artificial Intelligence technology come widespread concerns about how machines will behave in morally charged situations.Addressing these concerns raises the major challenge of quantifying societal expectations about the ethical principles that should guide machine behavior.To address this challenge, we deployed the Moral Machine, an Internet-based experimental platform that is designed to explore the multi-dimensional moral dilemmas faced by autonomous vehicles.This platform enabled us to gather 40 million decisions in ten languages from over 2.3 million people in 233 countries and territories, and thus to assess the paths and obstacles to machine ethics.First, we summarize global moral preferences, some of them strong (e.g., sparing more lives or younger individuals), and some of them weak (e.g., sparing women, sparing pedestrians).Second, we document individual variations of these preferences, based on respondents' demographics (e.g., age, gender, income), and find that demographics do not change the direction of any preference.Third, we report cross-cultural ethical variation, and uncover three major clusters of countries exhibiting substantial differences along key moral preferences about autonomous vehicles.Fourth, we show that these differences correlate with modern institutions, but also with deep cultural traits.We discuss how these three layers of preferences can contribute to developing global, harmonious, and socially acceptable principles for machine ethics.All data used in this article can be accessed and downloaded at https://goo.gl/JXRrBP .We are entering an age in which machines are not only tasked to promote well-being and minimize harm, but also to distribute the well-being they create, and the harm they cannot eliminate.Distributing well-being and harm inevitably creates tradeoffs, whose resolution falls in the moral domain 1 , 2 , 3 .Think of an autonomous vehicle (AV) that is about to crash, and cannot find a trajectory that would save everyone.Should it swerve onto one jaywalking teenager to spare its three elderly passengers?Should it swerve away from a group of children and kill a single adult passenger instead?Even in the more common instances in which harm is not inevitable, but just possible, AVs will need to decide how to divvy up the risk of harm between the different stakeholders on the road.Car manufacturers and policymakers are currently struggling with these moral dilemmas, in large part because they cannot be solved by any simple normative ethical principles like Asimov's laws of robotics 4 . Asimov's laws were not designed to solve the problem of universal machine ethics, and they were not even designed to let machines distribute harm between humans.They were a narrative device whose goal was to generate good stories, by showcasing how challenging it is to create moral machines with a dozen lines of code.And yet, we do not have the luxury to give up on creating moral machines 5 , 6 , 7 , 8 .AVs will cruise our roads soon, necessitating agreement on the principles that should apply when, inevitably, life-threatening dilemmas emerge.The frequency at which these dilemma will emerge is extremely hard to estimate, just as it is extremely hard to estimate the rate at which human drivers find themselves in comparable situations.Human drivers who die in crashes cannot report whether they were faced with a dilemma; and human drivers who survive a crash may not have realized that they were in a dilemma situation.Note though that ethical guidelines for AV choices in dilemma situations do not depend on the frequency of these situations.Whether these cases are rare, very rare, or extremely rare, we need to agree beforehand on how they should be solved. The keyword here is "we".As emphasized by former U.S. president Barack Obama 9 , consensus in this matter is going to be important.Decisions about the ethical principles that will guide AVs cannot be left to solely to either the engineers or the ethicists.For consumers to switch from traditional human-driven cars to AVs, and for the wider public to accept the proliferation of AI-driven vehicles on their roads, both groups will need to understand the origins of the ethical principles programmed into these vehicles 10 .In other words, even if ethicists were to agree on how AVs should solve moral dilemmas, their work would be useless if citizens were to disagree with their solution, and thus opt out of the future that AVs promise in lieu of the status quo.Any attempt to devise AI ethics must be at least cognizant of public morality. Accordingly, we need to gauge social expectations about the way AVs should solve moral dilemmas.This enterprise, however, is not without challenges 11 .The first challenge comes from the high-dimensionality of the problem.In a typical survey, one may test whether people prefer to spare many lives rather than few 9,12 , 13 ; or whether people prefer to spare the young rather than the elderly 14 , 15 ; or whether people prefer to spare pedestrians who cross legally, rather than pedestrians who jaywalk; or yet some other preference, or a simple combination of two or three of these preferences.But combining a dozen of such preferences leads to millions of possible scenarios, requiring a sample size that defies any conventional method of data collection.
hedgeddirect- Asserted: passes — The authors put it forward as true.
- Checkable: fails — It may be too vague to check against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. It may be too vague to check against evidence.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In sum, the individual variations we observe are theoretically important, but not essential information for policymakers.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > INDIVIDUAL VARIATIONS
plain- Asserted: passes — The authors put it forward as true.
- Checkable: unclear — It is unclear whether it could be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
In the paper
“Attempts at establishing broad ethical codes for intelligent machines, like the Asilomar AI Principle s 30 , often recommend that machine ethics should be aligned with human values.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > DISCUSSION
What the sources say
- Not matchedcites [30] Haidt, J. The Righteous Mind: Why Good People Are Divided by Politics and Religion (Knopf Doubleday, New York, 2012).
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
- Not matched
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“But the fact that broad regions of the world displayed relative agreement suggests that our journey to consensual machine ethics is not doomed from the start.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > DISCUSSION
hedgedcorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim.
What the paper did — it cites no source for this, so there is none to check. The quote shows it is in the paper.
In the paper
“we did not introduce uncertainty about the fates of the characters, and we did not introduce any uncertainty about the classification of these characters.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > DISCUSSION
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
In the paper
“our project builds on a long tradition of investigating cultural variations in ethical judgments 29 .”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > DISCUSSION
What the sources say
- Not matchedcites [29] Future of Life Institute. Asilomar AI Principles https://futureoflife.org/ai-principles/ (2017).
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
- Not matched
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“what the Moral Machine captured is the state of moral preferences before AVs are deployed on the road, and before they start facing moral dilemmas.”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > DISCUSSION
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. It may state two separate claims as one.
What the paper did — it cites no source for this, so there is none to check. The quote shows it is in the paper.
In the paper
“its downside is that our sample is self-selected, and not guaranteed to exactly match the socio-demographics of each country.”
- Also stated in: § DISCUSSION · plain
“But the fact that our samples are not guaranteed to be representative”
§ The Moral Machine Experiment: 40 Million Decisions and the Path to Universal Machine Ethics > DISCUSSION
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Also stated in: § DISCUSSION · plain
The paper relies on this without support.
The 130 countries with at least 100 respondents are representative of their countries' moral preferences.
“We selected the 130 countries with at least 100 respondents (N range = [101 -448,125])”
The paper relies on this without support.
No research ever attempted to measure moral preferences using a 9-dimensional experimental design, in more than 200 countries.
“No research ever attempted to measure moral preferences using a 9-dimensional experimental design, in more than 200 countries.”
The authors state this as an assumption.
In the Moral Machine scenarios, characters were recognized as adults, children, etc. with 100% certainty, and life-and-death outcomes were predicted with 100% certainty.
“In our scenarios, characters were recognized as adults, children, etc. with 100% certainty, and life-and-death outcomes were predicted with 100% certainty.”
“These assumptions are technologically unrealistic, but they were necessary to keep the project tractable.”
The paper relies on this without support.
The country from which users access the website, geo-localized through IP address, is a valid proxy for the country-level moral preferences of that country.
“The country from which users access the website is geo-localized through the IP address of their computer or mobile device.This information is used to compute a vector of moral preferences for each country.”
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“These results by contrast clash with many rules such as the IEEE code of ethics and some of the German ethical rules that clearly prohibit any distinction based on personal features such as age. However, results of the moral machine show that people strongly preferred sparing the young (such as children).”
From Public acceptance and perception of autonomous vehicles: a comprehensive review (2021) · cites this paper as [153]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“At this level of abstraction meaningful guidance may be impossible.”
From Principles alone cannot guarantee ethical AI (2019) · cites this paper as [44]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“The platform gathered the opinion of millions of people from 233 countries with 40 million decisions in ten languages.”
From Public acceptance and perception of autonomous vehicles: a comprehensive review (2021) · cites this paper as [153]
The cited source says what the claim attributes to it.
In the citing paper
“However, results show broad differences in relative preferences among different countries [153]. However, results show countries within close proximity to one another showed closer moral preferences, with three dominant clusters in the West, East, and South [154].”
From Public acceptance and perception of autonomous vehicles: a comprehensive review (2021) · cites this paper as [153]
The cited source says what the claim attributes to it.
In the citing paper
“Generally, results of the moral machine show that people prefer sparing humans over animals, sparing more lives, and sparing young lives [153].”
From Public acceptance and perception of autonomous vehicles: a comprehensive review (2021) · cites this paper as [153]
The cited source says what the claim attributes to it.
In the citing paper
“Previous work has shown that cultural and geographical differences play a key role in perceptions towards algorithmic systems [10,49,97].”
From Disentangling Fairness Perceptions in Algorithmic Decision-Making: the Effects of Explanations, Human Oversight, and Contestability (2023) · cites this paper as [10]
The cited source says what the claim attributes to it.
In the citing paper
“While some people are willing to sacrifice the elderly to save the young 169 , there are cultural differences on this preference 170 .”
From Using social and behavioural science to support COVID-19 pandemic response (2020) · cites this paper as [170]
The cited source says what the claim attributes to it.
In the citing paper
“Empirical work such as the Moral Machine Experiment has demonstrated that ethical preferences vary widely across cultures [8].”
From Ethics-Aware Safe Reinforcement Learning for Rare-Event Risk Control in Interactive Urban Driving (2026) · cites this paper as [8]
In the citing paper
“Ethics should ensure the design of AI technologies for human flourishing around the world [192,193], but this is a very complex matter given that, as the work of Awad et al. [194,195] clearly demonstrates, universally valid and accepted ethical principles do not exist.”
From The sustainability of artificial intelligence: an urbanistic viewpoint from the lens of smart and sustainable cities (2020) · cites this paper as [194]
In the citing paper
“Humans have a wide range of culturally diverse beliefs about right and wrong, and aligning AI systems to human values and preferences is challenging even within narrow domains such as automated driving (151).”
From The impact of generative artificial intelligence on socioeconomic inequalities and policy making (2024) · cites this paper as [151]
Citations
Cites
The Worldwide Governance Indicators: Methodology and Analytical Issues
Kaufmann, D., Kraay, A., & Mastruzzi, M. (2011) · Hague Journal on the Rule of Law · doi:10.1017/s1876404511200046 · Cited by 5,621
The righteous mind: why good people are divided by politics and religion
(2012) · Choice Reviews Online · doi:10.5860/choice.49-7185 · Cited by 4,347
(2005) · Cited by 3,670
In Search of Homo Economicus: Behavioral Experiments in 15 Small-Scale Societies
Henrich, J., Boyd, R. T., Bowles, S., Camerer, C., Fehr, E., Gintis, H. M., & McElreath, R. L. (2001) · American Economic Review · doi:10.1257/aer.91.2.73 · Cited by 2,619
Self-citation(2016) · Cited by 2,334
Hainmueller, J., Hopkins, D. J., & Yamamoto, T. (2014) · Political Analysis · doi:10.1093/pan/mpt024 · Cited by 1,990
A Natural History of Human Thinking
Tomasello, M. (2014) · doi:10.4159/9780674726369 · Cited by 1,923
The worldwide governance indicators : methodology and analytical issues
Kaufmann, D., Kraay, A., & Mastruzzi, M. (2010) · doi:10.1596/1813-9450-5430 · Cited by 1,552
The social dilemma of autonomous vehicles
Bonnefon, J.-F., Shariff, A., & Rahwan, I. (2016) · Science · doi:10.1126/science.aaf2654 · Cited by 1,527
Self-citationMoral tribes: emotion, reason, and the gap between us and them
(2014) · Choice Reviews Online · doi:10.5860/choice.51-4949 · Cited by 978
Cited by
Sorted by citations
Items 101–200 of 1,864
Collective Reflective Equilibrium in Practice (CREP) and controversial novel technologies
Savulescu, J., Gyngell, C., & Kahane, G. (2021) · Bioethics · doi:10.1111/bioe.12869 · Cited by 94
Explanations and trust: What happens to trust when a robot partner does something unexpected?
Lyons, J. B., Hamdan, I. A., & Vo, T. Q. (2022) · Computers in Human Behavior · doi:10.1016/j.chb.2022.107473 · Cited by 102
Public attitudes value interpretability but prioritize accuracy in Artificial Intelligence
Nußberger, A.-M., Luo, L., Celis, L. E., & Crockett, M. J. (2022) · Nature Communications · doi:10.1038/s41467-022-33417-3 · Cited by 105
“…for future work, especially in light of recent work suggesting that expectations towards machine-made decisions can vary substantially across countries and cultures 44 and amidst reports about the potential of disproportionately harmful impacts of AI on the lives of low-income populations 45 .”
BackgroundEthics in human–AI teaming: principles and perspectives
Pflanzer, M., Traylor, Z., Lyons, J. B., Dubljević, V., & Nam, C. S. (2022) · AI and Ethics · doi:10.1007/s43681-022-00214-z · Cited by 106
“…with these issues in the domain of autonomous vehicles, among other AI-based systems, “… for the wider public to accept the proliferation of AI-driven vehicles on their roads, both groups will need to understand the origins of the ethical principles that are programmed into these vehicles” [24].”
BackgroundAlmaatouq, A., Griffiths, T. L., Suchow, J. W., Whiting, M. E., Evans, J., & Watts, D. J. (2022) · Behavioral and Brain Sciences · doi:10.1017/s0140525x22002874 · Cited by 134
“…perception example where Srinivasan & Martinez, 2018 use a computer vision algorithm to annotate action units in the internet images; Yitzhak et al., 2017; Benitez-Quiroz et al., 2016), citizen science (Awad et al, 2018, 2020; Hilton & Mehr, 2021), and gamification (Long et al., 2023 Nature).”
Theorizing Moral Cognition: Culture in Action, Situations, and Relationships
Luft, A. (2020) · Socius Sociological Research for a Dynamic World · doi:10.1177/2378023120916125 · Cited by 95
“Slightly more than a year ago, a paper was published in Nature reporting the results of a study of self-driving cars (Awad et al. 2018).13 Developed by the MIT Media Lab, the purpose of the study was to crowd-source individuals’ decisions on how cars should prioritize lives in potentially dangerous…”
BackgroundMethodImpacts of Autonomous Vehicles on Public Health: A Conceptual Model and Policy Recommendations
Sohrabi, S., Khreis, H., & Lord, D. (2020) · Sustainable Cities and Society · doi:10.1016/j.scs.2020.102457 · Cited by 99
Artificial intelligence in government: Concepts, standards, and a unified framework
Straub, V. J., Morgan, D., Bright, J., & Margetts, H. Z. (2023) · Government Information Quarterly · doi:10.1016/j.giq.2023.101881 · Cited by 154
Baumer, E. P. S., Blythe, M., & Tanenbaum, T. J. (2020) · ACM Designing Interactive Systems Conference (DIS) · doi:10.1145/3357236.3395464 · Cited by 88
How Culture Shapes What People Want From AI
Ge, X., Xu, C., Misaki, D., Markus, H. R., & Tsai, J. L. (2024) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642660 · Cited by 121
“In other words, in these contexts with independent models of self, artificial agents can play an active role in human social life rather than operate in the background [7, 74].”
Accident analysis and prevention
Guastello, S. J. (2003) · PsycEXTRA Dataset · doi:10.1037/e483012008-001 · Cited by 86
Algorithms for Ethical Decision-Making in the Clinic: A Proof of Concept
Meier, L. J., Hein, A., Diepold, K. J., & Buyx, A. (2022) · The American Journal of Bioethics · doi:10.1080/15265161.2022.2040647 · Cited by 98
Ethical Decision Making in Autonomous Vehicles: The AV Ethics Project
Evans, K., de Moura, N., Chauvier, S., Chatila, R., & Dogan, E. (2020) · Science and Engineering Ethics · doi:10.1007/s11948-020-00272-8 · Cited by 82
Human-centred artificial intelligence: a contextual morality perspective
van Berkel, N., Tag, B., Gonçalves, J., & Hosio, S. J. (2020) · Behaviour and Information Technology · doi:10.1080/0144929x.2020.1818828 · Cited by 90
Moderator Chatbot for Deliberative Discussion
Kim, S., Eun, J., Seering, J., & Lee, J. (2021) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3449161 · Cited by 111
“Participants discussed two ethical dilemmas: the moral machine dilemma of self-driving cars [3] and the self-aware AI dilemma [71].”
BackgroundEffect of Information Presentation on Fairness Perceptions of Machine Learning Predictors
van Berkel, N., Goncalves, J., Russo, D., Hosio, S., & Skov, M. B. (2021) · Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3411764.3445365 · Cited by 84
Using Social and Behavioural Science to Support COVID-19 Pandemic ResponsePreprint
Bavel, J. J. V., D, K. B. A., Boggio, P. S., Capraro, V., Cichocka, A. K., Cikara, M., Crockett, M. J., Crum, A. J., Douglas, K. M., Druckman, J. N., et al. (2022) · SSRN Electronic Journal · doi:10.2139/ssrn.4178356 · Cited by 89
Self-citationAI Ethics – Too Principled to Fail?
Mittelstadt, B. (2019) · SSRN Electronic Journal · doi:10.2139/ssrn.3391293 · Cited by 82
“39,40 The great diversity of stakeholders and interests involved necessarily pushes the search for common values and norms towards a high level of abstraction.”
On doing hydrology with dragons: Realizing the value of perceptual models and knowledge accumulation
Wagener, T., Gleeson, T., Coxon, G., Hartmann, A., Howden, N. J. K., Pianosi, F., Rahman, M., Rosolem, R., Stein, L., & Woods, R. A. (2021) · Wiley Interdisciplinary Reviews Water · doi:10.1002/wat2.1550 · Cited by 97
“The Moral Machine online experiment collected 40 million decisions from people in 233 countries and territories (Awad et al., 2018).”
BackgroundA new sociology of humans and machines
Tsvetkova, M., Yasseri, T., Pescetelli, N., & Werner, T. (2024) · Nature Human Behaviour · doi:10.1038/s41562-024-02001-8 · Cited by 122
“People’s perceptions of machines vary by age, environment, personality and geography 46,170 .”
Bridging Machine Learning and Mechanism Design towards Algorithmic Fairness
Finocchiaro, J., Maio, R., Monachou, F., Patro, G. K., Raghavan, M., Stoica, A.-A., & Tsirtsis, S. (2021) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3442188.3445912 · Cited by 89
Qian, L., Yin, J., Huang, Y., & Liang, Y. (2022) · Technological Forecasting and Social Change · doi:10.1016/j.techfore.2022.122267 · Cited by 84
Telkamp, J. B., & Anderson, M. H. (2022) · Journal of Business Ethics · doi:10.1007/s10551-022-05057-6 · Cited by 89
“For example, Awad et al. (2018) conducted the “moral machine experiment,” which asked participants their opinions regarding who to hit if a self-driving car needs to hit someone, based on such criteria as age, gender, status, etc.”
BackgroundThe dark side of AI identity: Investigating when and why AI identity entitles unethical behavior
Cao, L., Chen, C., Dong, X., Wang, M., & Qin, X. (2023) · Computers in Human Behavior · doi:10.1016/j.chb.2023.107669 · Cited by 90
Using games to understand the mind
Allen, K. R., Brändle, F., Botvinick, M., Fan, J. E., Gershman, S. J., Alison Gopnik, C., Griffiths, T. L., Hartshorne, J. K., Hauser, T. U., Ho, M. K., et al. (2024) · Nature Human Behaviour · doi:10.1038/s41562-024-01878-9 · Cited by 130
“Making tasks intrinsically rewarding can also lead to an increase in the amount of collected data 35 as well as the diversity of participants 36–39 , including specific populations such as people born with missing limbs or patients with epilepsy 40,41 .”
Self-citationThe Gradient of Generative AI Release: Methods and Considerations
Solaiman, I. (2023) · 2023 ACM Conference on Fairness, Accountability, and Transparency · doi:10.1145/3593013.3593981 · Cited by 90
“Moral experiments show varying approaches to ethical problems by background and culture [3], which are urgently needed perspectives in building, evaluating, and deploying new AI systems.”
BackgroundHow inferred motives shape moral judgements
Carlson, R. W., Bigman, Y., Gray, K. J., Ferguson, M. J., & Crockett, M. J. (2022) · Nature Reviews Psychology · doi:10.1038/s44159-022-00071-x · Cited by 88
Moral Uncanny Valley: A Robot’s Appearance Moderates How its Decisions are Judged
Laakasuo, M., Palomäki, J., & Köbis, N. (2021) · International Journal of Social Robotics · doi:10.1007/s12369-020-00738-6 · Cited by 69
Artificial intelligence and human rights: a business ethical assessment
KRIEBITZ, A., & LÜTGE, C. (2020) · Business and Human Rights Journal · doi:10.1017/bhj.2019.28 · Cited by 76
Stapleton, L., Lee, M. H., Qing, D., Wright, M., Chouldechova, A., Holstein, K., Wu, Z. S., & Zhu, H. (2022) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3531146.3533177 · Cited by 96
Sloane, M., & Moss, E. (2019) · Nature Machine Intelligence · doi:10.1038/s42256-019-0084-6 · Cited by 76
Cunneen, M., Mullins, M., Murphy, F., Shannon, D., Furxhi, I., & Ryan, C. (2019) · Cybernetics & Systems · doi:10.1080/01969722.2019.1660541 · Cited by 72
An Overview on the Current Status and Future Perspectives of Smart Cars
Arena, F., Pau, G., & Severino, A. G. (2020) · Infrastructures · doi:10.3390/infrastructures5070053 · Cited by 76
Artificial Intelligence in Brain Tumour Surgery—An Emerging Paradigm
Williams, S. C., Horsfall, H. L., Funnell, J. P., HANRAHAN, J. G., Khan, D. Z., Muirhead, W. R., Stoyanov, D. V., & Marcus, H. J. (2021) · Cancers · doi:10.3390/cancers13195010 · Cited by 80
From Ethics Washing to Ethics Bashing: A Moral Philosophy View on Tech Ethics
Bietti, E. (2021) · Journal of Social Computing · doi:10.23919/jsc.2021.0031 · Cited by 170
Digital Twin: Values, Challenges and Enablers
Rasheed, A., San, O., & Kvamsdal, T. (2019) · arXiv · doi:10.48550/arxiv.1910.01719 · Cited by 81
“With the rapid development of the AI technologies in mind, the moral machine experiment [87] provides many insights how machines might make moral decisions.”
BackgroundFrom Ethics Washing to Ethics Bashing: A View on Tech Ethics from Within Moral PhilosophyPreprint
Bietti, E. (2021) · SSRN Electronic Journal · doi:10.2139/ssrn.3914119 · Cited by 79
Yurrita, M., Draws, T., Balayn, A., Murray-Rust, D., Tintarev, N., & Bozzon, A. (2023) · Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3544548.3581161 · Cited by 108
Promises and Perils of Experimentation: The Mutual-Internal-Validity Problem
Lin, H., Werner, K. M., & Inzlicht, M. (2021) · Perspectives on Psychological Science · doi:10.1177/1745691620974773 · Cited by 82
AI Literacy in K-12 and Higher Education in the Wake of Generative AI: An Integrative Review
Gu, X., & Ericson, B. J. (2025) · arXiv · doi:10.48550/arxiv.2503.00079 · Cited by 73
Hensel, L., Witte, M., Caria, S. A., et al. (2020) · doi:10.31234/osf.io/3kfmh · Cited by 72
Rhim, J., Lee, G.-B., & Lee, J. (2019) · Computers in Human Behavior · doi:10.1016/j.chb.2019.08.010 · Cited by 79
Niszczota, P., & Kaszás, D. (2020) · PLoS ONE · doi:10.1371/journal.pone.0239277 · Cited by 82
Trustworthy human-AI partnerships
Ramchurn, S. D., Stein, S., & Jennings, N. R. (2021) · iScience · doi:10.1016/j.isci.2021.102891 · Cited by 73
Human Perceptions on Moral Responsibility of AI: A Case Study in AI-Assisted Bail Decision-Making
Lima, G., Grgić-Hlača, N., & Cha, M. (2021) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3411764.3445260 · Cited by 91
“Literature exists on the public perception of moral and legal issues concerning AI [6, 7, 62].”
BackgroundArtificial Intelligence in Drug Design: Algorithms, Applications, Challenges and Ethics
Arabi, A. A. (2021) · Future Drug Discovery · doi:10.4155/fdd-2020-0028 · Cited by 75
Calhoun, C. S., Bobko, P., Gallimore, J. J., & Lyons, J. B. (2019) · Journal of Trust Research · doi:10.1080/21515581.2019.1579730 · Cited by 74
Values and Ethics in Information Systems
Spiekermann, S., Krasnova, H., Hinz, O. F., Baumann, A., Benlian, A., Gimpel, H., Heimbach, I., Köster, A., Maedche, A., Niehaves, B., et al. (2022) · Business & Information Systems Engineering · doi:10.1007/s12599-021-00734-8 · Cited by 82
“The work on ethical algorithms culminated in MIT’s ‘‘Moral Machine Experiment’’ to inform the evasive actions of autonomous cars (Awad et al. 2018) with the help of ‘‘trolley economics.’’”
Closing the loop – The human role in artificial intelligence for education
Ninaus, M., & Sailer, M. (2022) · Frontiers in Psychology · doi:10.3389/fpsyg.2022.956798 · Cited by 89
Ethics of artificial intelligence in medicine
Savulescu, J., Giubilini, A., Vandersluis, R., & Mishra, A. (2024) · Singapore Medical Journal · doi:10.4103/singaporemedj.smj-2023-279 · Cited by 68
“They indicate our distillation of the key benefits as captured in the literature, as well as through our research.”
The future of road transport : implications of automated, connected, low-carbon and shared mobility
MARIA, A. R., Ciuffo, B., PATRICIA, A. D., FULVIO, A., PHILIPPE, A. J., GIANMARCO, B., CLAUDIA, B., DARINA, B., SILVIA, B., ROBERT, B., et al. (2019) · EUR · doi:10.2760/668964 · Cited by 93
An ethical decision-making framework with serious gaming: a smart water case study on flooding
Ewing, G., & Demir, İ. (2021) · Journal of Hydroinformatics · doi:10.2166/hydro.2021.097 · Cited by 61
Autonomous vehicles and mobility for people with special needs
Kaßens-Noor, E., Cai, M., Kotval-K, Z., & Decaminada, T. (2021) · Transportation Research Part A Policy and Practice · doi:10.1016/j.tra.2021.06.014 · Cited by 69
“387 However, the impact of AVs, and how they are to be implemented, is still subject to a myriad of planning, political, technological, and ethical debate (Beiker, 2012; Litman 2017a, 2017b; Awad et al., 2018; Pearl 2019).”
BackgroundUnderstanding responsibility in Responsible AI. Dianoetic virtues and the hard problem of context
Constantinescu, M., Voinea, C., Uszkai, R., & Vică, C. (2021) · Ethics and Information Technology · doi:10.1007/s10676-021-09616-9 · Cited by 66
“The framework of virtue ethics seems to suggest that we view AI systems not as autonomous agents equal to humans, but rather similar to what contemporary research describes as assistive companions (Maes, 1995; Savulescu & Maslen, 2015; Voinea et al., 2020) or, broadly put, as tools (Balkin, 2017).”
ResultBagó, B., Kovács, M., Protzko, J., Nagy, T., Kekecs, Z., Pálfi, B., Adamkovic, M., Adamus, S., Albalooshi, S., Albayrak‐Aydemir, N., et al. (2022) · Nature Human Behaviour · doi:10.1038/s41562-022-01319-5 · Cited by 91
“These measures, such as economic status 39 , individual-level individualism–collectivism 39 and religiosity 40 , have been previously shown to be related to moral judgement.”
BackgroundMethodCultural Bias in Explainable AI Research: A Systematic Analysis
Peters, U., & Carman, M. (2024) · Journal of Artificial Intelligence Research · doi:10.1613/jair.1.14888 · Cited by 95
“Relatedly, we recently found that the scope of generalizations in many XAI user studies was only poorly correlated with the size of the studies’ samples suggesting that hasty, overly broad extrapolations may have been common (Peters & Carman, 2023).”
ResultLateef, F., Kas, M., & Ruichek, Y. (2021) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2021.3053178 · Cited by 72
Liu, P., Du, Y., & Xu, Z. (2019) · Accident Analysis & Prevention · doi:10.1016/j.aap.2019.02.012 · Cited by 65
Autonomous cars: Recent developments, challenges, and possible solutions
Singh, S. P., & Saini, B. (2021) · IOP Conference Series: Materials Science and Engineering · doi:10.1088/1757-899x/1022/1/012028 · Cited by 49
Ethical dilemmas are really important to potential adopters of autonomous vehicles
Gill, T. (2021) · Ethics and Information Technology · doi:10.1007/s10676-021-09605-y · Cited by 72
AI Literacy in K-12 and Higher Education in the Wake of Generative AI: An Integrative Review
Gu, X., & Ericson, B. J. (2025) · ACM Conference on International Computing Education Research (ICER) · doi:10.1145/3702652.3744217 · Cited by 135
“When evaluated in this way, the use of AI has raised ethical concerns, including AI accountability for its decisions, latent biases, environmental costs, and privacy intrusions during training data collection [6, 15, 106].”
Training data influence analysis and estimation: a survey
Hammoudeh, Z., & Lowd, D. (2024) · Machine Learning · doi:10.1007/s10994-023-06495-7 · Cited by 136
The looming challenges of regulating high level autonomous vehicles
Mordue, G., Yeung, A., & Wu, F. (2019) · Transportation Research Part A Policy and Practice · doi:10.1016/j.tra.2019.11.007 · Cited by 65
Scaling up psychology via Scientific Regret Minimization
Agrawal, M., Peterson, J. C., & Griffiths, T. L. (2020) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1915841117 · Cited by 66
“The Moral Machine experiment (7) is one recent example of a large-scale online study.”
BackgroundMethodThe Role of Explainable AI in the Research Field of AI Ethics
Vainio-Pekka, H., Agbese, M., Jantunen, M., Vakkuri, V., Mikkonen, T., Rousi, R. A., & Abrahamsson, P. (2023) · ACM Transactions on Interactive Intelligent Systems · doi:10.1145/3599974 · Cited by 85
“Massachusetts Institute of Technology’s “Moral Machine” research [ 16 ] collected 40 million answers to their online experiment, which studied decisions in ethical situations related to autonomous driving.”
BackgroundAwad, E., Dsouza, S., Bonnefon, J., Shariff, A., & Rahwan, I. (2020) · Communications of the ACM · doi:10.1145/3339904 · Cited by 63
Self-citationUnveiling the Mind of the Machine
Clegg, M., Hofstetter, R., de Bellis, E., & Schmitt, B. H. (2023) · Journal of Consumer Research · doi:10.1093/jcr/ucad075 · Cited by 75
Horowitz, M. C., & Kahn, L. (2024) · International Studies Quarterly · doi:10.1093/isq/sqae020 · Cited by 68
Fakhimi, A., Garry, T., & Biggemann, S. (2023) · Australasian Marketing Journal (AMJ) · doi:10.1177/14413582231181140 · Cited by 67
“Previous literature in this field has tended to focus on the attribution of behavioural aspects, emotional states and mental states (Awad et al., 2018; Minsky, 2007).”
BackgroundEthical artificial intelligence (AI): principles and practices
Law, R., Ye, H., & Lei, S. I. (2024) · International Journal of Contemporary Hospitality Management · doi:10.1108/ijchm-04-2024-0482 · Cited by 80
Kallioinen, N., Pershina, M., Zeiser, J., Nezami, F. N., Pipa, G., Stephan, A., & Peter König, B. (2019) · Frontiers in Psychology · doi:10.3389/fpsyg.2019.02415 · Cited by 63
Public Perspectives on COVID-19 Vaccine Prioritization
Persad, G., Emanuel, E., Sangenito, S., Glickman, A., Phillips, S. A., & Largent, E. A. (2021) · JAMA Network Open · doi:10.1001/jamanetworkopen.2021.7943 · Cited by 59
Unmasking inequalities of the code: Disentangling the nexus of AI and inequality
Bircan, T., & Özbilgin, M. F. (2024) · Technological Forecasting and Social Change · doi:10.1016/j.techfore.2024.123925 · Cited by 83
Intelligent energy management system for conventional autonomous vehicles
Phan, D., Bab‐Hadiashar, A., Lai, C. Y., et al. (2019) · Energy · doi:10.1016/j.energy.2019.116476 · Cited by 61
Bartneck, C., Belpaeme, T., Eyssel, F., Kanda, T., Keijsers, M., & Šabanović, S. (2020) · Cambridge University Press eBooks · doi:10.1017/9781108676649.006 · Cited by 61
Queue questions: Ethics of COVID‐19 vaccine prioritization
Giubilini, A., Savulescu, J., & Wilkinson, D. J. C. (2021) · Bioethics · doi:10.1111/bioe.12858 · Cited by 59
Pöllänen, E., Read, G. J. M., Lane, B. R., Thompson, J., & Salmon, P. M. (2020) · Ergonomics · doi:10.1080/00140139.2020.1744064 · Cited by 57
“Indeed, some cross-cultural differences in ethical autonomous vehicle decision-making were identified via the large-scale ‘Moral Machine’ experiment conducted at MIT (Awad et al. 2018).”
BackgroundMethodHeinz, N., & Koessler, A. (2020) · Ecological Economics · doi:10.1016/j.ecolecon.2021.106987 · Cited by 63
Himmelreich, J. (2022) · AI & Society · doi:10.1007/s00146-021-01357-z · Cited by 71
Gopinath, K., & Narayanamurthy, G. (2022) · International Journal of Information Management · doi:10.1016/j.ijinfomgt.2022.102536 · Cited by 63
Quesque, F., Coutrot, A., Cox, S., de Souza, L. C., Báez, S., Cardona, J. F., Mulet‐Perreault, H., Flanagan, E. C., Neely‐Prado, A., Clarens, M. F., et al. (2022) · Neuropsychology · doi:10.1037/neu0000817 · Cited by 69
Accident Analysis and Prevention
Darçın, M. (2020) · IntechOpen eBooks · doi:10.5772/intechopen.78841 · Cited by 58
What's next for responsible artificial intelligence: a way forward through responsible innovation
Herrmann, H. (2023) · Heliyon · doi:10.1016/j.heliyon.2023.e14379 · Cited by 62
“Awad et al. [169] conducted this experiment for a safety-engineering project to solve ethical dilemmas in unavoidable accidents.”
BackgroundHow games can make behavioural science better
Long, B., Simson, J., Buxó‐Lugo, A., Watson, D. G., & Mehr, S. A. (2023) · Nature · doi:10.1038/d41586-023-00065-6 · Cited by 50
The Era of AI: Upholding Ethical Leadership
Uddin, A. S. M. A. (2023) · Open Journal of Leadership · doi:10.4236/ojl.2023.124019 · Cited by 66
Almaatouq, A., Griffiths, T. L., Suchow, J. W., Whiting, M. E., Evans, J. A., & Watts, D. J. (2022) · doi:10.31234/osf.io/anjkm · Cited by 57
The European perspective on responsible computing
Inverardi, P. (2019) · Communications of the ACM · doi:10.1145/3311783 · Cited by 54
Design Bioethics: A Theoretical Framework and Argument for Innovation in Bioethics Research
Pavarini, G., McMillan, R., Robinson, A., & Singh, I. (2021) · The American Journal of Bioethics · doi:10.1080/15265161.2020.1863508 · Cited by 55
“For example, Awad et al. (2018) developed an online game to investigate moral preferences around dilemmas faced by autonomous vehicles during unavoidable accidents.”
BackgroundCitizens from 13 countries share similar preferences for COVID-19 vaccine allocation priorities
Duch, R. M., Roope, L. S. J., Violato, M., Becerra, M. F., Robinson, T. S., Bonnefon, J., Friedman, J. R., Loewen, P. J., Mamidi, P., Melegaro, A., et al. (2021) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2026382118 · Cited by 53
“Awad et al. (28) employed conjoint experiments that generate 40 million decisions to determine the ethical principles the public thinks should guide machine behavior.”
Self-citationHow do we assess the trustworthiness of AI? Introducing the trustworthiness assessment model (TrAM)
Schlicker, N., Baum, K. L., Uhde, A., Sterz, S., Hirsch, M. C., & Langer, M. (2025) · Computers in Human Behavior · doi:10.1016/j.chb.2025.108671 · Cited by 106
Brauner, P., Glawe, F., Liehner, G. L., Vervier, L., & Ziefle, M. (2025) · Technological Forecasting and Social Change · doi:10.1016/j.techfore.2025.124304 · Cited by 58
Knowledge of cultural moral norms in large language models
Ramezani, A., & Xu, Y. (2023) · Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2023.acl-long.26 · Cited by 54
Sparrow, R. (2021) · AI & Society · doi:10.1007/s00146-020-01132-6 · Cited by 55
Harm to Nonhuman Animals from AI: a Systematic Account and Framework
Coghlan, S., & Parker, C. (2023) · Philosophy & Technology · doi:10.1007/s13347-023-00627-6 · Cited by 62
“The Moral Machine experiment (Awad et al., 2018) for S. Coghlan, C. Parker 1 3 25 Page 16 of 34 self-driving cars showed that various cultures prioritise human over animal life in unavoidable collisions (though some are more ‘animal-friendly’ than others).”
BackgroundMultimodality of AI for Education: Toward Artificial General Intelligence
Lee, G., Shi, L., Latif, E., Gao, Y., Bewersdorff, A., Nyaaba, M., Guo, S., Liu, Z., Mai, G., Liu, T., et al. (2025) · IEEE Transactions on Learning Technologies · doi:10.1109/tlt.2025.3574466 · Cited by 199
“The Massachusetts Institute of Technology Media Lab further advances ethical AI literacy by involving students in projects, such as the “Moral Machine,” encouraging critical reflection on AI decision-making processes and promoting student agency [186].”
Soliciting Stakeholders’ Fairness Notions in Child Maltreatment Predictive Systems
Cheng, H.-F., Stapleton, L., Wang, R., Bullock, P., Chouldechova, A., Wu, Z. S. S., & Zhu, H. (2021) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3411764.3445308 · Cited by 71
Empirica: a virtual lab for high-throughput macro-level experiments
Almaatouq, A., Becker, J. A., Houghton, J. P., Paton, N., Watts, D. J., & Whiting, M. E. (2021) · DSpace@MIT (Massachusetts Institute of Technology) · Cited by 51
Data deprivations, data gaps and digital divides: Lessons from the COVID-19 pandemic
Naudé, W. A., & Vinuesa, R. (2021) · Big Data & Society · doi:10.1177/20539517211025545 · Cited by 54
Parental Acceptance of Children’s Storytelling Robots: A Projection of the Uncanny Valley of AI
Lin, C., Šabanović, S., Dombrowski, L. S., Miller, A., Brady, E. L., & MacDorman, K. F. (2021) · Frontiers in Robotics and AI · doi:10.3389/frobt.2021.579993 · Cited by 62
Tragic Choices and the Virtue of Techno-Responsibility Gaps
Danaher, J. (2022) · Philosophy & Technology · doi:10.1007/s13347-022-00519-1 · Cited by 53
“A study by ProPublica journalists suggested that these algorithms were systematically biased against black defendants/prisoners, consistently resulting in more false positive risk scores for them than for white defendants/prisoners (Angwin et al. 2016).”
Background
Metadata
- Authors
- Awad, E., Dsouza, S., Kim, R., Schulz, J., Henrich, J., Shariff, A., Bonnefon, J.-F., & Rahwan, I.
- Affiliations
- MIT Media Lab (US)
- Harvard University (US)
- University of British Columbia (CA)
- Centre National de la Recherche Scientifique (FR)
- Corresponding author
- Shariff, Azim
- Bonnefon, Jean-François
- Rahwan, Iyad
- ORCID
- Venue
- Nature
- Published
- 2018-10-24
- Volume
- 563
- Issue
- 7729
- Pages
- 59-64
- ISSN
- 0028-0836 (ISSN-L)
- 1476-4687
- OpenAlex
- W2896252141
- Cited by
- 1,859
- References
- 31
- Citations by year
- 2026: 214 · 2025: 263 · 2024: 222 · 2023: 270 · 2022: 273 · 2021: 263 · 2020: 223 · 2019: 120 · 2018: 10
- FWCI
- 91.60
- Type
- journal article
- Language
- en
- Open access
- green
- hdl.handle.net
- Licence
- www.springer.com/tdm (tdm)
- Locations
- Nature (publishedVersion)
- PubMed (publishedVersion)
- Open Research Exeter (University of Exeter) (acceptedVersion)
- DSpace@MIT (Massachusetts Institute of Technology) (submittedVersion)
- RePEc: Research Papers in Economics (submittedVersion)
- Figshare (submittedVersion)
- MPG.PuRe (Max Planck Society) (submittedVersion)
- Funders
- Agence Nationale de la Recherche (ANR-17-EURE-0010, ANR-10-LABX-0029)
- Topics
- Life Sciences › Neuroscience › Cognitive Neuroscience › Psychology of Moral and Emotional Judgment
- Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
- Social Sciences › Business, Management and Accounting › Organizational Behavior and Human Resource Management › Ethics in Business and Education
- Keywords
- autonomous vehicles, machine ethics, ethical dilemmas, ethical preferences, cross-cultural variation, individual differences, demographic variation, modern institutions, cultural characteristics
- MeSH terms
- Accidents, Traffic, Pedestrians, Artificial Intelligence, Data Collection, Decision Making, Female, Humans, Male, Morals, Public Opinion, Robotics, Translating, Motor Vehicles, Internet, Internationality, Harm Reduction
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
