The DOI matches its Crossref record; 49 of the 59 references it cites were found in a registry or an index; none of the 65 sources its claims cite is checked for support yet.
Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies
Open accessAbstract
The significant advancements in applying artificial intelligence (AI) to healthcare decision-making, medical diagnosis, and other domains have simultaneously raised concerns about the fairness and bias of AI systems. This is particularly critical in areas like healthcare, employment, criminal justice, credit scoring, and increasingly, in generative AI models (GenAI) that produce synthetic media. Such systems can lead to unfair outcomes and perpetuate existing inequalities, including generative biases that affect the representation of individuals in synthetic data. This survey study offers a succinct, comprehensive overview of fairness and bias in AI, addressing their sources, impacts, and mitigation strategies. We review sources of bias, such as data, algorithm, and human decision biases—highlighting the emergent issue of generative AI bias, where models may reproduce and amplify societal stereotypes. We assess the societal impact of biased AI systems, focusing on perpetuating inequalities and reinforcing harmful stereotypes, especially as generative AI becomes more prevalent in creating content that influences public perception. We explore various proposed mitigation strategies, discuss the ethical considerations of their implementation, and emphasize the need for interdisciplinary collaboration to ensure effectiveness. Through a systematic literature review spanning multiple academic disciplines, we present definitions of AI bias and its different types, including a detailed look at generative AI bias. We discuss the negative impacts of AI bias on individuals and society and provide an overview of current approaches to mitigate AI bias, including data pre-processing, model selection, and post-processing. We emphasize the unique challenges presented by generative AI models and the importance of strategies specifically tailored to address these. Addressing bias in AI requires a holistic approach involving diverse and representative datasets, enhanced transparency and accountability in AI systems, and the exploration of alternative AI paradigms that prioritize fairness and ethical considerations. This survey contributes to the ongoing discussion on developing fair and unbiased AI systems by providing an overview of the sources, impacts, and mitigation strategies related to AI bias, with a particular focus on the emerging field of generative AI.
Overview
AI snapshot
Written from the paper’s full text.
TL;DR
This survey reviews sources, impacts, and mitigation strategies for fairness and bias in AI, with particular attention to generative AI bias. It argues that bias arises from data, algorithms, and users, can perpetuate inequality, and requires holistic, interdisciplinary mitigation.
Key findings
- Bias in AI is defined as systematic error in decision-making that leads to unfair outcomes; sources include data bias, algorithmic bias, and user bias, with generative bias as an emergent issue in GenAI outputs.
- Real-world examples cited include COMPAS being biased against African-American defendants, a healthcare mortality algorithm biased against African-American patients, NIST finding facial recognition less accurate for darker skin tones, and text-to-image models (StableDiffusion, DALL-E, Midjourney) predominantly generating men for CEOs and people of color for criminals or terrorists.
- Mitigation approaches reviewed include pre-processing data (oversampling, undersampling, synthetic data, adversarial debiasing), model selection (group fairness, individual fairness, regularization, ensemble methods), and post-processing (e.g., equalized odds); GenAI is said to require holistic strategies, audits, monitoring, and feedback loops.
- Fairness types discussed include group fairness (demographic parity, disparate mistreatment, equal opportunity), individual fairness, counterfactual fairness, procedural fairness, and causal fairness; the paper distinguishes fairness as a deliberate goal from bias, which can be unintentional.
Methods
- Design: a survey described as a systematic literature review spanning multiple academic disciplines, providing an overview of AI bias sources, impacts, and mitigation strategies.
- Data or sample: no primary dataset or sample is analyzed; the paper draws on prior literature and illustrative examples from healthcare, criminal justice, hiring, credit scoring, facial recognition, and generative AI.
- Main analysis: narrative synthesis, with tables characterizing types of AI bias, mitigation approaches and challenges, and definitions of fairness; no quantitative meta-analysis or formal search protocol is reported.
Limitations
- Stated: lack of diverse and representative training data; difficulty identifying and measuring different types of bias; trade-offs between fairness and accuracy; ethical questions about prioritizing types of bias or groups; and limitations of mitigation methods, including pre-processing being time-consuming or ineffective if data are already biased, model selection lacking consensus on fairness, and post-processing being complex and data-intensive.
- Stated: fairness definitions vary and may conflict; group and individual fairness approaches have trade-offs; fairness metrics may not capture intersectionality; and mitigation attempts can have unintended consequences, such as predictive policing efforts potentially increasing racial disparities in arrests.
- Evident: as a brief survey, the paper does not report a systematic search strategy, inclusion/exclusion criteria, or quantitative synthesis; its examples are illustrative rather than exhaustive.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
48 Not checked yet · 17 Own interpretation · 8 No basis given
None of the sources cited for this claim have been checked yet.
In the paper
“Additionally, model selection methods may be limited by the lack of consensus on what constitutes fairness, and post-processing methods can be complex and require large amounts of additional data [32].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 4.1. Overview of Current Approaches to Mitigate Bias in AI, Including Pre-Processing Data, Model Selection, and Post-Processing Decisions
hedgedconditional- 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.
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
“One of the main challenges is the lack of diverse and representative training data.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 2. Unintended consequences on the distribution of outcomes
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.
In the paper
“Biased algorithms can lead to the underrepresentation of certain groups, such as people of color or those from lower socioeconomic backgrounds, in credit scoring systems, making it harder for them to access loans or mortgages [25].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 3.1. Negative Impacts of Bias in AI on Individuals and Society, Including Discrimination and Perpetuation of Existing Inequalities
What the sources say
- Not checked yetcites [25] Dwork, C., Hardt, M., Pitassi, T., Reingold, O., and Zemel, R. (2012, January 8–10). Fairness through awareness. Proceedings of the 3rd Innovations in Theoretical Computer Science Conference, Cambridge, MA, USA. source
plaindirect- 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.
In the paper
“The same GenAI models that exhibit gender bias, perhaps unsurprisingly, also portray criminals or terrorists as people of color.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 3.1. Negative Impacts of Bias in AI on Individuals and Society, Including Discrimination and Perpetuation of Existing Inequalities
What the sources say
- Not checked yetcites [23] Mittelstadt, 2016, The ethics of algorithms: Mapping the debate, Big Data Soc. source
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.
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
“GenAI models trained on internet-sourced images likely suffer from this bias, as the data mirrored existing disparities.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 2.3. Real-World Examples of Bias in AI
hedgedconditional- 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
“A study by the National Institute of Standards and Technology (NIST) found that facial recognition technology was significantly less accurate for people with darker skin tones, leading to higher rates of false positives [16].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 2.3. Real-World Examples of Bias in AI
What the sources say
- Not checked yetcites [16] Schwartz, R., Vassilev, A., Greene, K., Perine, L., Burt, A., and Hall, P. (2022). Towards a Standard for Identifying and Managing Bias in Artificial Intelligence. source
plaindirect- 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.
None of the sources cited for this claim have been checked yet.
In the paper
“Researchers and practitioners have proposed various mitigation strategies, such as improving data quality [8] and designing explicitly fair algorithms [9][10][11].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 1. Introduction
What the sources say
- Not checked yetcites [8] Asan, 2020, Artificial intelligence and human trust in healthcare: Focus on clinicians, J. Med. Internet Res. source
- Not checked yetcites [9][10][11]
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.
In the paper
“Furthermore, when prompted to generate images of criminals or terrorists, the models' output overwhelmingly more people of color.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 2.3. Real-World Examples of Bias in AI
What the sources say
- Not checked yetcites [23] Mittelstadt, 2016, The ethics of algorithms: Mapping the debate, Big Data Soc. source
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.
In the paper
“A striking instance of GenAI bias was reported, where text-to-image models like StableDiffusion, OpenAI's DALL-E, and Midjourney exhibited racial and stereotypical biases in their outputs [23].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 2.3. Real-World Examples of Bias in AI
What the sources say
- Not checked yetcites [23] Mittelstadt, 2016, The ethics of algorithms: Mapping the debate, Big Data Soc. source
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.
None of the sources cited for this claim have been checked yet.
In the paper
“Multiple studies have identified biases against certain groups in AI systems, such as the facial recognition systems studied by Buolamwini and Gebru [1] and hiring algorithms examined by Dastin [2].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 1. Introduction
What the sources say
- Not checked yetcites [1] Buolamwini, J., and Gebru, T. (2018, January 23–24). Gender shades: Intersectional accuracy disparities in commercial gender classification. Proceedings of the 1st Conference on Fairness, Accountability and Transparency, New York, NY, USA.
- Not checked yetcites [2] Dastin, J. (2018). Ethics of Data and Analytics, Auerbach Publications.
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.
In the paper
“When prompted to generate images of CEOs, these models predominantly produced images of men, reflecting gender bias.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 2.3. Real-World Examples of Bias in AI
What the sources say
- Not checked yetcites [23] Mittelstadt, 2016, The ethics of algorithms: Mapping the debate, Big Data Soc. source
plaindirect- 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.
In the paper
“Another study found similar biases in a similar system used in the state of Wisconsin [19].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 2.3. Real-World Examples of Bias in AI
What the sources say
- Not checked yetcites [19] Angwin, J., Larson, J., Mattu, S., and Kirchner, L. (2016). Ethics of Data and Analytics, Auerbach Publications.
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.
In the paper
“A study conducted by Obermeyer et al. [20] found that the system was more likely to assign higher-risk scores to African-American patients, even when other factors, such as age and health status, were the same.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 2.3. Real-World Examples of Bias in AI
What the sources say
- Not checked yetcites Obermeyer et al. [20] Obermeyer, 2019, Dissecting racial bias in an algorithm used to manage the health of populations, Science source
plaindirect- 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.
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
“The public deployment of these systems can lead to serious consequences, such as denial of services, job opportunities, or even wrongful arrests or convictions.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 3.1. Negative Impacts of Bias in AI on Individuals and Society, Including Discrimination and Perpetuation of Existing Inequalities
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.
In the paper
“When generative AI (GenAI) models are prompted to create images of CEOs, they tend to reinforce stereotypes by depicting CEOs predominantly as men [23].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 3.1. Negative Impacts of Bias in AI on Individuals and Society, Including Discrimination and Perpetuation of Existing Inequalities
What the sources say
- Not checked yetcites [23] Mittelstadt, 2016, The ethics of algorithms: Mapping the debate, Big Data Soc. source
plaindirect- 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.
In the paper
“For example, group fairness approaches may result in unequal treatment of individuals within a group, while individual fairness approaches may not address systemic biases that affect entire groups [26].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 6.2. Discussion of the Limitations and Challenges of These Approaches
What the sources say
- Not checked yetcites [26] Ananny, 2018, Seeing without knowing: Limitations of the transparency ideal and its application to algorithmic accountability, New Media Soc. source
hedgedconditional- 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.
None of the sources cited for this claim have been checked yet.
In the paper
“It is clear that these powerful computational tools, if not diligently designed and audited, have the potential to perpetuate and even amplify existing biases, particularly those related to race, gender, and other societal constructs [40][41][42][43].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 7. Conclusions
What the sources say
- Not checked yetcites [40] Crenshaw, K. (1989). Feminist Legal Theories, Routledge.
- Not checked yetcites [41] Nicoletti, L., and Bass, D. (Bloomberg Technology + Equality, 2023). Humans Are Biased: Generative AI Is Even Worse, Bloomberg Technology + Equality.
- Not checked yetcites [42] Cirillo, 2020, Sex and gender differences and biases in artificial intelligence for biomedicine and healthcare, NPJ Digit. Med. source
boostedconditional- 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: Bias can arise due to various factors, such as biased data or algorithmic design.
In the paper
“For example, group fairness metrics may not consider intersectionality or the ways in which different dimensions of identity (such as race, gender, and socioeconomic status) interact and affect outcomes [40].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 6.2. Discussion of the Limitations and Challenges of These Approaches
What the sources say
- Not checked yetcites [40] Crenshaw, K. (1989). Feminist Legal Theories, Routledge.
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.
In the paper
“For example, some researchers have found that attempts to mitigate bias in predictive policing algorithms may increase racial disparities in arrests [34].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 6.2. Discussion of the Limitations and Challenges of These Approaches
What the sources say
- Not checked yetcites [34] Ferguson, 2012, Predictive policing and reasonable suspicion, Emory LJ
hedgedconditional- 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.
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
“However, this may result in reduced accuracy for certain groups or in certain contexts.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 2. Unintended consequences on the distribution of outcomes
hedgedconditional- 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: unclear — It is unclear why the paper says it holds.
In the paper
“Research has shown that AI systems used to predict healthcare outcomes can be biased against certain groups, such as African Americans [25].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 5.3. Real-World Examples of Fairness in AI
What the sources say
- Not checked yetcites [25] Dwork, C., Hardt, M., Pitassi, T., Reingold, O., and Zemel, R. (2012, January 8–10). Fairness through awareness. Proceedings of the 3rd Innovations in Theoretical Computer Science Conference, Cambridge, MA, USA. source
plaindirect- 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.
None of the sources cited for this claim have been checked yet.
In the paper
“Corbett-Davies and collaborators introduced risk-minimization approaches aimed at minimizing disparities [27,28].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 6.1. Overview of Current Approaches to Ensure Fairness in AI, including Group Fairness and Individual Fairness
What the sources say
- Not checked yetcites [27] Corbett-Davies, S., Pierson, E., Feller, A., Goel, S., and Huq, A. (2017, January 13–17). Algorithmic decision making and the cost of fairness. Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Halifax, NS, Canada. source
- Not checked yetcites [28] Corbett-Davies, S., and Goel, S. (2018). The measure and mismeasure of fairness: A critical review of fair machine learning. arXiv.
plaindirect- 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.
None of the sources cited for this claim have been checked yet.
In the paper
“To combat bias, this paper has highlighted strategies such as robust data augmentation, the application of counterfactual fairness, and the imperative for diverse, representative datasets alongside unbiased data collection methods [49][50][51].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 7. Conclusions
What the sources say
- Not checked yetcites [49] Chouldechova, A., and Roth, A. (2018). The frontiers of fairness in machine learning. arXiv.
- Not checked yetcites [50] Mehrabi, 2021, A survey on bias and fairness in machine learning, ACM Comput. Surv. (CSUR) source
- Not checked yetcites [51] Verma, S., and Rubin, J. (2018, January 29). Fairness definitions explained. Proceedings of the International Workshop on Software Fairness, Gothenburg, Sweden. source
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.
- Rests on an assumption: Group fairness can be achieved through various techniques such as re-sampling, pre-processing, or post-processing of the data used to train the AI model.
- Rests on an assumption: Individual fairness can be achieved through techniques such as counterfactual fairness or causal fairness.
In the paper
“Research has shown that AI recruitment systems can be biased against women, as they may be less likely to be selected for male-dominated roles [26].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 5.3. Real-World Examples of Fairness in AI
What the sources say
- Not checked yetcites [26] Ananny, 2018, Seeing without knowing: Limitations of the transparency ideal and its application to algorithmic accountability, New Media Soc. source
hedgedconditional- 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.
None of the sources cited for this claim have been checked yet.
In the paper
“To address this issue, some companies have implemented “gender decoder” tools that analyze job postings and suggest changes to reduce gender bias [17,37].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 5.3. Real-World Examples of Fairness in AI
What the sources say
plaindirect- 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.
In the paper
“For example, group fairness approaches may result in unequal treatment of individuals within a group, while individual fairness approaches may not address systemic biases that affect entire groups [26].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 6.2. Discussion of the Limitations and Challenges of These Approaches
What the sources say
- Not checked yetcites [26] Ananny, 2018, Seeing without knowing: Limitations of the transparency ideal and its application to algorithmic accountability, New Media Soc. source
hedgedconditional- 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.
In the paper
“For example, a study by Buolamwini and Gebru [1] demonstrated that oversampling darker-skinned individuals improved the accuracy of facial recognition algorithms for this group.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 4.1. Overview of Current Approaches to Mitigate Bias in AI, Including Pre-Processing Data, Model Selection, and Post-Processing Decisions
What the sources say
- Not checked yetcites [1] Buolamwini, J., and Gebru, T. (2018, January 23–24). Gender shades: Intersectional accuracy disparities in commercial gender classification. Proceedings of the 1st Conference on Fairness, Accountability and Transparency, New York, NY, USA.
boosteddirect- 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.
In the paper
“For example, a study by Kamiran and Calders [31] proposed a method for selecting classifiers that achieve demographic parity, ensuring that the positive and negative outcomes are distributed equally across different demographic groups.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 4.1. Overview of Current Approaches to Mitigate Bias in AI, Including Pre-Processing Data, Model Selection, and Post-Processing Decisions
What the sources say
- Not checked yetcites [31] Kamiran, 2012, Data preprocessing techniques for classification without discrimination, Knowl. Inf. Syst. source
plaindirect- 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.
In the paper
“For example, researchers have proposed post-processing methods that adjust the decisions made by a model to achieve equalized odds, which ensures that false positives and false negatives are equally distributed across different demographic groups [11].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 4.1. Overview of Current Approaches to Mitigate Bias in AI, Including Pre-Processing Data, Model Selection, and Post-Processing Decisions
What the sources say
- Not checked yetcites [11] Yan, S., Kao, H.T., and Ferrara, E. (July, January 26). Fair class balancing: Enhancing model fairness without observing sensitive attributes. Proceedings of the 29th ACM International Conference on Information & Knowledge Management, Online.
plaindirect- 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.
None of the sources cited for this claim have been checked yet.
In the paper
“Researchers have proposed using model selection methods that prioritize fairness, such as those based on group fairness [11] or individual fairness [30].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 4.1. Overview of Current Approaches to Mitigate Bias in AI, Including Pre-Processing Data, Model Selection, and Post-Processing Decisions
What the sources say
- Not checked yetcites [11] Yan, S., Kao, H.T., and Ferrara, E. (July, January 26). Fair class balancing: Enhancing model fairness without observing sensitive attributes. Proceedings of the 29th ACM International Conference on Information & Knowledge Management, Online.
- Not checked yetcites [30] Zafar, M.B., Valera, I., Gomez Rodriguez, M., and Gummadi, K.P. (2017, January 3–7). Fairness beyond disparate treatment & disparate impact: Learning classification without disparate mistreatment. Proceedings of the 26th International Conference on World Wide Web, Perth, Australia. source
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.
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
“Moreover, mitigation approaches may introduce trade-offs between fairness and accuracy.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 2. Unintended consequences on the distribution of outcomes
hedgedconditional- 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 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
“Finally, there are concerns about the potential for unintended consequences and harmful outcomes resulting from attempts to ensure fairness in AI.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 6.2. Discussion of the Limitations and Challenges of These Approaches
hedgedconditional- 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. 26
- Probably not a substantial claim. 24
- The quote only partly states this claim. 1
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
“For example, pre-processing data can be time-consuming and may not always be effective, especially if the data used to train models are already biased.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 4.1. Overview of Current Approaches to Mitigate Bias in AI, Including Pre-Processing Data, Model Selection, and Post-Processing Decisions
hedgedconditional- 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.
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
“Algorithmic bias can be difficult to detect and quantify, especially when the algorithms are complex or opaque.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 2. Unintended consequences on the distribution of outcomes
plainconditional- 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: fails — The paper may give no reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
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
“Additionally, the sources of bias may be difficult to isolate, as bias can arise from multiple sources, such as the data, the algorithm, and the user.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 2. Unintended consequences on the distribution of outcomes
hedgedconditional- 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: fails — The paper may give no reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
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
“However, collecting diverse and representative data can be challenging, especially when dealing with sensitive or rare events.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 2. Unintended consequences on the distribution of outcomes
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: fails — The paper may give no reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
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
“Additionally, there may be privacy concerns when collecting certain types of data, such as medical records or financial information.”
- Also stated in: § 2. Unintended consequences on the distribution of outcomes · plain
“These challenges can limit the effectiveness of dataset augmentation as a mitigation approach.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 2. Unintended consequences on the distribution of outcomes
plaindirect- 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.
The quote only partly states this claim.
- Also stated in: § 2. Unintended consequences on the distribution of outcomes · plain
In the paper
“These biases can perpetuate systemic discrimination and inequality, with detrimental effects on individuals and communities in areas like hiring, lending, and criminal justice [3][4][5][6][7].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 1. Introduction
What the sources say
- Not checked yetcites [3][4][5][6][7]
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.
None of the sources cited for this claim have been checked yet.
In the paper
“In the realm of generative AI, addressing bias is even more challenging as it requires a holistic strategy [14,21,22].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 4.1. Overview of Current Approaches to Mitigate Bias in AI, Including Pre-Processing Data, Model Selection, and Post-Processing Decisions
What the sources say
- Not checked yetcites [14] Ferrara, E. (2023). Should ChatGPT be Biased? Challenges and Risks of Bias in Large Language Models. First Monday, 28. source
- Not checked yetcites [21] Ferrara, E. (2023). GenAI against humanity: Nefarious applications of generative artificial intelligence and large language models. arXiv. source
- Not checked yetcites [22] Ferrara, E. (2023). The butterfly effect in artificial intelligence systems: Implications for AI bias and fairness. arXiv. source
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: fails — It may be too vague to check 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. It may be too vague to check against evidence. 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
“For example, an AI system used in a hiring process may disproportionately exclude candidates from marginalized groups, limiting their ability to access employment opportunities and contribute to society.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 3.2. Discussion of the Ethical Implications of Biased AI
hedgedconditional- 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: fails — The paper may give no reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds. 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
“Such systems can lead to unfair outcomes and perpetuate existing inequalities, including generative biases that affect the representation of individuals in synthetic data.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > Abstract
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: unclear — It is unclear why the paper says it holds.
It may state two separate claims as one.
In the paper
“The importance of fairness and bias in AI is widely recognized by researchers, policymakers, and the academic community [1,[12][13][14][15][16].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 1. Introduction
What the sources say
- Not checked yetcites [1,[12][13][14][15][16]
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
“Bias in AI can perpetuate and even amplify existing inequalities, leading to discrimination against marginalized groups and limiting their access to essential services.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 3. Impacts of Bias in AI
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: fails — The paper may give no reason why it holds.
Probably not a substantial claim. The paper may give no reason why it holds. It may state two separate claims as one.
In the paper
“For instance, facial recognition algorithms trained on data primarily consisting of men can struggle to recognize female faces accurately, perpetuating gender bias in security systems [1].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 3.1. Negative Impacts of Bias in AI on Individuals and Society, Including Discrimination and Perpetuation of Existing Inequalities
What the sources say
- Not checked yetcites [1] Buolamwini, J., and Gebru, T. (2018, January 23–24). Gender shades: Intersectional accuracy disparities in commercial gender classification. Proceedings of the 1st Conference on Fairness, Accountability and Transparency, New York, NY, USA.
plaindirect- 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 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
“When AI systems are biased, they can limit individual freedoms and reinforce societal power dynamics.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 3.2. Discussion of the Ethical Implications of Biased AI
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.
Probably not a substantial claim. 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
“In addition to perpetuating gender stereotypes and discrimination, it can also lead to new forms of discrimination based on skin color, ethnicity, or physical appearance.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 3. Impacts of Bias in AI
plaindirect- 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: fails — The paper may give no reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
In the paper
“Additionally, model selection methods may be limited by the lack of consensus on what constitutes fairness, and post-processing methods can be complex and require large amounts of additional data [32].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 4.1. Overview of Current Approaches to Mitigate Bias in AI, Including Pre-Processing Data, Model Selection, and Post-Processing Decisions
What the sources say
- Not checked yetcites [32] Barocas, 2016, Big data’s disparate impact, Calif. Law Rev.
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.
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
“On a societal level, the widespread use of such biased AI systems can entrench discriminatory narratives and hinder efforts toward equality and inclusivity.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 3.1. Negative Impacts of Bias in AI on Individuals and Society, Including Discrimination and Perpetuation of Existing Inequalities
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.
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
“Machine learning models, a type of AI system, can learn and replicate patterns of bias present in the data used to train them, resulting in unfair or discriminatory outcomes.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 2.1. Definition of Bias in AI and Its Different Types
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: 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
“Moreover, the use of biased AI systems may undermine public trust in technology, leading to decreased adoption and even rejection of new technologies.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 3.2. Discussion of the Ethical Implications of Biased AI
hedgedconditional- 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: fails — The paper may give no reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no 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
“In the context of AI, bias can arise from various sources, including data collection, algorithm design, and human interpretation.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 2. Sources of Bias in AI
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: fails — The paper may give no reason why it holds.
Probably not a substantial claim. The paper may give no reason why it holds.
In the paper
“A study by ProPublica found that the system was biased against African-American defendants, as they were more likely to be labeled as high-risk even if they had no prior convictions.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 2.3. Real-World Examples of Bias in AI
What the sources say
- Not checked yetcites ProPublica
plaindirect- 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.
In the paper
“For example, biased algorithms used in the criminal justice system can lead to unfair treatment of certain groups, particularly people of color, who are more likely to be wrongly convicted or receive harsher sentences [1].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 3.1. Negative Impacts of Bias in AI on Individuals and Society, Including Discrimination and Perpetuation of Existing Inequalities
What the sources say
- Not checked yetcites [1] Buolamwini, J., and Gebru, T. (2018, January 23–24). Gender shades: Intersectional accuracy disparities in commercial gender classification. Proceedings of the 1st Conference on Fairness, Accountability and Transparency, New York, NY, USA.
plaindirect- 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.
In the paper
“Different people and groups may have different definitions of fairness, and these definitions may change over time [39].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 6.2. Discussion of the Limitations and Challenges of These Approaches
What the sources say
- Not checked yetcites [39] Ezzeldin, Y.H., Yan, S., He, C., Ferrara, E., and Avestimehr, S. (2023, January 7–14). Fairfed: Enabling group fairness in federated learning. Proceedings of the AAAI 2023—37th AAAI Conference on Artificial Intelligence, Washington, DC, USA.
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.
In the paper
“This can be performed through techniques such as regularization, which penalizes models for making discriminatory predictions, or through ensemble methods, which combine multiple models to reduce bias [25].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 4.1. Overview of Current Approaches to Mitigate Bias in AI, Including Pre-Processing Data, Model Selection, and Post-Processing Decisions
What the sources say
- Not checked yetcites [25] Dwork, C., Hardt, M., Pitassi, T., Reingold, O., and Zemel, R. (2012, January 8–10). Fairness through awareness. Proceedings of the 3rd Innovations in Theoretical Computer Science Conference, Cambridge, MA, USA. source
plaindirect- 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.
None of the sources cited for this claim have been checked yet.
In the paper
“Other approaches include transparency, accountability, and explainability.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 6.1. Overview of Current Approaches to Ensure Fairness in AI, including Group Fairness and Individual Fairness
What the sources say
- Not checked yetcites [26] Ananny, 2018, Seeing without knowing: Limitations of the transparency ideal and its application to algorithmic accountability, New Media Soc. source
- Not checked yetcites [38] Donovan, J., Caplan, R., Matthews, J., and Hanson, L. (2018). Algorithmic Accountability: A Primer, Data & Society.
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.
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
“Achieving both fairness and accuracy can be challenging and requires careful consideration of the trade-offs involved.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 2. Unintended consequences on the distribution of outcomes
plain- Asserted: passes — The authors put it forward as true.
- Checkable: fails — It may be too vague to check 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. It may be too vague to check against evidence. The paper’s conclusions may not depend on it. It may state two separate claims as one.
In the paper
“Additionally, different types of fairness may conflict with each other, and trade-offs may need to be made to achieve fairness in specific contexts [26].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 5.1. Definition of Fairness in AI and Its Different Types
What the sources say
- Not checked yetcites [26] Ananny, 2018, Seeing without knowing: Limitations of the transparency ideal and its application to algorithmic accountability, New Media Soc. source
hedgedconditional- 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.
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
“There are several types of fairness proposed in the literature, including group fairness, individual fairness, and counterfactual fairness.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 5. Fairness in AI
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: fails — The paper may give no reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
None of the sources cited for this claim have been checked yet.
In the paper
“Additionally, it may be difficult to determine which types of fairness are most appropriate for a given context and how to balance them appropriately [4][5][6].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 6.2. Discussion of the Limitations and Challenges of These Approaches
What the sources say
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.
In the paper
“To address this issue, the Northpointe COMPAS was modified to include a “race-neutral” version of the algorithm that achieved similar accuracy while reducing racial bias [19].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 5.3. Real-World Examples of Fairness in AI
What the sources say
- Not checked yetcites [19] Angwin, J., Larson, J., Mattu, S., and Kirchner, L. (2016). Ethics of Data and Analytics, Auerbach Publications.
plaindirect- 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.
In the paper
“To address this issue, researchers have proposed using techniques such as subgroup analysis to identify and address bias in the data used to train AI models [11].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 5.3. Real-World Examples of Fairness in AI
What the sources say
- Not checked yetcites [11] Yan, S., Kao, H.T., and Ferrara, E. (July, January 26). Fair class balancing: Enhancing model fairness without observing sensitive attributes. Proceedings of the 29th ACM International Conference on Information & Knowledge Management, Online.
plaindirect- 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
“In other words, bias can be viewed as a technical issue, while fairness is a social and ethical issue [26].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 5.2. Comparison of Fairness and Bias in AI
What the sources say
- Not checked yetcites [26] Ananny, 2018, Seeing without knowing: Limitations of the transparency ideal and its application to algorithmic accountability, New Media Soc. 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.
- Rests on an assumption: Bias can be unintentional, while fairness is inherently a deliberate and intentional goal.
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
“It is important to note that these different types of fairness are not mutually exclusive and may overlap in practice.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 5.1. Definition of Fairness in AI and Its Different Types
hedgedcorrelational- 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
“There are several different types of fairness that have been proposed in the literature, including group fairness, individual fairness, and counterfactual fairness [30].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 5.1. Definition of Fairness in AI and Its Different Types
What the sources say
- Not checked yetcites [30] Zafar, M.B., Valera, I., Gomez Rodriguez, M., and Gummadi, K.P. (2017, January 3–7). Fairness beyond disparate treatment & disparate impact: Learning classification without disparate mistreatment. Proceedings of the 26th International Conference on World Wide Web, Perth, Australia. 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.
In the paper
“One example is the COMPAS (Correctional Offender Management Profiling for Alternative Sanctions) system, which is used to predict the likelihood of recidivism in criminal defendants.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 5.3. Real-World Examples of Fairness in AI
What the sources say
- Not checked yetcites [19] Angwin, J., Larson, J., Mattu, S., and Kirchner, L. (2016). Ethics of Data and Analytics, Auerbach Publications.
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.
None of the sources cited for this claim have been checked yet.
In the paper
“Diversifying the teams involved in AI development and evaluation is equally crucial, as it brings a multiplicity of perspectives that can better identify and correct for biases [55][56][57].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 7. Conclusions
What the sources say
- Not checked yetcites [55] Chauhan, 2022, The Role of Data and Artificial Intelligence in Driving Diversity, Equity, and Inclusion, Computer source
- Not checked yetcites [56] Holstein, K., Wortman Vaughan, J., Daumé, H., Dudik, M., and Wallach, H. (2019, January 4–9). Improving fairness in machine learning systems: What do industry practitioners need?. Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems, Glasgow, UK. source
- Not checked yetcites [57] Stathoulopoulos, K., Mateos-Garcia, J.C., and Owen, H. (2023, December 15). Gender Diversity in AI Research. Available online: https://www.nesta.org.uk/report/gender-diversity-ai/.
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: passes — The paper gives a reason why it holds.
Probably not a substantial claim. It may be too vague to check against evidence.
In the paper
“Another difference is that bias can be either positive or negative, whereas fairness is only concerned with negative bias or discrimination [36].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 5.2. Comparison of Fairness and Bias in AI
What the sources say
- Not checked yetcites [36] 2017, Measuring discrimination in algorithmic decision making, Data Min. Knowl. Discov. 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.
None of the sources cited for this claim have been checked yet.
In the paper
“We also considered the ethical implications of AI in preserving privacy and the necessity for transparency, oversight, and continuous evaluation of AI systems [52][53][54].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 7. Conclusions
What the sources say
- Not checked yetcites [52] Lipton, 2018, The mythos of model interpretability: In machine learning, the concept of interpretability is both important and slippery, Queue source
- Not checked yetcites [53] Mitchell, M., Wu, S., Zaldivar, A., Barnes, P., Vasserman, L., Hutchinson, B., Spitzer, E., Raji, I.D., and Gebru, T. (2019, January 29–31). Model cards for model reporting. Proceedings of the Conference on Fairness, Accountability, and Transparency, Atlanta, GA, USA. source
- Not checked yetcites [54] Raji, I.D., and Buolamwini, J. (2019, January 29–31). Actionable auditing: Investigating the impact of publicly naming biased performance results of commercial AI products. Proceedings of the Conference on Fairness, Accountability, and Transparency, Atlanta, GA, USA. source
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.
Probably not a substantial claim. It may state two separate claims as one.
In the paper
“Research has shown that the system was biased against African-American defendants, as it was more likely to falsely predict that they would re-offend compared to white defendants [19].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 5.3. Real-World Examples of Fairness in AI
What the sources say
- Not checked yetcites [19] Angwin, J., Larson, J., Mattu, S., and Kirchner, L. (2016). Ethics of Data and Analytics, Auerbach Publications.
plaindirect- 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 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
“These ethical considerations can add complexity to the development and implementation of bias mitigation approaches.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 2. Unintended consequences on the distribution of outcomes
plaindirect- 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
“However, it is important to recognize that bias is not the only factor that can lead to unfairness, and achieving fairness may require additional efforts beyond bias mitigation [27].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 5.2. Comparison of Fairness and Bias in AI
What the sources say
- Not checked yetcites [27] Corbett-Davies, S., Pierson, E., Feller, A., Goel, S., and Huq, A. (2017, January 13–17). Algorithmic decision making and the cost of fairness. Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Halifax, NS, Canada. source
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.
- Rests on an assumption: Addressing bias in training data or algorithms can help reduce the likelihood of unfair outcomes.
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
“While these approaches have shown promising results in promoting fairness in AI, they are not without limitations and challenges.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 6.2. Discussion of the Limitations and Challenges of These Approaches
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.
- Rests on an assumption: Many current approaches to ensuring fairness in AI rely on statistical methods and assumptions that may not accurately capture the complexity of human behavior and decision-making.
The paper relies on this without support.
The use of AI systems is growing.
“The growing use of AI systems has intensified discussions around fairness and bias in artificial intelligence as potential biases and discrimination become more apparent.”
The paper relies on this without support.
Generative AI is becoming more prevalent in creating content that influences public perception.
“especially as generative AI becomes more prevalent in creating content that influences public perception.”
The paper relies on this without support.
Bias can be unintentional, while fairness is inherently a deliberate and intentional goal.
“One key difference between fairness and bias is that while bias can be unintentional, fairness is inherently a deliberate and intentional goal.”
The paper relies on this without support.
Addressing bias in training data or algorithms can help reduce the likelihood of unfair outcomes.
“For example, addressing bias in training data or algorithms can help reduce the likelihood of unfair outcomes.”
The paper relies on this without support.
Group fairness can be achieved through various techniques such as re-sampling, pre-processing, or post-processing of the data used to train the AI model.
“This can be achieved through various techniques such as re-sampling, pre-processing, or postprocessing of the data used to train the AI model.”
The paper relies on this without support.
Individual fairness can be achieved through techniques such as counterfactual fairness or causal fairness.
“Individual fairness can be achieved through techniques such as counterfactual fairness or causal fairness.”
The paper relies on this without support.
Many current approaches to ensuring fairness in AI rely on statistical methods and assumptions that may not accurately capture the complexity of human behavior and decision-making.
“Furthermore, many of the current approaches to ensuring fairness in AI rely on statistical methods and assumptions that may not accurately capture the complexity of human behavior and decision-making.”
The paper relies on this without support.
Bias can arise due to various factors, such as biased data or algorithmic design.
“Bias can arise due to various factors, such as biased data or algorithmic design, but fairness requires a conscious effort to ensure that the algorithm does not discriminate against any group or individual.”
The paper relies on this without support.
Fairness requires a conscious effort to ensure that the algorithm does not discriminate against any group or individual.
“Bias can arise due to various factors, such as biased data or algorithmic design, but fairness requires a conscious effort to ensure that the algorithm does not discriminate against any group or individual.”
The paper relies on this without support.
If an AI model is trained on data that are biased toward a particular group, re-sampling techniques can be used to create a balanced dataset where each group is represented equally.
“For example, if an AI model is trained on data that are biased toward a particular group, re-sampling techniques can be used to create a balanced dataset where each group is represented equally.”
The paper relies on this without support.
Pre-processing or post-processing can be used to adjust the output of the AI model to ensure that it does not unfairly disadvantage any group.
“Other techniques, such as pre-processing or post-processing, can be used to adjust the output of the AI model to ensure that it does not unfairly disadvantage any group.”
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?
In the citing paper
“Moreover, different sources of bias can contribute to the generation of bias, including the data source, algorithm, and users, which makes the identification more cumbersome.124”
From Bias in artificial intelligence for medical imaging: fundamentals, detection, avoidance, mitigation, challenges, ethics, and prospects (2024) · cites this paper as [124]
In the citing paper
“However, although sharing data is essential for developing robust AI tools, patient privacy when collecting medical information can pose significant challenges.124”
From Bias in artificial intelligence for medical imaging: fundamentals, detection, avoidance, mitigation, challenges, ethics, and prospects (2024) · cites this paper as [124]
In the citing paper
“In the healthcare context, the lack of interpretability [10] and explainability [11] in ML and DL models can lead to ethical issues and a loss of trust [12], [13].”
From Ethical Framework for Harnessing the Power of AI in Healthcare and Beyond (2024) · cites this paper as [12]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 5.2. Comparison of Fairness and Bias in AI
“fairness is only concerned with negative bias or discrimination [36].”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 5.2. Comparison of Fairness and Bias in AI
“fairness in AI refers to the absence of discrimination or favoritism towards any individual or group based on protected characteristics such as race, gender, age, or religion [25].”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 4.1. Overview of Current Approaches to Mitigate Bias in AI, Including Pre-Processing Data, Model Selection, and Post-Processing Decisions
“Pre-processing data involves identifying and addressing biases in the data before the model is trained.”
§ Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies > 4.1. Overview of Current Approaches to Mitigate Bias in AI, Including Pre-Processing Data, Model Selection, and Post-Processing Decisions
“or through adversarial debiasing, which involves training the model to be resilient to specific types of bias [11].”
Citations
Cites
Crenshaw, K. (2018) · Living with Contradictions · doi:10.4324/9780429499142-5 · Cited by 7,232
Dissecting racial bias in an algorithm used to manage the health of populations
Obermeyer, Z., Powers, B., Vogeli, C., & Mullainathan, S. (2019) · Science · doi:10.1126/science.aax2342 · Cited by 7,137 · cited 1×
“A study conducted by Obermeyer et al. [20] found that the system was more likely to assign higher-risk scores to African-American patients, even when other factors, such as age and health status, were the same.”
A Survey on Bias and Fairness in Machine Learning
Mehrabi, N., Morstatter, F., Saxena, N., Lerman, K., & Galstyan, A. (2022) · ACM Computing Surveys · doi:10.1145/3457607 · Cited by 5,812
Lipton, Z. C. (2018) · Queue · doi:10.1145/3236386.3241340 · Cited by 5,427
The ethics of algorithms: Mapping the debate
Mittelstadt, B. D., Allo, P., Taddeo, M., Wachter, S., & Floridi, L. (2016) · Big Data & Society · doi:10.1177/2053951716679679 · Cited by 4,567 · cited 3×
“When generative AI (GenAI) models are prompted to create images of CEOs, they tend to reinforce stereotypes by depicting CEOs predominantly as men [23].”
BackgroundMethodInfluential(2018) · Cited by 3,510
Dwork, C., Hardt, M., Pitassi, T., Reingold, O., & Zemel, R. (2012) · Proceedings of the 3rd Innovations in Theoretical Computer Science Conference · doi:10.1145/2090236.2090255 · Cited by 3,501 · cited 4×
“Similarly, an image generation model trained on datasets with limited diversity in human portraits may struggle to accurately represent a broad range of ethnicities.”
BackgroundMethodInfluentialBarocas, S., & Selbst, A. D. (2014) · doi:10.2139/ssrn.2477899 · Cited by 3,048 · cited 2×
“Can be complex and require large amounts of additional data [32].”
BackgroundSemantics derived automatically from language corpora contain human-like biases
Caliskan, A., Bryson, J. J., & Narayanan, A. (2017) · Science · doi:10.1126/science.aal4230 · Cited by 2,887 · cited 1×
“This survey study delves into the complex and multifaceted issues surrounding fairness and bias in AI, covering the sources of bias, their impacts, and proposed mitigation strategies.”
BackgroundAlgorithms of Oppression: How Search Engines Reinforce Racism
Noble, S. U. (2020) · doi:10.2307/j.ctt1pwt9w5 · Cited by 2,709
Research works printed but not linked: 3
- Eubanks, V. (2018). Automating Inequality: How High-Tech Tools Profile, Police, and Punish the Poor, St. Martin’s Press.
- Corbett-Davies, S., and Goel, S. (2018). The measure and mismeasure of fairness: A critical review of fair machine learning. arXiv.
- Crenshaw, K. (1989). Feminist Legal Theories, Routledge.
Cited by
Sorted by citations
AI‐driven adaptive learning for sustainable educational transformation
Strielkowski, W., Grebennikova, V., Lisovskiy, A., Rakhimova, G., & Vasileva, T. (2025) · Sustainable Development · doi:10.1002/sd.3221 · Cited by 581
Generative AI for Customizable Learning Experiences
Pesovski, I., Santos, R., Henriques, R., & Trajkovik, V. (2024) · Sustainability · doi:10.3390/su16073034 · Cited by 274
“Mitigation strategies include the resampling and pre-processing of data used for training the models [68], using diverse and representative datasets [69], feedback loops [70], etc.”
MethodSynthetic data generation methods in healthcare: A review on open-source tools and methods
Pezoulas, V. C., Zaridis, D. I., Mylona, E., Androutsos, C., Apostolidis, K., Tachos, N. S., & Fotiadis, D. I. (2024) · Computational and Structural Biotechnology Journal · doi:10.1016/j.csbj.2024.07.005 · Cited by 270
Ethical and legal considerations in healthcare AI: innovation and policy for safe and fair use
Pham, T. D. (2025) · Royal Society Open Science · doi:10.1098/rsos.241873 · Cited by 367
Dhal, S. B., & Kar, D. (2025) · Discover Applied Sciences · doi:10.1007/s42452-025-06472-w · Cited by 255
“In food safety, bias can impact decision-making related to product quality, supplier evaluations, and compliance checks [247].”
Metadata
- Authors
- Ferrara, E.
- Affiliations
- University of Southern California (US)
- Corresponding author
- Ferrara, Emilio
- ORCID
- Ferrara, Emilio
- Venue
- Sci
- Publisher
- MDPI AG
- Published
- 2023-12-26
- Volume
- 6
- Issue
- 1
- Pages
- 3
- ISSN
- 2413-4155 (ISSN-L)
- OpenAlex
- W4390229867
- Semantic Scholar
- 53b04ccd2a001467d7ce168e9ce20b16a9466a69
- Cited by
- 875
- References
- 59
- Citations by year
- 2026: 271 · 2025: 433 · 2024: 162 · 2023: 6
- FWCI
- 218.36
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Sci (publishedVersion, cc-by)
- SSRN Electronic Journal (acceptedVersion)
- arXiv (Cornell University) (submittedVersion, cc-by)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- arXiv (Cornell University) (cc-by)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
- Physical Sciences › Computer Science › Artificial Intelligence › Explainable Artificial Intelligence (XAI)
- Social Sciences › Business, Management and Accounting › Organizational Behavior and Human Resource Management › AI and HR Technologies
- Keywords
- algorithmic fairness, algorithmic bias, generative AI bias, generative AI, synthetic media, data bias, bias mitigation, bias mitigation strategies, data preprocessing, model selection, post-processing, societal impact of AI, stereotypes, artificial intelligence in healthcare, AI ethics
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 2026-10-09, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-09 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
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
