CoChat
Sign Up
Cited by 94Ethics and Social Impacts of AIExplainable Artificial Intelligence (XAI)Digitalization, Law, and Regulation

AI, Opacity, and Personal Autonomy

Open access
Vaassen, Bram·2022·Philosophy & Technology
DOI 10.1007/s13347-022-00577-5
Open this paper in CoChat — free accountOpen PDF

Abstract

Advancements in machine learning have fuelled the popularity of using AI decision algorithms in procedures such as bail hearings, medical diagnoses and recruitment. Academic articles, policy texts, and popularizing books alike warn that such algorithms tend to be opaque : they do not provide explanations for their outcomes. Building on a causal account of transparency and opacity as well as recent work on the value of causal explanation, I formulate a moral concern for opaque algorithms that is yet to receive a systematic treatment in the literature: when such algorithms are used in life-changing decisions, they can obstruct us from effectively shaping our lives according to our goals and preferences, thus undermining our autonomy. I argue that this concern deserves closer attention as it furnishes the call for transparency in algorithmic decision-making with both new tools and new challenges.

AI snapshot

Written from the paper’s full text.

TL;DR

The paper argues that opaque AI decision algorithms can undermine personal autonomy by hiding the causal pathways through which affected parties could shape life-changing outcomes, even when the algorithms are reliable, fair, accountable, and negotiable. It develops a causal account of transparency/opacity and uses it to motivate transparency demands, while also identifying practical and legal obstacles.

Key findings

  • Opacity and transparency are characterized causally: an AI system is opaque to someone to the extent they cannot grasp the causal explanations of its outcomes, and transparent to the extent they can.
  • Because causal explanations enable prediction and control, opaque algorithms in life-changing decisions can obstruct people from effectively shaping their lives according to their goals and preferences, thereby threatening autonomy.
  • This autonomy threat is distinct from other worries: an opaque algorithm could in principle be reliable, fair, trusted, accountable, and negotiable, yet still hide salient ways of affecting its outcomes.
  • The paper notes that a meta-study of over 100 XAI texts lists 28 desiderata but does not mention opacity undermining affected parties’ ability to influence outcomes, and a review of 34 AI ethics documents similarly focuses on users rather than affected parties.
  • Current legal instruments are inadequate: the GDPR does not provide a right to explanation of individual decisions, and the European Commission’s AI Act proposal would demand transparency toward users but not affected parties.

Methods

  • Conceptual/philosophical analysis; no empirical sample, dataset, or formal experiment is reported.
  • The paper builds a causal account of transparency and opacity from interventionist accounts of causal explanation (e.g., Woodward, Pearl, Spirtes et al.) and connects it to autonomy and responsibility.
  • It analyzes XAI literature, policy and legal texts (e.g., EU HLEG guidelines, GDPR, AI Act, ECHR), and uses hypothetical examples such as the GOV-1 government-job algorithm.

Limitations

  • Stated: a detailed comparison of the paper’s use of “autonomy” with standard accounts in the literature is postponed to another occasion.
  • Stated: whether forensic/glass-box approaches can reliably provide accurate explanations that fit affected parties’ practical perspectives is an outstanding empirical question, and whether a “fitting” translation can be legally entrenched remains open.
  • Evident from the text: the argument is normative and conceptual rather than empirically tested; the central GOV-1 case is hypothetical, and the legal discussion is selective rather than a comprehensive doctrinal analysis.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1007/s13347-022-00577-5

Cites

79 of 79 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 66 cited sources checked · 0 supported · 0 flagged

  • Interpretationreasoning

    There is a distinct concern with opaque algorithms that has not received systematic treatment: opacity can undermine users' autonomy by hiding salient pathways of affecting the algorithm's outcomes.

    “I argue that there is a distinct concern with opaque algorithms that has not received a systematic treatment in the literature: opacity can undermine the users' autonomy by hiding salient pathways of affecting the algorithm's outcomes.”
    Quote in the paperNot a substantial claim
  • Interpretationreasoning

    Even when an opaque algorithm fulfills its task excellently and its results are treated as negotiable, its opacity can threaten the autonomy of affected parties.

    “Even in cases where an opaque algorithm fulfills its task excellently, and its results are treated as negotiable, its opacity can threaten the autonomy of the parties affected by its outcomes.”
    Quote in the paperNot a substantial claim
  • Interpretationreasoning

    Motivating the demand for transparency with its relevance to autonomy would not require a double standard of the kind Zerilli et al. (2019) worry about.

    “Given that such a threat to our autonomy can occur regardless of whether the decision algorithm is executed by an AI system or by a human, motivating the demand for transparency with its relevance to autonomy would not require a double standard of the kind Zerilli et al. (2019) worry about.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Zerilli et al., 2019 source

      Not checked in the free run.

  • Premisecitation

    For many current AI algorithms, information about causal factors is exceedingly likely to be encoded implicitly rather than explicitly.

    “For many current AI algorithms, it is exceedingly likely that such information is encoded implicitly rather than explicitly”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Burrell, 2016 source

      Not checked in the free run.

  • Interpretationreasoning

    Opaque decision algorithms hide effective strategies of affecting and predicting their outcomes from affected parties.

    “In effect, opaque decision algorithms hide effective strategies of affecting and predicting their outcomes from the affected parties.”
    Quote in the paperSubstantial claim
  • Classified as probably not a substantial claim. 38
  • May combine two claims. 19
  • The quote only partly states this claim. 2

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    The right degree of transparency will often require that we grasp hypothetical causal explanations.

    “As will become apparent in the examples to follow, the right degree of transparency will often require that we grasp such hypothetical causal explanations.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    Addressing the threat opacity poses to autonomy will require a certain degree of transparency in decision-making.

    “Consequently, this threat can remain even if other worries about opaque algorithms are successfully resolved and addressing this threat will require a certain degree of transparency in decision-making.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Langer et al. (2021) list twenty-eight desiderata found in the literature but make no mention of how opacity can undermine the ability of affected parties to effectively influence outcomes according to their goals.

    “For example, a recent meta-study of over 100 texts on XAI by Langer et al. (2021), lists twenty-eight desiderata found in the literature but makes no mention of how opacity can undermine the ability of affected parties to effectively influence the outcomes according to their goals.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Langer et al., 2021

      Not checked in the free run.

  • Classified as probably not a substantial claim. May combine two claims.

    Premisereasoning

    Causal explanation can be available before the fact in hypothetical form, and events that never occurred can be causally explained in a similar way.

    “Causal explanation can be available before the fact in hypothetical form, and events that never occurred can be causally explained in a similar way.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    Fjeld et al. (2020) review thirty-four AI ethics documents, and similar remarks apply: no mention is made of opacity undermining the ability to influence outcomes.

    “Similar remarks apply to a recent review of thirty-four AI ethics documents published by civil society, the private sector, governments, intergovernmental organizations, and multi-stakeholder organizations by Fjeld et al. (2020).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Fjeld et al., 2020

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    An adequate standard of transparency for AI systems would demand only restricted access to the workings of the system.

    “An adequate standard of transparency for AI systems would thus similarly demand only restricted access to the workings of the system.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

    Premisereasoning

    Not all causal explanations will help us understand the outcome, nor will all causal explanations be of interest to us.

    “Crucially, not all causal explanations will help us understand the outcome, nor will all causal explanations be of interest to us.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

    Premisecitation

    Autonomy is taken to be some brand of self-determination, and self-determination requires long-term control over one's life and plans.

    “First, autonomy is taken to be some brand of self-determination and there is growing attention for the fact that self-determination requires a long-term control over one's life and plans”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites MacIntyre, 1983

      Not checked in the free run.

    • cites Raz, 1986

      Not checked in the free run.

    • cites Bratman, 2000

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    Undermining autonomy undermines responsibility.

    “Third, undermining autonomy undermines responsibility.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    Opacity and a lack of knowledge about how to manipulate are definitionally inseparable.

    “According to the causal account proposed above, opacity and a lack of knowledge about how to manipulate are definitionally inseparable.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    Our lack of knowledge of how to affect the outcomes is integral to an AI system being opaque to us.

    “By contrast, our lack of knowledge of how to affect the outcomes is integral to an AI system being opaque to us.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Interpretationreasoning

    The outcomes of an opaque algorithm could be treated as negotiable by allowing users to double-check them using another decision algorithm.

    “The outcomes could even be treated as negotiable by allowing users to double-check the outcomes using another decision algorithm, be it a human or an artificial one.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    There is nothing in principle that stands in the way of an algorithm fulfilling its function perfectly without the relevant stakeholders having knowledge of what produces its outcomes.

    “but there is nothing that in principle stands in the way of an algorithm fulfilling its function perfectly without the relevant stakeholders having knowledge of what produces its outcomes.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The importance of autonomy is picked up in AI ethics and legal theory.

    “The importance of autonomy is picked up in other discussions within AI ethics (e.g., Kim et al., 2021), and legal theory as well (e.g., Marshall, 2008;McLean, 2009).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites Kim et al., 2021

      Not checked in the free run.

    • cites Marshall, 2008

      Not checked in the free run.

    • cites McLean, 2009

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation

    Autonomy takes center stage in several ethical theories.

    “Autonomy takes center stage in several ethical theories.”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites Kant, 1993

      Not checked in the free run.

    • cites Mill, 1999

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Attaining full physical, algorithmic, or design transparency is extraordinarily difficult, and demanding it might be in tension with copyright laws.

    “Attaining either of these is taken to be extraordinarily difficult as well, and demanding it might even be in tension with copyright laws”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Burrell, 2016 source

      Not checked in the free run.

    • cites Ananny & Crawford, 2018 source

      Not checked in the free run.

    • cites Wachter et al., 2017

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    It is unlikely that knowledge of robust correlations will require full physical details or design details, but it is also unlikely that the required explanation will always be found at the intentional level.

    “It is unlikely that knowledge of such robust correlations will require full physical details or design details, but it is also unlikely that the required explanation will always be found at the intentional level.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    Opacity can undermine autonomy and autonomy has moral value.

    “In conclusion, it appears that opacity can undermine autonomy and autonomy has moral value.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    The GDPR fails to provide affected parties with a right to explanation of individual decisions, and instead delivers only a vague right to be informed of whether an AI is used and what 'logic' underlies the algorithm.

    “Wachter et al. (2017) point out that the GDPR fails to provide affected parties with a right to explanation of individual decisions, but instead delivers a rather vague right to be informed of (i) whether an AI is used, and (ii) what”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Wachter et al. (2017)

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    For 'forensic' strategies that explain black box outcomes without opening the black box to safeguard autonomy, they need to reliably provide accurate explanations that fit the needs and practical perspective of the affected parties.

    “In order for these”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Interpretationreasoning

    There is a real risk that the details omitted by forensic approaches are crucial to the true causal story of how the decision was produced.

    “Such neglect of detail is necessary to provide explanations that are understandable for human agents, and, so the worry goes, there is a real risk that the omitted details are in fact crucial to the true causal story of how the decision was in fact produced”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Rudin, 2019 source

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    A right to an explanation will not suffice; the explanation needs to fit the explanatory needs and perspective of the affected party.

    “The upshot is that the right to an explanation will not suffice.The explanation needs to fit the explanatory needs and perspective of the affected party.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    In the worst case, lack of detail may make for mistaken explanations altogether, taking mere correlation for causation.

    “In the worst case, this lack of detail may make for mistaken explanations altogether, taking mere correlation for causation.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationcitation

    External auditing mechanisms could help enforce the right of redress and justify accusations of bias, but they would not provide explanations and thus would not provide affected parties with action-enabling information.

    “Wachter et al. (2017, p. 98) make note of this challenge and suggest that external auditing mechanisms can be employed in such cases, but while such mechanisms could help to enforce the right of redress and to justify accusations of bias, they would not provide explanations, and thus not provide affected parties with action-enabling information.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Wachter et al. (2017, p. 98)

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Premisenone given

    There are ongoing attempts to provide explanations of black box outcomes without 'opening' black boxes.

    “There are ongoing attempts to make provide explanations of black box outcomes without”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Premisenone given

    Neither the US Constitution nor the ECHR explicitly mention autonomy.

    “Neither the US Constitution or the ECHR explicitly mention autonomy.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

    Interpretationreasoning

    If respecting personal autonomy requires providing adequate explanations for potentially life-changing decisions, then the right to personal autonomy requires a right to explanation.

    “If respecting personal autonomy requires providing adequate explanations for potentially life-changing decisions, then the right to personal autonomy requires a right to explanation.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Respecting personal autonomy requires providing adequate explanations for potentially life-changing decisions.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    The link between transparency and autonomy provides a promising extra tool for legally entrenching a right to explanation.

    “The link between transparency and autonomy provides a promising extra tool for legally entrenching a right to explanation.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    Some legal scholars maintain that the right to personal autonomy is enshrined in the European Convention of Human Rights and the US Constitution.

    “and some legal scholars maintain that the very right to personal autonomy is enshrined in the European Convention of Human Rights (ECHR) (e.g., Marshall, 1 3 2008) and the US Constitution (e.g., McLean, 2009).”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites Marshall, 2008

      Not checked in the free run.

    • cites McLean, 2009

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    The articles in the Charter are explicitly restricted to cases where the administration is 'implementing Union Law' (Art. 51), which puts a further burden on affected parties to argue that the governmental decision they demand explained counts as an implementation of Union Law.

    “Moreover, the articles in the charter are explicitly restricted to cases where the administration are”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Art.51

      Not checked in the free run.

  • May combine two claims. The quote only partly states this claim.

    Interpretationreasoning

    Article 41(2c) focuses on a right to good administration, where 'administration' refers to the institutions, bodies, offices and agencies of the Union, and is therefore of no help to those seeking reasons for decisions made by insurance companies, banks, or private educational facilities.

    “However, this article focuses on a right to good administration, where”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    Even if transparency has a distinct moral value in serving autonomy, it should not be pursued at all costs.

    “Even if transparency has a distinct moral value in serving autonomy, it should not be pursued at all costs.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    There is still plenty of legislative work to be done before we can feel safe that a right to explanation is legally protected.

    “Even so, there is still plenty of legislative work to be done before we can feel safe that such a right is legally protected.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given

    Other concerns with AI opacity, such as assessing reliability and fairness, will require us to open the black box.

    “This is not to say that other concerns with AI opacity, such as assessing reliability and fairness will not require us to open the black box.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

    Premisenone given

    Previous research indicated that transparency is conducive to negotiability and accountability, helps fine-tune reliability and avoid bias, and that we are more likely to trust the results of algorithms if we understand what causes their outcomes.

    “Previous research indicated that transparency is conducive to negotiability and accountability, helps fine-tune reliability and avoid bias, and that we are more likely to trust the results of algorithms if we understand what causes their outcomes.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    If we only use transparent algorithms, we might come to see that some unexpected or intractable reasons feature in the production of the decision, and in the long run this might incite us to use less reliable mechanisms that only employ reasons and inferences tractable to us.

    “If we only use transparent algorithms, we might come to see that some unexpected or simply intractable reasons feature in the production of the decision.In the long run, this might incite us to use less reliable mechanisms that only employ reasons and inferences that are tractable to us.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    The AI act document explicitly distinguishes between users and affected persons (p. 30).

    “This is despite the fact that the document explicitly distinguishes between users and affected persons (p.30).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites AI act

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    There are several reasons to demand that impactful decision algorithms are transparent, regardless of whether they are implemented by humans or AI systems.

    “There are several reasons to demand that impactful decision algorithms are transparent.This is true regardless of whether they are implemented by humans or AI systems.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    The kinds of explanations required to respect our autonomy can be hard to come by.

    “On the other hand, it appears that resolving previous concerns about opacity will not automatically address the threat to the autonomy of the affected parties, and the kinds of explanations required to respect our autonomy can be hard to come by.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

    Interpretationreasoning

    Respecting our autonomy might not require us to open the black box, but it does require us to take into account the perspectives of rational planning agents with life-goals and dreams.

    “It might not require us to open the black box, but it does require us to take into account the perspectives of rational planning agents with life-goals and dreams.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims. The quote only partly states this claim.

    Interpretationreasoning

    Affected parties must be able to convincingly argue that the decisions affecting them count as 'high risk', which risks putting an undue burden of proof on victims who do not have the means to litigate in grey area cases.

    “This means that affected parties must be able to convincingly argue that the decisions affecting them count as”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    Such knowledge can play an integral part in planning and shaping our lives as rational, self-determining agents.

    “Such knowledge can play an integral part in planning and shaping our lives as rational, self-determining agents.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Premisenone given

    One of the oft-cited reasons for using automated decision algorithms is that they appear to discover patterns that are not readily appreciable by human observers.

    “In fact, one of the oft-cited reasons for using automated decision algorithms is that they appear to discover patterns that are not readily appreciable by human observers.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Premisenone given

    Extraction algorithms require a certain causal structure to the data set in order to be successful.

    “Admittedly, such extraction algorithms require a certain causal structure to the data set in order to be successful.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

    Interpretationreasoning

    If the broad consensus that autonomy carries moral weight is correct, then transparency is worth demanding.

    “If the broad consensus that autonomy carries moral weight is correct, this means transparency is worth demanding.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Autonomy carries moral weight.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    Nguyen's arguments about transparency limiting reasons appear to transfer quite easily to the automatized decision-making case.

    “While Nguyen does not focus on automatized decision-making, his arguments appear to transfer quite easily to the automatized case.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    In both the human and the AI case, we are interested in knowing the robust patterns of correlation that allow us to reliably affect and predict their outcomes.

    “In both the human and the AI case, we are interested in knowing the robust patterns of correlation that allow us to reliably affect and predict their outcomes.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The texts reviewed by Langer et al. (2021) focused on the autonomy of the users but made no mention of the autonomy of the affected parties.

    “This oversight is reminiscent of the general oversight in the texts reviewed byLanger et al. (2021), which focused on the autonomy of the users, but made no mention of the autonomy of the affected parties (see Section 3).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Langer et al. (2021)

      Not checked in the free run.

48 of 48

  • The authors state this as an assumption.

    Understanding a decision-making process requires a certain amount of transparency.

    “I am assuming here that understanding a decision-making process requires a certain amount of transparency”

  • The paper relies on this without support.

    Autonomy carries moral weight.

    “If the broad consensus that autonomy carries moral weight is correct, this means transparency is worth demanding.”

  • The paper relies on this without support.

    Respecting personal autonomy requires providing adequate explanations for potentially life-changing decisions.

    “If respecting personal autonomy requires providing adequate explanations for potentially life-changing decisions, then the right to personal autonomy requires a right to explanation.”

  • The paper relies on this without support.

    For each individual case, the inputs cause the outputs, and not vice versa.

    “We know, for example, that for each individual case, the inputs cause the outputs, and not vice versa.”
4 of 4

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?

  • Widespread use of RLHF fine-tuned LLMs can lead to linguistic use being flattened on the characteristic style of such systems, making linguistic usage less diverse, less authentic, and less adequate for capturing the expressive practices and needs of different communities.

    “Widespread use of RLHF fine-tuned LLMs can in addition lead to linguistic use being flattened on the characteristic style of such systems, making linguistic usage less diverse, less authentic, and less adequate for capturing the expressive practices and needs of different communities (which in their turn bring yet other risks to their autonomy, cf. Fricker 2010; Vaassen 2022).”

    From Helpful, harmless, honest? Sociotechnical limits of AI alignment and safety through Reinforcement Learning from Human Feedback (2025) · cites this paper as Vaassen 2022

    Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

    Experimental — accuracy not yet measured

  • Legal and ethical challenges concerning data privacy and patient consent are highlighted by authors of [28], who stress the importance of robust legal frameworks adaptable to evolving AI.

    “The legal and ethical challenges, particularly concerning data privacy and patient consent, are highlighted by the authors of [28], who stress the importance of robust legal frameworks that are adaptable and responsive to the evolving nature of AI applications.”

    From Balancing Privacy and Progress: A Review of Privacy Challenges, Systemic Oversight, and Patient Perceptions in AI-Driven Healthcare (2024) · cites this paper as [28]

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.

  • GDPR and AI Act citation years are reversed

    Minor

    Section 4.2 cites the GDPR as the 2021 document and the AI Act as the 2016 document. The reference list identifies the GDPR as the 2016 document and the AI Act as the 2021 document.

1 of 1

Cites

Flagged first, then by citations

Research works printed but not linked: 8

  1. Loewer, B.M. (2007). Mental causation, or something near enough. In B.P. McLaughlin J.D. Cohen (Eds.) Contemporary debates in philosophy of mind (pp. 243–64). Blackwell.
  2. MacIntyre, A.C. (1983). After virtue: A study in moral theory. University of Notre Dame Press.
  3. Mill, J.S. (1999). On Liberty. Broadview Press.
  4. O’Neill, O. (2002). A question of trust: The BBC Reith lectures 2002. Cambridge: Cambridge University Press.
  5. Raz, J. (1986). The morality of freedom. New York: Oxford University Press.
  6. Selbst, A.D., & Barocas, S. (2018). The intuitive appeal of explainable machines. Fordham. L Reviews, 87, 1085.
  7. Tubella, A.A., Theodorou, A., Dignum, V., & Dignum, F. (2019). Governance by glass-box: Implementing transparent moral bounds for AI behaviour. arXiv:1905.04994 · arXiv
  8. Wachter, S., Mittelstadt, B., & Russell, C. (2017). Counterfactual explanations without opening the black box: Automated decisions and the GPDR. Harv. JL & Tech., 31, 841.

Cited by

Sorted by citations

Related

Authors
Vaassen, B.
Affiliations
Rutgers, The State University of New Jersey (US)
Corresponding author
Vaassen, Bram
ORCID
Vaassen, Bram
Venue
Philosophy & Technology
Publisher
Springer Science and Business Media LLC
Published
2022-09-22
Volume
35
Issue
4
ISSN
  • 2210-5433 (ISSN-L)
  • 2210-5441
DOI
10.1007/s13347-022-00577-5
OpenAlex
W4296638419
Semantic Scholar
0aa23f751523de433205baebce2a33e7af965198
Cited by
94
References
79
Citations by year
2026: 40 · 2025: 29 · 2024: 13 · 2023: 12
FWCI
15.64
Type
journal article
Language
en
Open access
Licence
Locations
  • Philosophy & Technology (publishedVersion, cc-by)
  • arXiv (Cornell University) (submittedVersion)
  • PhilPapers (PhilPapers Foundation) (submittedVersion)
  • DiVA at Umeå University (Umeå University) (submittedVersion, other-oa)
  • arXiv (Cornell University) (cc-by)
Funders
  • Deutsche Forschungsgemeinschaft (FOR 2495)
  • Vetenskapsrådet (2019-6381)
  • Umea University
Topics
  • Social Sciences › Safety Research › Ethics and Social Impacts of AI
  • Physical Sciences › Computer Science › Artificial Intelligence › Explainable Artificial Intelligence (XAI)
  • Social Sciences › Law › Digitalization, Law, and Regulation
Keywords
algorithmic opacity, individual autonomy, algorithmic transparency, causal explanation, algorithmic decision-making, algorithmic fairness

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.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-09.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-09.

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