On the Quest for Effectiveness in Human Oversight: Interdisciplinary Perspectives

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

Human oversight has been subject to criticism, with concerns about the efficacy of oversight in practice and its potential role as a mere vehicle to legitimize imperfect AI systems.

“Human oversight has also been subject to criticism, with concerns being raised about the efficacy of oversight in practice and its potential role as a mere vehicle to legitimize imperfect AI systems [31, 89].”

• cites [31] [31] Ben Green. 2022. The flaws of policies requiring human oversight of government algorithms. Computer Law & Security Review 45 (2022), 105681. https://doi.org/ 10.1016/j.clsr.2022.105681 - sourceSupported
• cites [89] [89] John Zerilli, Alistair Knott, James Maclaurin, and Colin Gavaghan. 2019. Algorithmic Decision-Making and the Control Problem. Minds and Machines 29, 4 (Dec. 2019), 555-578. https://doi.org/10.1007/s11023-019-09513-7 sourcePartly supported

  The Introduction and Conclusion describe how human supervisors can become complacent and hand effective control to an algorithm, supporting concern about oversight’s practical efficacy. The passages do not say that human oversight serves to legitimize imperfect AI systems; the discussion of an unvetted system concerns an assumption that the system itself had been validated. Page numbers are not available in our copy. (LLM, conf=0.85)

  Experimental — accuracy not yet measured

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

The main objective of human oversight is risk mitigation.

“Based on the claim that the main objective of human oversight is risk mitigation, we propose a viable understanding of effectiveness in human oversight”

• Also stated in: § 3.1 An objective-first approach to the effectiveness of human oversight · p. 3 · hedged

  “As we have discussed above, risk mitigation can be taken to be one of the main objectives of human oversight.”

• cites [19] [19] Lena Enqvist. 2023. 'Human oversight'in the EU artificial intelligence act: what, when and by whom? Law, Innovation and Technology 15, 2 (2023), 508-535. - sourceSupported
• cites [22] [22] European Parliament. 2024. Corrigendum to the position of the European Parliament adopted at first reading on 13 March 2024 with a view to the adoption of Regulation (EU) 2024/ ...... of the European Parliament and of the Council laying down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 (Artificial Intelligence Act) P9\_TA(2024)0138 (COM(2021)0206 - C9-0146/2021 2021/0106(COD)). https://www.europarl.europaNot checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

Empirical evidence suggests that humans may be ineffective in supervising AI-based systems.

“as empirical evidence suggests that humans may be ineffective in supervising AI-based systems [31].”

• cites [31] [31] Ben Green. 2022. The flaws of policies requiring human oversight of government algorithms. Computer Law & Security Review 45 (2022), 105681. https://doi.org/ 10.1016/j.clsr.2022.105681 - sourceSupported

May combine two claims.

Forcing people to think about a case themselves for some time before making a final decision partly counters anchoring effects, though it leads to less acceptance of the system.

“For example, although this leads to less acceptance of the system, it seems to partly counter such anchoring effects if people are forced to think about a given case themselves for some time before making a final decision [11].”

• cites [11] [11] Zana Buçinca, Maja Barbara Malaya, and Krzysztof Z. Gajos. 2021. To Trust or to Think: Cognitive Forcing Functions Can Reduce Overreliance on AI in AI-assisted Decision-making. Proceedings of the ACM on Human-Computer Interaction 5, CSCW1 (April 2021), 1-21. https://doi.org/10.1145/3449287 - sourceSupported

May combine two claims.

The standardization process is still in its early stages and limited to a few general guidelines.

“However, this standardization process is still in its early stages and is limited to a few general guidelines [49].”

• cites [49] [49] Johann Laux. 2023. Institutionalised distrust and human oversight of artificial intelligence: towards a democratic design of AI governance under the European Union AI Act. AI & SOCIETY (Oct. 2023). https://doi.org/10.1007/s00146-02301777-z - sourceSupported

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing). May combine two claims.

Measures in accordance with (b) seem out of place because they introduce a concrete psychological aspect that sticks out from more technical and operational aspects, especially since automation bias is just one of many relevant biases, heuristics, and psychological phenomena.

“Measures in accordance with (b), which are intended to maintain an awareness of the automation bias, seem out of place: it introduces a concrete psychological aspect that sticks out from the more technical and operational aspects outlined in the other provisions, especially considering that automation bias is just one of many relevant biases (e.g., hindsight bias, confidence bias [36, 66]), heuris”

• cites [36] [36] Martin Hilbert. 2012. Toward a synthesis of cognitive biases: How noisy information processing can bias human decision making. Psychological Bulletin 138, 2 (March 2012), 211-237. https://doi.org/10.1037/a0025940 - sourceNot checked — no full text of the source
• cites [66] [66] Neal J. Roese and Kathleen D. Vohs. 2012. Hindsight Bias. Perspectives on Psychological Science 7, 5 (Sept. 2012), 411-426. https://doi.org/10.1177/ 1745691612454303 - sourceNot checked — no full text of the source
• cites [83] [83] Amos Tversky and Daniel Kahneman. 1973. Availability: A heuristic for judging frequency and probability. Cognitive Psychology 5, 2 (Sept. 1973), 207-232. https: //doi.org/10.1016/0010-0285(73)90033-9 - sourceNot checked — no full text of the source
• cites [68] [68] Richard M. Ryan and Edward L. Deci. 2000. Intrinsic and Extrinsic Motivations: Classic Definitions and New Directions. Contemporary Educational Psychology 25, 1 (Jan. 2000), 54-67. https://doi.org/10.1006/ceps.1999.1020 - sourceNot checked — no full text of the source
• cites [73] [73] Sabine Sonnentag, Louis Tay, and Hadar Nesher Shoshan. 2023. A review on health and well-being at work: More than stressors and strains. Personnel Psychology 76, 2 (Jan. 2023), 473-510. https://doi.org/10.1111/peps.12572 - sourceCould not verify

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

  Experimental — accuracy not yet measured

Classified as probably not a substantial claim.

Risk mitigation is defined as a key objective of human oversight measures that should contribute to the overall trustworthiness of high-risk systems.

“Risk mitigation is therefore defined as a key objective of human oversight measures that should contribute to the overall trustworthiness of high-risk systems [19].”

• cites [19] [19] Lena Enqvist. 2023. 'Human oversight'in the EU artificial intelligence act: what, when and by whom? Law, Innovation and Technology 15, 2 (2023), 508-535. - sourceSupported

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

If oversight persons receive AI output after making up their own mind, they are not initially affected and may better judge appropriateness.

“Alternatively, oversight persons may only receive the AI-based output after they have made up their own mind about a specific case because then they would not initially be affected by the AIbased output and may be in a better position to judge whether the AI-based output is appropriate [31, 89].”

• cites [31] [31] Ben Green. 2022. The flaws of policies requiring human oversight of government algorithms. Computer Law & Security Review 45 (2022), 105681. https://doi.org/ 10.1016/j.clsr.2022.105681 - sourcePartly supported

  Section 4.1.3 says algorithmic output can alter how people decide, supporting the concern about its influence on an initial judgment. But neither that section nor §4.1.2 evaluates showing the output only after a person has formed a judgment, so the claimed improvement in judging its appropriateness is not established by this paper. (LLM, conf=0.76)

  Experimental — accuracy not yet measured
• cites [89] [89] John Zerilli, Alistair Knott, James Maclaurin, and Colin Gavaghan. 2019. Algorithmic Decision-Making and the Control Problem. Minds and Machines 29, 4 (Dec. 2019), 555-578. https://doi.org/10.1007/s11023-019-09513-7 sourceSupported

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

The requirement for effective oversight is particularly highlighted in the European AI Act.

“The requirement for effective oversight is particularly highlighted in the European AI Act.”

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

Conditions (a)-(c) (causal power, epistemic access, self-control) are jointly sufficient for moral responsibility.

“We point out that (a)-(c) are jointly sufficient for moral responsibility.”

• Also stated in: § 3.2 Further considerations · p. 6 · plain

  “an oversight person is morally responsible for risk mitigation (or any other part of their job) if and only if they have (1) causal power, (2) epistemic access, and (3) self-control in that respect.”

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

Competence is not part of the definition of effectiveness, but a competent oversight person is more likely to be effective.

“there is good reason why competence is not part of our definition of effectiveness”

• Also stated in: § 3.2 Further considerations · p. 6 · hedged

  “A competent oversight person will be more likely to establish and retain sufficient epistemic access, self-control, and to use their causal powers in the right way.”

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

An oversight person is effective if and only if they are morally responsible and have fitting intentions.

“Furthermore, we argue that this is equivalent to saying that an oversight person is effective if and only if they are morally responsible and have fitting intentions.”

• Also stated in: § 1 INTRODUCTION · p. 2 · plain

  “Effectiveness = Moral Responsibility + Fitting Intentions.”

• Also stated in: § ABSTRACT · p. 1 · plain

  “An oversight person is effective if and only if they are morally responsible and have fitting intentions.”

Classified as probably not a substantial claim.

Most conceptions of moral responsibility share three aspects: a causal connection between the agent and the event, sufficient knowledge of the decision situation, and a certain autonomy of the agent.

“Most conceptions of moral responsibility have a focus on three aspects in common: a causal connection between the agent and the event for which the agent might be responsible, the agent having sufficient knowledge of their decision situation, and a certain autonomy of the agent [26, 59, 82].”

• cites [26] [26] John Martin Fischer and Mark Ravizza. 1998. Responsibility and Control: A Theory of Moral Responsibility . Cambridge University Press, New York. - sourceNot checked — no full text of the source
• cites [59] [59] Merel Noorman. 2020. Computing and Moral Responsibility. In The Stanford Encyclopedia of Philosophy (Spring 2020 ed.), Edward N. Zalta (Ed.). Metaphysics Research Lab, Stanford University. -Not checked yet

  Not checked in the free run.

• cites [82] [82] Matthew Talbert. 2019. Moral Responsibility. In The Stanford Encyclopedia of Philosophy (Winter 2019 ed.), Edward N. Zalta (Ed.). Metaphysics Research Lab, Stanford University. -Not checked yet

  Not checked in the free run.

May combine two claims.

The paper's conceptualization of effective human oversight provides a more general and more practically useful conceptualization compared to the respective stipulations in the AI Act.

“we argue that our conceptualization provides both a more general and a more practically useful conceptualization of effective human oversight compared to the respective stipulations in the AI Act.”

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

Human-system decision accuracy is higher when the human choice is restricted to the outputs that are in the proposed set.

“It was shown that for certain tasks, the human-system decision accuracy indeed is higher when the human choice is restricted to the outputs that are in the proposed set [78].”

• cites [78] [78] Eleni Straitouri and Manuel Gomez Rodriguez. 2023. Designing Decision Support Systems Using Counterfactual Prediction Sets. CoRR abs/2306.03928 (2023). https://doi.org/10.48550/ARXIV.2306.03928 arXiv:2306.03928 - sourceNot checked yet

  Checked before — its earlier verdict is held back until a new check runs.

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

The intentional vagueness in the definition of effective human oversight (e.g., 'sufficient', 'fitting') is an essential feature, not a drawback.

“This intentional vagueness is not a drawback but rather an essential feature.”

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

The definition and conditions for achieving effective human involvement remain underexplored, highlighting a critical gap in the current state of the art of AI governance.

“Nevertheless, the definition and conditions for achieving effective human involvement remain underexplored, highlighting a critical gap in the current state of the art of AI governance.”

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

Automation bias is a likely-to-observe phenomenon where people tend to use the outcome of an automated decision entity as a heuristic replacement for vigilant information seeking and processing.

“automation bias (cf. Section 4.2) is a likely-to-observe phenomenon where people tend to use the outcome of an automated decision entity 'as a heuristic replacement for vigilant information seeking and processing' ([56] p. 205).”

• cites [56] [56] Kathleen L. Mosier, Linda J. Skitka, Mark D. Burdick, and Susan T. Heers. 1996. Automation Bias, Accountability, and Verification Behaviors. Proceedings of the Human Factors and Ergonomics Society Annual Meeting 40, 4 (Oct. 1996), 204-208. https://doi.org/10.1177/154193129604000413 - sourceNot checked — no full text of the source

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

The European AI Act mandates human oversight for high-risk AI applications.

“the European AI Act [22], which mandates human oversight for high-risk AI applications.”

• cites [22] [22] European Parliament. 2024. Corrigendum to the position of the European Parliament adopted at first reading on 13 March 2024 with a view to the adoption of Regulation (EU) 2024/ ...... of the European Parliament and of the Council laying down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 (Artificial Intelligence Act) P9\_TA(2024)0138 (COM(2021)0206 - C9-0146/2021 2021/0106(COD)). https://www.europarl.europaNot checked yet

  Not checked in the free run.

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

Research reports mixed results regarding human oversight.

“Overall, research reports mixed results regarding human oversight.”

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

Effectiveness in human oversight is not a binary concept but a matter of degrees.

“Moreover, 'effectiveness' is not a binary concept but a matter of degrees: one exercise of oversight might be more effective than another.”

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

An oversight person is effective if and only if they have causal power, epistemic access, self-control, and fitting intentions.

“we propose a viable understanding of effectiveness in human oversight: for human oversight to be effective, the oversight person has to have (a) sufficient causal power with regard to the system and its effects, (b) suitable epistemic access to relevant aspects of the situation, (c) self-control, and (d) fitting intentions for their role.”

• Also stated in: § 1 INTRODUCTION · p. 2 · plain

  “we propose that an oversight person is effective if and only if they have (a) sufficient causal power with regards to the system and its effects, (b) suitable epistemic access to relevant aspects of the situation, (c) self-control over their own actions, and (d) fitting intentions for their role.”

• Also stated in: § 3.1 An objective-first approach to the effectiveness of human oversight · p. 5 · plain

  “Definition 3.5 (Effectiveness). An oversight person is effective in their human oversight if and only if they have causal power , epistemic access , self-control , and fi tting intentions in the above senses.”

• AssumptionThe authors state this as an assumption: The oversight person has certain basic knowledge, such as what a stop button is or how to operate a computer on a consumer-level.
• AssumptionThe authors state this as an assumption: Human oversight should not be an end in itself.

Classified as probably not a substantial claim.

As stated in Article 14(2), oversight shall aim to prevent or minimize risks to health, safety, or fundamental rights.

“As stated in Article 14(2), this oversight shall aim to prevent or minimize risks to health, safety, or fundamental rights.”

• cites [22] [22] European Parliament. 2024. Corrigendum to the position of the European Parliament adopted at first reading on 13 March 2024 with a view to the adoption of Regulation (EU) 2024/ ...... of the European Parliament and of the Council laying down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 (Artificial Intelligence Act) P9\_TA(2024)0138 (COM(2021)0206 - C9-0146/2021 2021/0106(COD)). https://www.europarl.europaNot checked yet

  Not checked in the free run.

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

If both an oversight person and the deployer had causal power, epistemic access, and self-control with regard to a certain risk, they would both be responsible.

“If, for example, both an oversight person and the deployer had causal power, epistemic access, and self-control with regard to a certain risk, they would both be responsible.”

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

Our approach is well suited to assist standard-setting organizations and to clarify the broad requirements of Article 14(4).

“Our approach is, therefore, well suited to assist standard-setting organizations and to clarify the rather broad requirements of Article 14(4).”

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

An oversight person who lacks self-control might also simultaneously lack sufficient causal power or epistemic access.

“An oversight person who lacks self-control might also simultaneously lack sufficient causal power or epistemic access (e.g., in case of a seizure).”

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

Highly conscientious oversight persons will arguably be more likely to maintain self-control, sufficient epistemic access, and fitting intentions.

“Highly conscientious oversight persons will arguably be more likely to maintain self-control, sufficient epistemic access, and fitting intentions.”

Classified as probably not a substantial claim.

The paper identified facilitators and inhibitors of effectiveness in three categories: the technical design of the system, individual characteristics of oversight persons, and the environmental circumstances in which oversight occurs.

“We have identified facilitators and inhibitors of effectiveness in three categories, namely the technical design of the system, individual characteristics of oversight persons, and the environmental circumstances in which oversight occurs.”

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

Our approach can assist courts in the future by providing a framework for defining and assessing the effectiveness of human oversight.

“Finally, our approach can assist courts in the future by providing a framework for defining and assessing the effectiveness of human oversight.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

Dealing with the abstraction and vagueness issues in complex scenarios requires involving individuals with extensive expertise in the task at hand and fostering collaboration in multidisciplinary teams.

“Dealing with these issues, especially in complex scenarios, requires (a) involving individuals with extensive expertise in the task at hand and (b) fostering collaboration in multidisciplinary teams to assess the impact on human oversight practice.”

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

Our proposed framework provides guidance to practitioners who do not wish to rely solely on technical standards.

“In addition, our proposed framework provides guidance to practitioners who do not wish to rely solely on technical standards.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing). May combine two claims.

Learned carelessness can be interpreted as a problem of self-control or as an issue that leads to intentions that no longer fit the duties of an oversight person.

“This phenomenon could be interpreted as a problem of self-control or as an issue that leads to intentions that no longer fit the duties of an oversight person.”

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

The AI Act could benefit from incorporating more nuanced and structured conditions, such as those proposed in this paper.

“Our analysis suggests that the AI Act could benefit from incorporating more nuanced and structured conditions, such as those proposed in this paper.”

Classified as probably not a substantial claim.

The measures in Article 14(4) reflect our conditions of effectiveness.

“These measures reflect our conditions of effectiveness.”

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

Adaptations that help oversight persons acquire and process information better increase their epistemic access.

“This increases the oversight person's epistemic access if the adaptations help him to acquire and process the information better.”

Classified as probably not a substantial claim.

Article 14(3) obliges the provider to enable its deployer to perform effective oversight by identifying 'appropriate measures'.

“Article 14(3) obliges the provider to enable its deployer to perform effective oversight by identifying 'appropriate measures'.”

• cites [22] [22] European Parliament. 2024. Corrigendum to the position of the European Parliament adopted at first reading on 13 March 2024 with a view to the adoption of Regulation (EU) 2024/ ...... of the European Parliament and of the Council laying down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 (Artificial Intelligence Act) P9\_TA(2024)0138 (COM(2021)0206 - C9-0146/2021 2021/0106(COD)). https://www.europarl.europaNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing). May combine two claims.

This streamlining may decrease the legal uncertainty in Article 14, fostering a more consistent and predictable environment for AI providers and deployers.

“This streamlining may also decrease the legal uncertainty in Article 14, fostering a more consistent and predictable environment for AI providers and deployers alike, while providing a practical definition of effectiveness for deciding courts.”

Probably not a substantial claim: may not pass S1 (asserted), S3 (load-bearing), S4 (based). May combine two claims.

A felt obligation to explain oneself will motivate maintaining high self-control and incentivize upholding fitting intentions.

“It can also be hypothesized that a felt obligation to explain oneself will motivate to maintain high self-control in order to keep up with the expectations of the job and that it will incentivize one to uphold intentions that fit one's duties.”

May combine two claims.

Self-control may be an antecedent of causal power and epistemic access, but not vice versa.

“In fact, self-control may, in many situations, be an antecedent of causal power and epistemic access - but not vice versa.”

Classified as probably not a substantial claim.

Article 26(2) emphasizes the need for competence, training, authority, and support in oversight functions, hence requiring epistemic access and causal power.

“Furthermore, Article 26(2) [22] emphasizes the need for the necessary competence, training, authority, and support in oversight functions, hence requiring epistemic access and causal power.”

• cites [22] [22] European Parliament. 2024. Corrigendum to the position of the European Parliament adopted at first reading on 13 March 2024 with a view to the adoption of Regulation (EU) 2024/ ...... of the European Parliament and of the Council laying down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 (Artificial Intelligence Act) P9\_TA(2024)0138 (COM(2021)0206 - C9-0146/2021 2021/0106(COD)). https://www.europarl.europaNot checked yet

  Not checked in the free run.

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

Defining effectiveness using our four conditions provides a practical framework that highlights the key aspects of effective oversight.

“Defining effectiveness using our four conditions provides a practical framework that highlights the key aspects of effective oversight.”

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

This principle can be understood as a condition for a minimum level of effectiveness of the oversight measures implemented, comparable to a threshold that must be reached.

“This principle can be understood as a condition for a minimum level of effectiveness of the oversight measures implemented, comparable to a threshold that must be reached.”

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

Automation bias targeted in Article 14(4)(b) may interfere with the fittingness of the oversight person's intentions.

“However, one might consider that the automation bias targeted in Article 14(4) (b) may interfere with the fittingness of the oversight person's intentions (cf. Section 4.2).”

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

The definition of effectiveness will be one of the key tasks of the deciding courts in the coming years.

“the definition of effectiveness will be one of the key tasks of the deciding courts in the coming years, leaving these courts also in need of further clarification on the conditions of effectiveness.”

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

For some facilitators (or inhibitors), which of the four conditions they contribute to (or hinder) depends on the concrete context.

“for some of the facilitators (or inhibitors), the question of which of the four conditions they contribute to (or hinder) depends on the concrete context”

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

The paper is helpful in determining whether an instance of human oversight is effective but less helpful in assessing whether something is human oversight.

“while our paper is helpful in determining whether an instance of human oversight is effective, it is less helpful in assessing whether something is human oversight”

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

The breadth and structure of our conceptualization simplifies the assessment process for regulators.

“One of the benefits of the breadth and structure of our conceptualization is that it simplifies the assessment process for regulators, making it easier to evaluate different AI systems against a structured set of principles that are framed by the four conditions.”

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

Our conceptualization is designed for broad applicability across various AI-based systems and shares the AI Act's criticism of residing on a high level of abstraction and of vagueness.

“our conceptualization is designed for broad applicability across various AI-based systems, and thus shares a common criticism with the AI Act, namely that of residing on a high level of abstraction and that of vagueness”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing). May combine two claims.

The job of an oversight person needs to be well designed, meaning it is motivating, satisfying, and not overly demanding.

“The job of an oversight person needs to be well designed [63], which means that the job is, e.g., motivating, satisfying, and not overly demanding.”

• cites [63] [63] Sharon K. Parker and Gudela Grote. 2020. Automation, Algorithms, and Beyond: WhyWorkDesignMatters More Than Ever in a Digital World. Applied Psychology 71, 4 (Feb. 2020), 1171-1204. https://doi.org/10.1111/apps.12241 - sourceNot checked yet

  Not checked in the free run.

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

Article 14(4)'s vagueness and varying levels of abstraction are insufficient for determining true effectiveness in specific cases.

“While Article 14(4) offers some guidance, its vagueness and varying levels of abstraction are insufficient for determining true effectiveness in specific cases.”

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

Without a concrete idea of what effectiveness is in the context of human oversight, no productive discussion of it is possible.

“Without a concrete idea of what effectiveness is in the context of human oversight, no productive discussion of it is possible.”

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

The paper did not discuss the concept of human oversight itself in much detail, focusing only on the effectiveness of human oversight.

“we did not discuss the concept of human oversight itself in much detail, but focused only effectiveness of human oversight”

Classified as probably not a substantial claim.

The effectiveness of human oversight, as outlined in Article 14(1), serves as a foundational principle that requires oversight measures to be capable of achieving their risk mitigation objective.

“Consequently, the effectiveness of human oversight, as outlined in Article 14(1), serves as a foundational principle that requires oversight measures to be capable of achieving their risk mitigation objective.”

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

The definition does not allow for group responsibility but can be adapted to do so.

“For reasons of simplicity, the definition does not allow for group responsibility but can be adapted to do so.”

Classified as probably not a substantial claim.

In the event of a legal dispute, courts will have to examine whether the specific design of human oversight measures meets the overarching requirement of effectiveness in Article 14(1).

“In the event of a legal dispute, the courts will therefore have to examine whether the specific design of the human oversight measures meets the overarching requirement of effectiveness in Article 14(1).”

• AssumptionRests on an assumption: Technical standards created for the AI Act only have a presumptive effect for providers and deployers of AI systems that are subject to the AI Act.

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

Overseer training could, in principle, target all of the conditions of effective human oversight.

“Overseer training could, in principle, target all of the conditions of effective human oversight.”

• AssumptionRests on an assumption: Overseer training could, in principle, target all of the conditions of effective human oversight.

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

For an oversight person to have causal power over the system, there must be options that allow the oversight person to intervene with, control, overwrite, or undo system decisions and actions.

“For an oversight person to have causal power over the system, there must be options that allow the oversight person to intervene with, control, overwrite, or undo system decisions and actions.”

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

Understanding the system's input can be equally crucial as understanding the system's output for epistemic access.

“while understanding the system's output (c) can enhance epistemic access, understanding of the system's input can be equally crucial”

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

To ensure uniform application of the AI Act by lower courts, a comprehensible definition of effective human oversight is essential.

“To ensure uniform application of the AI Act by lower courts, a comprehensible definition of effective human oversight is essential.”

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

The four conditions of effectiveness could form a pivot for future research on human oversight.

“We, therefore, believe that these conditions could form a pivot for future research on human oversight.”

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

The points listed in Table 1 can be understood as examples of how an assessment of the effectiveness of measures could be carried out.

“the points listed in Table 1 can be understood as examples of how an assessment of the effectiveness of measures could be carried out”

Classified as probably not a substantial claim.

If an oversight person had ill intent, they could use their position to foster risks instead of mitigating them.

“Even worse, if an oversight person had ill intent, they could use their position to foster risks instead of mitigating them.”

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

For certain cases, the mapping from facilitators and inhibitors to conditions needs considerable empirical research, outside the scope of this conceptual paper.

“for certain cases, the mapping from facilitators and inhibitors to conditions needs considerable empirical research, outside the scope of this conceptual paper”

Classified as probably not a substantial claim.

In the context of human oversight, motivation is crucial to initiate and maintain fitting intentions and self-control for the goal of trying to mitigate risk.

“In the context of human oversight, motivation is crucial to initiate and maintain fitting intentions and self-control for the goal of trying to mitigate risk.”

• AssumptionRests on an assumption: Motivation leads people to initiate, intensify, and persist with goal-directed behavior.
• AssumptionRests on an assumption: Motivation is often influenced by individual factors, environmental factors, and the perceived significance of the desired outcomes.

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

Our definition of effectiveness is broad enough to be applied to many roles along the AI pipeline, not only to oversight persons.

“our definition of effectiveness is broad enough to be applied to many roles along the AI pipeline, not only to oversight persons”

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

By defining boundary conditions or by using oversight support tools, oversight persons could gain better epistemic access when only having to review a subset of outputs.

“By defining boundary conditions or by using oversight support tools, oversight persons could gain better epistemic access when only having to review a subset of outputs.”

May combine two claims.

If oversight persons feel accountable, this may influence their attentiveness or lead to feeling more stressed.

“If oversight persons feel accountable for their actions this may influence their attentiveness or may lead to them feeling more stressed [35].”

• cites [35] [35] Angela T. Hall, Dwight D. Frink, and M. Ronald Buckley. 2015. An accountability account: A review and synthesis of the theoretical and empirical research on felt accountability. Journal of Organizational Behavior 38, 2 (Sept. 2015), 204-224. https://doi.org/10.1002/job.2052 - sourceNot checked — no full text of the source

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

Interpretability aims at improving epistemic access.

“Interpretability thus aims at improving epistemic access.”

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

In a hiring scenario, a conflict arises when addressing unfairness requires inviting additional less suitable candidates, conflicting with the HR manager's goal of selecting the best candidates only.

“A conflict arises when the only method to address unfairness involves inviting additional, apparently less suitable candidates, conflicting with the HR manager's goal of selecting the best candidates only.”

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

The job characteristics model highlights that skill variety, task identity, perceived task significance, autonomy, and feedback promote motivation and satisfaction with the job.

“For example, the job characteristics model from work psychology [33] highlights that certain job characteristics promote motivation and satisfaction with the job, namely skill variety, task identity, perceived task significance, autonomy, and feedback from the job.”

• cites [33] [33] J.Richard Hackman and Greg R. Oldham. 1976. Motivation through the design of work: test of a theory. Organizational Behavior and Human Performance 16, 2 (Aug. 1976), 250-279. https://doi.org/10.1016/0030-5073(76)90016-7 - sourceNot checked — no full text of the source

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

If an oversight person were unwilling to make the efforts necessary to mitigate a risk, they would not be effective in their oversight even if they had causal power, epistemic access, and self-control.

“If the oversight person were unwilling to make the efforts necessary to mitigate a risk, they would not be effective in their oversight - even if they had causal power, epistemic access, and self-control.”

May combine two claims.

Anchors make oversight persons selectively aware of anchor-consistent information, which undermines epistemic access.

“Such anchors make oversight persons selectively aware of anchor-consistent information [18] which undermines epistemic access.”

• cites [18] [18] Birte Englich, Thomas Mussweiler, and Fritz Strack. 2006. Playing Dice With Criminal Sentences: The Influence of Irrelevant Anchors on Experts' Judicial Decision Making. Personality and Social Psychology Bulletin 32, 2 (Feb. 2006), 188-200. https://doi.org/10.1177/0146167205282152 - sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

The three domains of facilitators and inhibitors can aid in giving further structure for certain use cases.

“The three domains of facilitators and inhibitors can aid in giving further structure for certain use cases.”

May combine two claims.

How and which requirements apply depends heavily on the individual case and system design.

“Not only how these requirements apply, but also which requirements are applicable depends heavily on the individual case and the design of the system.”

May combine two claims.

The AI Act's approach provides flexibility but may also be associated with greater legal uncertainty.

“Although AI Act's approach provides flexibility that accommodates the sometimes beneficial vagueness of the effectiveness of human oversight also mentioned in Section 3.1.5, it may also be associated with greater legal uncertainty.”

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

For many relevant oversight contexts, the four conditions of effectiveness seem to provide a better picture of when and how human oversight is useful.

“For many relevant oversight contexts, the four conditions of effectiveness seem to provide a better picture of when and how human oversight is useful.”

Classified as probably not a substantial claim.

According to Article 14(1) of the AI Act proposal, high-risk AI systems shall be designed and developed in such a way that they can be effectively overseen during their use.

“According to Article 14(1) of the proposal, what the AI Act defines as high-risk AI systems shall be designed and developed in such a way that they can be effectively overseen during their use.”

• cites [22] [22] European Parliament. 2024. Corrigendum to the position of the European Parliament adopted at first reading on 13 March 2024 with a view to the adoption of Regulation (EU) 2024/ ...... of the European Parliament and of the Council laying down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 (Artificial Intelligence Act) P9\_TA(2024)0138 (COM(2021)0206 - C9-0146/2021 2021/0106(COD)). https://www.europarl.europaNot checked yet

  Not checked in the free run.

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

Article 14(4) contains certain aspects of what the proposed AI Act envisages as indicators of the effectiveness of these measures.

“Article 14(4) contains certain aspects of what the proposed AI Act envisages as indicators of the effectiveness of these measures.”

• cites [22] [22] European Parliament. 2024. Corrigendum to the position of the European Parliament adopted at first reading on 13 March 2024 with a view to the adoption of Regulation (EU) 2024/ ...... of the European Parliament and of the Council laying down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 (Artificial Intelligence Act) P9\_TA(2024)0138 (COM(2021)0206 - C9-0146/2021 2021/0106(COD)). https://www.europarl.europaNot checked yet

  Not checked in the free run.

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

The account allows for shared responsibility between multiple agents.

“This account also allows for shared responsibility between multiple agents, which is desirable in the context of high-risk AI systems.”

Classified as probably not a substantial claim.

Lack of regulatory clarity could hinder practical applicability of Article 14 and introduction of effective human oversight.

“The lack of regulatory clarity could hinder the practical applicability of Article 14 and, thus, the introduction of truly effective human oversight.”

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

There may be cases where oversight tasks require a mechanism for selecting system outputs that are to be reviewed by the oversight person.

“There may be cases where oversight tasks require a mechanism for selecting system outputs that are to be reviewed by the oversight person.”

Classified as probably not a substantial claim.

Conditions (a) and (c) relate to the understanding of the system, requiring the provider to enable oversight persons to understand limitations, monitor effectively, and interpret results correctly.

“Conditions (a) and (c) relate to the understanding of the system, as they require the provider to enable the oversight person to understand the limitations of the system (though not necessarily all details of the system), to monitor it effectively, and to interpret its results correctly.”

Classified as probably not a substantial claim.

Requirements (d) and (e) underscore the EU's recognition of the importance of oversight persons being able to influence the AI system, either by modifying the system's output or by stopping its operation.

“Requirements (d) and (e) underscore the EU's recognition of the importance of oversight persons being able to influence the AI system, either by modifying the system's output or by stopping its operation.”

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

The requirement of a stop button in (e) is highly concrete; the requirement in (a) to create understanding of limitations is vague.

“For example, the requirement of a 'stop' button in (e) is highly concrete. In contrast, the requirement in (a) to create an understanding of the limitations of the system is vague and leaves a lot of room for interpretation.”

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

A feeling of accountability could promote efforts to achieve sufficient epistemic position.

“Thus, a feeling of accountability could promote efforts of oversight persons to get themselves in a sufficient epistemic position.”

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

Our conditions for effectiveness, when applied with expertise, make the need to understand the system's input evident, whereas this need is challenging to identify when relying solely on Article 14(4).

“Our conditions for effectiveness, when applied with expertise, make this need evident; whereas this need can be challenging to identify when relying solely on Article 14(4), which does not mention the system's input at all.”

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

Our objective-first approach to effectiveness can be used to integrate the piecemeal list of different requirements listed in the AI Act into a coherent system.

“Ourobjective-first approach to effectiveness can be used to integrate the piecemeal list of different requirements listed in the AI Act into a coherent system.”

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

Article 14(4) appears to be a loose collection of items without much structural consistency.

“Thus, Article 14(4) appears to be a loose collection of items that were deemed useful in the legislative process without much structural consistency.”

Classified as probably not a substantial claim.

The measures in Article 14(4) are of varying degrees of abstraction.

“It is striking that the measures appearing in Article 14(4) are of varying degrees of abstraction: While some measures contain specific requirements for the design of the AI system, other measures are less specific.”

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

The criterion of fitting intentions has little direct equivalent in the legal text, which focuses on actions not subjective intentions.

“The criterion of fitting intentions of the agents involved has little direct equivalent in the legal text, which is not focused on subjective intentions but on actions.”

Classified as probably not a substantial claim.

Requirements (d) and (e) manifest the need for both causal power and self-control.

“These requirements manifest the need for both causal power and self-control, ensuring that individuals can influence the decisions of the AI system, or decide not to use the system's output or to modify it.”

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

Our approach is advantageous for a legally sound application of the AI Act, as it helps to structure the requirements for and responsibilities of the different parties involved.

“This is advantageous for a legally sound application of the AI Act, as it helps to structure the requirements for and responsibilities of the different parties involved.”

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

The specific measures taken must be assessed on a case-by-case basis.

“As a result, the specific measures taken must be assessed on a case-by-case basis.”

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

Standardization organisations risk developing technical standards for the AI Act that may not withstand legal scrutiny.

“Thereby, the standardization organisations risk developing technical standards that may not withstand legal scrutiny.”

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

Training to prepare for or enhance the role of the oversight person will be crucial.

“Training to prepare for or enhance the role of the oversight person will be crucial.”

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

Accountability could lead to unintended consequences such as unfit intentions: if oversight persons believe they would have no good explanation for overriding a system output, they may not do so when they have to file a report.

“However, it could also lead to unintended consequences such as leading to unfit intentions: if oversight persons believe that they would have no good explanation for overriding a system output, they may not do so in cases where they know that they have to file a report after overwriting system outputs.”