Pith. sign in

REVIEW 3 major objections 6 minor 179 references

AI Safety for Everyone

T0 review · 3 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Across 383 peer-reviewed papers, AI safety is mostly concrete robustness work, not existential-risk theory.

desk verdict A useful empirical map of peer-reviewed AI safety work, but the relative-frequency claim about existential risk is partly an artifact of the search query. read the letter →

arxiv 2502.09288 v2 pith:YOOEDBHE submitted 2025-02-13 cs.CY

Balint Gyevnar , Atoosa Kasirzadeh This is my paper · ORCID
classification cs.CY
keywords AIsafetyexistentialrisksystematicliteraturereviewsafereinforcementlearningadversarialrobustnessinterpretabilitytechnologicalpluralism
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues, on empirical grounds, that AI safety is not predominantly the study of existential risk from future superintelligent AI. A systematic review of 383 papers indexed in major peer-reviewed databases shows that the bulk of AI safety research addresses concrete, near-term problems such as robustness to noise, adversarial attacks, unsafe exploration, monitoring, and system misspecification. The authors interpret this pattern as evidence that AI safety is a natural extension of traditional technological and systems safety, and they propose a pluralistic, epistemically inclusive conception of the field. The motivation is practical: narrowing AI safety to existential scenarios excludes many researchers, misleads the public, and can breed resistance to safety measures.

What carries the argument

The machinery is the systematic literature review pipeline: a keyword hierarchy (AI terms AND safe*/robust*/align* AND lifecycle terms) run on two major indexing databases, narrowed to 383 papers by title/abstract/full-text screening, enriched by inductive coding and snowball sampling. From that corpus the authors derive their eight risk types and ten methodology categories. The empirical distribution of those categories—not any single theorem—does the argumentative work: it shows near-term technical safety concerns dominate and that they map onto traditional systems-safety concerns such as reliability, control, and adversarial resilience.

What would settle it

Re-run the same three-pronged keyword query (AI AND safe*/robust*/align* AND lifecycle terms) on a broad preprint server plus the main online alignment forums for the same period and compare the risk-type shares. If papers on existential risk, superintelligence, and AGI confinement outnumber any single near-term category—or if the combined non-peer-reviewed share overturns the 23.2% figure for noise and outliers—the review's central empirical claim fails.

Watch

Extended reading notes

Core claim

The paper's central claim is that the empirical record of peer-reviewed AI safety research undercuts the narrative that AI safety is primarily about existential risk from advanced AI. Reviewing 2,666 database hits plus snowballed papers down to 383, the authors identify eight families of safety risk and find the largest share concerns noise and outliers (23.2%), followed by lack of monitoring, system misspecification, and lack of control enforcement, with explicitly existential or AGI-confinement work accounting for a small fraction of the literature. They further show the field's methods—applied algorithms, safe-RL agent simulations, analysis frameworks, mechanistic interpretability, and design frameworks—mirror the practices of reliability engineering, control theory, and systems safety. On this evidence they argue AI safety should be understood pluralistically, as continuous with the safety engineering tradition rather than as a separate enterprise fixated on extinction scenarios.

Load-bearing premise

The whole picture rests on the assumption that the peer-reviewed, index-database corpus queried with the words safe, robust, or align fairly represents what 'AI safety' research is; because most existential-risk and alignment work appears on preprints and non-peer-reviewed forums, the measured small share of existential risk could be an artifact of where the authors looked.

Editorial extensions

If this is right

  • If the review's distribution is representative, regulators and funders should treat AI safety as a broad engineering discipline, not only as existential-risk mitigation.
  • The shared technical vocabulary across time horizons, e.g. corrigibility and adversarial robustness in reinforcement learning, supports integrated safety research rather than a near-term/long-term dichotomy.
  • Reconnecting AI safety to systems-safety practice in aviation, medicine, and nuclear power gives policymakers a ready stock of certification and assurance frameworks to adapt.
  • A pluralistic definition of AI safety can broaden participation from researchers who do not subscribe to existential-risk narratives, without eliminating existential risk as a research topic.
  • The authors stress that their findings do not negate the importance of existential risk, only its status as the defining lens.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A direct test of the paper's scope claim would be to run the same query hierarchy on a broad preprint server plus the main online alignment forums and recalculate the risk-type shares; if existential-risk and AGI-confinement papers then outnumber every near-term category, the quantitative 'sliver' claim would be an artifact of the peer-reviewed-only sample, a limitation the paper itself flags.
  • The governance consequence the authors leave implicit: if AI safety is continuous with systems safety, then standards bodies with existing safety-case patterns for machine-learning components have a head start on AI regulation, and 'AI safety certification' becomes more tractable than under an existential-risk framing.
  • The four-cluster map suggests a research hypothesis: adversarial robustness and safe exploration may be two sides of the same formal problem (worst-case perturbations versus constraint violations), and a unification would strengthen the paper's continuity claim.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper reports a systematic literature review of peer-reviewed AI safety research indexed in Web of Science and Scopus, using a multi-stage query that combines AI/ML terms with the safety-related keywords safe*, robust*, and align*, and adds lifecycle terms. From 383 selected papers, the authors inductively code eight risk types and ten methodology categories, reporting that near-term concerns such as robustness to noise and outliers dominate while existential risk appears in only a small minority of papers. On this basis, the paper argues that AI safety is a pluralistic field naturally connected to traditional technological and systems safety, and that the public narrative focusing on existential risk is too narrow. The authors propose a more inclusive framing and outline future research directions, including expanding the review to non-peer-reviewed sources.

Significance. If the empirical findings are accepted, the paper would make a useful contribution by providing a structured, reproducible map of peer-reviewed AI safety research and by grounding debates about the field's scope in data rather than anecdote. The authors are commendably explicit about their methodology and state their intention to release the annotated corpus and analysis code. However, the force of the central claim depends on whether the corpus is representative of AI safety as a field; the current manuscript does not yet demonstrate this, because the selection criteria and the borrowed risk taxonomy partly determine the observed distribution. The pluralism thesis is plausible and important, but the evidence as presented is provisional.

major comments (3)
  1. [§2.1, §4, §6] The central relative-frequency claim in §6—that AI safety research 'challenges the narrative that associates AI safety primarily with mitigating existential risks'—is computed from the 383-paper corpus selected by q3 = q1 ∧ (safe* ∨ robust* ∨ align*) and q4. Because q2 includes the broad wildcard 'robust*', the denominator is inflated by mainstream machine-learning, control, and reliability-engineering papers on noise, outliers, and adversarial perturbations that may not self-identify as AI safety research. The authors apply exclusion criterion (B) (motivation), but they do not report how many papers were excluded at each stage, nor do they provide a baseline comparison against general ML papers to show that the risk distribution in Figure 3 is specific to AI safety rather than an artifact of the robust* term. Without a sensitivity analysis (for example, re-running the review with q2 restricted to 'safe*' or to the phrase 'AI safety', or with additional terms), the claim that 'noise and outliers' is the dominant risk type (23.2%) and the resulting argument in §6 are not fully supported.
  2. [§2.1; Introduction, Limitations paragraph] The search protocol also systematically underrepresents existential-risk and long-term alignment research. Much of this work is disseminated on arXiv, LessWrong, and the AI Alignment Forum, and it often uses vocabulary such as 'existential risk', 'x-risk', 'superintelligence', or 'misalignment' without necessarily co-occurring with safe*, robust*, or align* in WoS/Scopus-indexed metadata. The Limitations paragraph concedes the exclusion of preprints and forums, but this concession is not quantified, and the paper's conclusion in §6 is not correspondingly hedged. A supplemental search of arXiv (and a citation-based snowball from known x-risk seeds) with the resulting papers coded into the same taxonomy would allow the authors to bound the effect of this omission and would materially strengthen the central claim.
  3. [§4, Fig. 3] The risk taxonomy is borrowed from DeepMind's blog post [67], and several categories overlap with the search terms themselves. 'Noise and outliers' and 'Lack of monitoring' are natural homes for papers captured by robust* and safe*, so the relative frequencies in Figure 3 partially reflect the query design. The authors should discuss this potential circularity and, ideally, re-code a random subset of papers with a taxonomy not derived from the safety-robustness-alignment vocabulary to verify that the qualitative ordering of risk types is stable. At minimum, the text should acknowledge that the taxonomy choice and the keyword filter are not independent.
minor comments (6)
  1. [Fig. 1 caption] The caption reads 'World cloud of morphologically standardised terms'; 'World' should be 'Word'.
  2. [§6] The text contains two typos: 'episemically-inclusive' should be 'epistemically inclusive', and 'demistify' should be 'demystify'.
  3. [§2.1, q4d] The wildcard 'decomission*' is misspelled; the intended term is 'decommission*'.
  4. [§5] 'Generalisaion' should be 'generalization' in the sentence describing domain-generalisation methods.
  5. [§2.1] The query is described in WoS notation only; the authors should state how the syntax was adapted for Scopus, including any differences in wildcard handling and field tags.
  6. [Limitations] The snowball sampling description does not report how many of the 117 snowballed papers were included in the final 383; reporting this overlap would help readers assess the contribution of the snowballing step.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the corpus is operationalized rather than derived, and the field-breadth conclusion is an empirical generalization supported by independent paper-level screening; minor self-citations are not load-bearing.

full rationale

This paper is a systematic literature review, not a formal derivation, so equation-level circularity does not directly apply. The corpus is defined in Section 2.1 by q3 = q1 ∧ q2, where q2 requires safe*, robust*, or align* in WoS/Scopus metadata. One might worry that the prominence of 'noise and outliers' (23.2%) and 'adversarial attacks' categories in Figure 3 is an artifact of including robust* as an inclusion keyword. However, the final 383-paper set was not produced by q3 alone: the authors applied exclusion criteria (A)-(C) requiring a primary focus on AI, a stated motivation for safe AI algorithms, and generalizability, and then manually annotated each full text. The risk-type frequencies are therefore a measured property of the screened corpus, not a logically forced consequence of the keyword filter. The risk taxonomy is borrowed from DeepMind [67], an external source, and applying it is standard deductive coding; the authors concede annotator bias in the Limitations section. The paper also explicitly concedes that excluding arXiv, LessWrong, and the AI Alignment Forum may underrepresent existential-risk work; that is a generalizability threat, not a circularity, because the paper's central claim is explicitly about 'primarily peer-reviewed research' as indexed by WoS/Scopus, and the existence of the 383 concrete safety papers is an independent empirical fact. Self-citations ([1], [16], [78], [148]) are used as illustrative examples of recent discourse, explainability research, and mechanistic interpretability; none is load-bearing, and no uniqueness theorem or prior-work-only premise is invoked to force the conclusions. The finding that AI safety research extends traditional technological safety is an interpretive synthesis of the annotated corpus, not a renaming of the query inputs. Overall, the paper shows minor self-citation but no circular step that reduces its central empirical claim to its own construction.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The central empirical claims rest on corpus representativeness, the search-term definition of AI safety, and the externally borrowed risk taxonomy. No numerical free parameters are fitted; the main load-bearing choices are qualitative: which databases, which keywords, and which coding categories. The normative conclusion adds a separate value premise.

assumptions (5)
  • domain assumption The WoS and Scopus peer-reviewed corpus plus snowball sampling is representative of AI safety research as a whole.
    This is required for the claim that existential risk is only a minor part of AI safety. Section 2.1 defines the corpus; the Limitations concede that excluding arXiv, preprints, and forums underrepresents parts of the field.
  • ad hoc to paper The keyword filter safe*, robust*, align* captures the relevant AI safety literature.
    q2 in Section 2.1 selects on these stems. Robustness papers enter through 'robust*' and alignment papers through 'align*', so the resulting distribution of risk types is partly determined by the query.
  • domain assumption The eight risk categories adapted from DeepMind's taxonomy are a valid coding scheme.
    Section 4 says the review builds on DeepMind's categorization [67]. The percentages in Figure 3 depend on this externally chosen taxonomy rather than emerging purely from the corpus.
  • domain assumption Subjective title, abstract, and full-text screening can be applied consistently.
    Section 2.1 applies exclusion criteria (A) through (C) without reporting inter-rater reliability; the Limitations acknowledge annotator bias.
  • domain assumption A more inclusive and pluralistic conception of AI safety is normatively desirable.
    The recommendation in Section 6 is a value judgment that the empirical finding of breadth does not by itself entail, though the authors argue for it.

how reviews work

0 comments
Cite this review

Pith. "Pith review of AI Safety for Everyone." pith.science (2026). https://pith.science/paper/YOOEDBHE

@misc{pith2026250209288,
  author       = {Pith},
  title        = {Pith review of: AI Safety for Everyone},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YOOEDBHE}},
  note         = {Machine review of arXiv:2502.09288}
}
read the original abstract

Recent discussions and research in AI safety have increasingly emphasized the deep connection between AI safety and existential risk from advanced AI systems, suggesting that work on AI safety necessarily entails serious consideration of potential existential threats. However, this framing has three potential drawbacks: it may exclude researchers and practitioners who are committed to AI safety but approach the field from different angles; it could lead the public to mistakenly view AI safety as focused solely on existential scenarios rather than addressing a wide spectrum of safety challenges; and it risks creating resistance to safety measures among those who disagree with predictions of existential AI risks. Through a systematic literature review of primarily peer-reviewed research, we find a vast array of concrete safety work that addresses immediate and practical concerns with current AI systems. This includes crucial areas like adversarial robustness and interpretability, highlighting how AI safety research naturally extends existing technological and systems safety concerns and practices. Our findings suggest the need for an epistemically inclusive and pluralistic conception of AI safety that can accommodate the full range of safety considerations, motivations, and perspectives that currently shape the field.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

179 extracted references · 67 canonical work pages

  1. [67]

    A., Maini, V

    Ortega, P. A., Maini, V. & the DeepMind safety team. Building safe artificial intel- ligence: specification, robustness, and assurance. https://deepmindsafetyresearch. medium.com/building-safe-artificial-intelligence-52f5f75058f1 (2018). Accessed: 2024-05-13

    2018

  2. [1]

    Two types of AI existential risk: Decisive and accumulative

    Kasirzadeh, A. Two types of AI existential risk: Decisive and accumulative. arXiv preprint arXiv:2401.07836 (2024)

    arXiv 2024

  3. [2]

    & Nelson, A

    Lazar, S. & Nelson, A. AI safety on whose terms? Science 381, 138–138 (2023)

    2023

  4. [3]

    & Wang, M

    Ahmed, S., Ja´ zwi´ nska, K., Ahlawat, A., Winecoff, A. & Wang, M. Building the epistemic community of AI safety. First Monday (forthcoming). URL https://papers.ssrn.com/sol3/papers.cfm?abstract id=4641526

  5. [4]

    Existential risks: Analyzing human extinction scenarios and related hazards

    Bostrom, N. Existential risks: Analyzing human extinction scenarios and related hazards. Journal of Evolution and Technology 9, 1–30 (2002)

    2002

  6. [5]

    Superintelligence: Paths, dangers, strategies (Oxford University Press, 2014)

    Bostrom, N. Superintelligence: Paths, dangers, strategies (Oxford University Press, 2014)

    2014

  7. [6]

    The precipice: Existential risk and the future of humanity (Hachette Books, 2020)

    Ord, T. The precipice: Existential risk and the future of humanity (Hachette Books, 2020)

    2020

  8. [7]

    The logic of effective altruism

    Boston Review. The logic of effective altruism. Boston Review (2015). URL https://www.bostonreview.net/forum/peter-singer-logic-effective-altruism/

    2015

Show all 179 references
  1. [8]

    Effective altruism and its critics

    Gabriel, I. Effective altruism and its critics. Journal of Applied Philosophy 34, 457–473 (2017)

    2017

  2. [9]

    AI is an existential threat—just not the way you think

    Eisikovits, N. AI is an existential threat—just not the way you think. https://www.scientificamerican.com/article/ ai-is-an-existential-threat-just-not-the-way-you-think/ (2023). Accessed: July 12, 2023

    2023

  3. [10]

    Statement on ai risk

    Center for AI Safety. Statement on ai risk. https://www.safe.ai/work/ statement-on-ai-risk (2023). Accessed on 2024-05-08

    2023

  4. [11]

    Training, A. S. Ai safety training. https://aisafety.training/ (2024). Accessed on 2024-05-08

    2024

  5. [12]

    AI safety — wikipedia, the free encyclopedia (2024)

    Wikipedia contributors. AI safety — wikipedia, the free encyclopedia (2024). URL https://en.wikipedia.org/wiki/AI safety. [Online; accessed 21-January-2024]

    2024

  6. [13]

    & Norvig, P

    Ag¨ uera y Arcas, B. & Norvig, P. Artificial general intelligence is already here. Noema Magazine (2023). Available at: https://www.noemamag.com/ artificial-general-intelligence-is-already-here/. 17

    2023

  7. [14]

    Roose, K. A.I. Poses ‘Risk of Extinction’, Industry Leaders Warn. The New York Times (2023). URL https://www.nytimes.com/2023/05/30/technology/ ai-threat-warning.html

    2023

  8. [15]

    PM should make ethics a priority at AI safety summit, say tech professionals (2023)

    BCS Comment. PM should make ethics a priority at AI safety summit, say tech professionals (2023). URL https://www.bcs.org/articles-opinion-and-research/ pm-should-make-ethics-a-priority-at-ai-safety-summit-say-tech-professionals/. Accessed: 27 January, 2024

    2023

  9. [16]

    & Gohdes, A

    Gilardi, F., Kasirzadeh, A., Bernstein, A., Staab, S. & Gohdes, A. We need to understand the effect of narratives about generative ai. Nature Human Behaviour 1–2 (2024)

    2024

  10. [17]

    Bender, E. M. Talking about a ‘schism’ is ahistorical (2023). URL https://medium. com/@emilymenonbender/talking-about-a-schism-is-ahistorical-3c454a77220f

    2023

  11. [18]

    Krause, S. S. Aircraft safety (McGraw-Hill Professional Publishing, 2003)

    2003

  12. [19]

    Boyd, D. D. A review of general aviation safety (1984–2017). Aerospace medicine and human performance 88, 657–664 (2017)

    2017

  13. [20]

    & Restani, P

    Pifferi, G. & Restani, P. The safety of pharmaceutical excipients. Il Farmaco 58, 541–550 (2003)

    2003

  14. [21]

    Leveson, N. et al. Applying system engineering to pharmaceutical safety. Journal of Healthcare Engineering 3, 391–414 (2012)

    2012

  15. [22]

    & Van Ouytsel, J

    De Kimpe, L., Walrave, M., Ponnet, K. & Van Ouytsel, J. Internet safety. The international encyclopedia of media literacy 1–11 (2019)

    2019

  16. [23]

    Salim, H. M. Cyber safety: A systems thinking and systems theory approach to managing cyber security risks. Ph.D. thesis, Massachusetts Institute of Technology (2014)

    2014

  17. [24]

    Leveson, N. G. Engineering a safer world: Systems thinking applied to safety (The MIT Press, 2016)

    2016

  18. [25]

    Varshney, K. R. Engineering safety in machine learning , 1–5 (IEEE, 2016)

    2016

  19. [26]

    Rismani, S. et al. From plane crashes to algorithmic harm: applicability of safety engineering frameworks for responsible ml , 1–18 (2023)

    2023

  20. [27]

    System safety and artificial intelligence , 1584–1584 (2022)

    Dobbe, R. System safety and artificial intelligence , 1584–1584 (2022)

    2022

  21. [28]

    Rismani, S. et al. Beyond the ml model: Applying safety engineering frameworks to text-to-image development , 70–83 (2023). 18

    2023

  22. [29]

    Amodei, D. et al. Concrete problems in AI safety.arXiv preprint arXiv:1606.06565 (2016)

    2016 arXiv

  23. [30]

    Raji, I. D. & Dobbe, R. Concrete problems in AI safety, revisited. arXiv preprint arXiv:2401.10899 (2023)

    2023 arXiv

  24. [31]

    & Charters, S

    Kitchenham, B. & Charters, S. Guidelines for performing Systematic Literature Reviews in Software Engineering. EBSE Technical Report EBSE-2007-01, School of Computer Science and Mathematics, Keele University, Keele, UK (2007)

    2007

  25. [32]

    Wohlin, C. Guidelines for snowballing in systematic literature studies and a replication in software engineering , EASE ’14, 1–10 (Association for Computing Machinery, New York, NY, USA, 2014)

    2014

  26. [33]

    & Amodei, D

    Irving, G., Christiano, P. & Amodei, D. AI safety via debate. arXiv (2018)

    2018

  27. [34]

    Ng, A. Y. & Russell, S. J. Algorithms for Inverse Reinforcement Learning, ICML ’00, 663–670 (Morgan Kaufmann Publishers Inc., San Francisco, CA, USA, 2000)

    2000

  28. [35]

    Hendrycks, D. et al. Aligning AI With Shared Human Values (ICLR, 2021). 2008.02275

    2021 arXiv

  29. [36]

    Yampolskiy, R. V. Artificial Intelligence Safety and Cybersecurity: A Timeline of AI Failures. arXiv (2016)

    2016

  30. [37]

    J., Abbeel, P

    Hadfield-Menell, D., Russell, S. J., Abbeel, P. & Dragan, A. Cooperative Inverse Reinforcement Learning, Vol. 29 (Curran Associates, Inc., 2016)

    2016

  31. [38]

    Xu, H., Zhu, T., Zhang, L., Zhou, W. & Yu, P. S. Machine Unlearning: A Survey. ACM Computing Surveys 56, 9:1–9:36 (2023)

    2023

  32. [39]

    & Tegmark, M

    Russell, S., Dewey, D. & Tegmark, M. Research Priorities for Robust and Beneficial Artificial Intelligence. AI magazine 36, 105–114 (2015)

    2015

  33. [40]

    & Abrecht, S

    Willers, O., Sudholt, S., Raafatnia, S. & Abrecht, S. Casimiro, A., Ortmeier, F., Schoitsch, E., Bitsch, F. & Ferreira, P. (eds) Safety Concerns and Mitigation Approaches Regarding the Use of Deep Learning in Safety-Critical Perception Tasks. (eds Casimiro, A., Ortmeier, F., S...

    2020

  34. [41]

    Mohseni, S. et al. Taxonomy of Machine Learning Safety: A Survey and Primer. ACM Computing Surveys 55, 1–38 (2022)

    2022

  35. [42]

    & Steinhardt, J

    Hendrycks, D., Carlini, N., Schulman, J. & Steinhardt, J. Unsolved Problems in ML Safety. arXiv (2022). 19

    2022

  36. [43]

    Boyatzis, R. E. Transforming qualitative information: Thematic analysis and code development (sage, 1998)

    1998

  37. [44]

    van Eck, N. J. & Waltman, L. Software survey: VOSviewer, a computer program for bibliometric mapping. Scientometrics 84, 523–538 (2010)

    2010

  38. [45]

    V., Strong, J

    Oster Jr, C. V., Strong, J. S. & Zorn, C. K. Analyzing aviation safety: Problems, challenges, opportunities. Research in transportation economics 43, 148–164 (2013)

    2013

  39. [46]

    S., Corrigan, J

    Donaldson, M. S., Corrigan, J. M. & Kohn, L. T. To err is human: building a safer health system (2000)

    2000

  40. [47]

    Bates, D. W. et al. The safety of inpatient health care. New England Journal of Medicine 388, 142–153 (2023)

    2023

  41. [48]

    Marais, K., Dulac, N., Leveson, N. et al. Beyond normal accidents and high reliability organizations: The need for an alternative approach to safety in complex systems, 1–16 (Citeseer, 2004)

    2004

  42. [49]

    Griffor, E. Handbook of system safety and security: cyber risk and risk manage- ment, cyber security, threat analysis, functional safety, software systems, and cyber physical systems (Syngress, 2016)

    2016

  43. [50]

    & Rohokale, V

    Prasad, R. & Rohokale, V. Cyber security: the lifeline of information and communication technology (Springer, 2020)

    2020

  44. [51]

    & Armstrong, S

    Soares, N., Fallenstein, B., Yudkowsky, E. & Armstrong, S. Corrigibility (Austin, Texas, USA, 2015)

    2015

  45. [52]

    & Orseau, L

    Ring, M. & Orseau, L. Schmidhuber, J., Th´ orisson, K. R. & Looks, M. (eds) Delusion, Survival, and Intelligent Agents . (eds Schmidhuber, J., Th´ orisson, K. R. & Looks, M.) Artificial General Intelligence , Lecture Notes in Computer Science, 11–20 (Springer, Berlin, Heidelbe...

    2011

  46. [53]

    Ganin, Y. et al. in Domain-Adversarial Training of Neural Networks (ed.Csurka, G.) Domain Adaptation in Computer Vision Applications 189–209 (Springer International Publishing, Cham, 2017)

    2017

  47. [54]

    & Chellappa, R

    Balaji, Y., Sankaranarayanan, S. & Chellappa, R. MetaReg: Towards Domain Generalization using Meta-Regularization, Vol. 31 (Curran Associates, Inc., 2018)

    2018

  48. [55]

    J., Shlens, J

    Goodfellow, I. J., Shlens, J. & Szegedy, C. Explaining and Harnessing Adversarial Examples (2015). 1412.6572

    2015 arXiv

  49. [56]

    & Vladu, A

    Madry, A., Makelov, A., Schmidt, L., Tsipras, D. & Vladu, A. Towards Deep Learning Models Resistant to Adversarial Attacks. arXiv (2019). 20

    2019

  50. [57]

    & Amodei, D

    Ray, A., Achiam, J. & Amodei, D. Benchmarking Safe Exploration in Deep Reinforcement Learning. OpenAI (2023). Unpublished article

    2023

  51. [58]

    & Krause, A

    Turchetta, M., Berkenkamp, F. & Krause, A. Wallach, H. et al. (eds) Safe Exploration for Interactive Machine Learning . (eds Wallach, H. et al. ) Advances in Neural Information Processing Systems 32 (NIPS 2019) , Vol. 32 (2019)

    2019

  52. [59]

    Ouyang, L. et al. Training language models to follow instructions with human feedback. arXiv (2022). OpenAI

    2022

  53. [60]

    Abbeel, P. & Ng, A. Y. Apprenticeship learning via inverse reinforcement learning, ICML ’04, 1 (Association for Computing Machinery, New York, NY, USA, 2004)

    2004

  54. [61]

    A., Chaudhuri, K

    Izzo, Z., Smart, M. A., Chaudhuri, K. & Zou, J. Approximate Data Deletion from Machine Learning Models , 2008–2016 (PMLR, 2021)

    2008

  55. [62]

    & Davidson, S

    Wu, Y., Dobriban, E. & Davidson, S. DeltaGrad: Rapid retraining of machine learning models, 10355–10366 (PMLR, 2020)

    2020

  56. [63]

    Adebayo, J. et al. Sanity Checks for Saliency Maps , Vol. 31 (Curran Associates, Inc., 2018)

    2018

  57. [64]

    Kim, B. et al. Interpretability beyond feature attribution: Quantitative testing with concept activation vectors (TCA V), 2668–2677 (PMLR, 2018)

    2018

  58. [65]

    & Gimpel, K

    Hendrycks, D. & Gimpel, K. A Baseline for Detecting Misclassified and Out-of- Distribution Examples in Neural Networks (arXiv, 2018). 1610.02136

    2018 arXiv

  59. [66]

    & Dietterich, T

    Hendrycks, D. & Dietterich, T. Benchmarking Neural Network Robustness to Common Corruptions and Perturbations (arXiv, 2019). 1903.12261

    2019 arXiv

  60. [68]

    Meng, Y. et al. Distantly-Supervised Named Entity Recognition with Noise-Robust Learning and Language Model Augmented Self-Training , 10367–10378 (2021)

    2021

  61. [69]

    & Guo, P

    Wang, K. & Guo, P. A Robust Automated Machine Learning System with Pseudoinverse Learning. Cognitive Computation 13, 724–735 (2021)

    2021

  62. [70]

    & Murphy, T

    Cappozzo, A., Greselin, F. & Murphy, T. B. A robust approach to model-based classification based on trimming and constraints: Semi-supervised learning in presence of outliers and label noise. Advances in Data Analysis and Classification 14, 327–354 (2020). 21

    2020

  63. [71]

    & Wang, Y

    Li, W. & Wang, Y. A robust supervised subspace learning approach for output- relevant prediction and detection against outliers. Journal of Process Control 106, 184–194 (2021)

    2021

  64. [72]

    & Krause, A

    Curi, S., Bogunovic, I. & Krause, A. Combining Pessimism with Optimism for Robust and Efficient Model-Based Deep Reinforcement Learning , Vol. 139, 2254–2264 (2021)

    2021

  65. [73]

    & Mintz, Y

    Dobbe, R., Gilbert, TK. & Mintz, Y. Hard choices in artificial intelligence. Artificial Intelligence 300 (2021)

    2021

  66. [74]

    & Feldman, V

    Dwork, C. & Feldman, V. Privacy-preserving Prediction, 1693–1702 (PMLR, 2018)

    2018

  67. [75]

    Elhage, N. et al. A mathematical framework for transformer circuits. Transformer Circuits Thread (2021). Unpublished article

    2021

  68. [76]

    & Mnih, A

    Kim, H. & Mnih, A. Disentangling by Factorising. arXiv (2019)

    2019

  69. [77]

    & Habli, I

    Ward, F. & Habli, I. An Assurance Case Pattern for the Interpretability of Machine Learning in Safety-Critical Systems , Vol. 12235 LNCS, 395–407 (2020)

    2020

  70. [78]

    & Schafer, B

    Gyevnar, B., Ferguson, N. & Schafer, B. Bridging the transparency gap: What can explainable ai learn from the ai act? , 964–971 (IOS Press, 2023)

    2023

  71. [79]

    & Kniesel-W¨ unsche, G.Safe-DS: A Domain Specific Language to Make Data Science Safe , 72–77 (2023)

    Reimann, L. & Kniesel-W¨ unsche, G.Safe-DS: A Domain Specific Language to Make Data Science Safe , 72–77 (2023)

    2023

  72. [80]

    & Lee, S.-W

    Dey, S. & Lee, S.-W. A Multi-layered Collaborative Framework for Evidence- driven Data Requirements Engineering for Machine Learning-based Safety-critical Systems, 1404–1413 (2023)

    2023

  73. [81]

    & Zimmert, J

    Wei, C.-Y., Dann, C. & Zimmert, J. A Model Selection Approach for Corruption Robust Reinforcement Learning, Vol. 167, 1043–1096 (2022)

    2022

  74. [82]

    & Singla, A

    Ghosh, A., Tschiatschek, S., Mahdavi, H. & Singla, A. Towards Deployment of Robust Cooperative AI Agents: An Algorithmic Framework for Learning Adaptive Policies. New Zealand (2020)

    2020

  75. [83]

    & Hutter, M

    Everitt, T. & Hutter, M. Avoiding Wireheading with Value Reinforcement Learning (2016). 1605.03143

    2016 arXiv

  76. [84]

    & Garrabrant, S

    Hubinger, E., van Merwijk, C., Mikulik, V., Skalse, J. & Garrabrant, S. Risks from learned optimization in advanced machine learning systems. CoRR abs/1906.01820 (2019). URL http://arxiv.org/abs/1906.01820

    2019 arXiv

  77. [85]

    & Yampolskiy, R

    Pistono, F. & Yampolskiy, R. V. The Age of Artificial Intelligence , Ch. Unethical Research: How to Create a Malevolent Artificial Intelligence (Vernon Press 22 Wilmington, DE, USA, 2016)

    2016

  78. [86]

    & Habli, I

    Picardi, C., Paterson, C., Hawkins, R., Calinescu, R. & Habli, I. Assurance argument patterns and processes for machine learning in safety-related systems , Vol. 2560, 23–30 (2020)

    2020

  79. [87]

    & Zeilinger, MN

    Wabersich, KJ., Hewing, L., Carron, A. & Zeilinger, MN. Probabilistic Model Predictive Safety Certification for Learning-Based Control. IEEE Transactions on Automatic Control 67, 176–188 (2022)

    2022

  80. [88]

    & Topcu, U

    Wen, M. & Topcu, U. Constrained Cross-Entropy Method for Safe Reinforcement Learning. IEEE Transactions on Automatic Control 66, 3123–3137 (2021)

    2021

  81. [89]

    Analyzing Information Leakage of Updates to Natural Language Models, 363–375 (2020)

    Zanella-B´ eguelin, S.et al. Analyzing Information Leakage of Updates to Natural Language Models, 363–375 (2020). 1912.07942

    2020 arXiv

  82. [90]

    & Dong, D

    Wang, Z., Chen, C. & Dong, D. A Dirichlet Process Mixture of Robust Task Models for Scalable Lifelong Reinforcement Learning. IEEE Transactions on Cybernetics 1–12 (2022)

    2022

  83. [91]

    Zou, A. et al. Universal and Transferable Adversarial Attacks on Aligned Language Models. arXiv (2023)

    2023

  84. [92]

    Ilyas, A. et al. Adversarial Examples Are Not Bugs, They Are Features , Vol. 32 (Curran Associates, Inc., 2019)

    2019

  85. [93]

    & Yin, YL

    He, RD., Han, ZY., Yang, Y. & Yin, YL. Not All Parameters Should Be Treated Equally: Deep Safe Semi-supervised Learning under Class Distribution Mismatch , 6874–6883 (2022)

    2022

  86. [94]

    & Vigna, G

    Aghakhani, H., Meng, D., Wang, Y.-X., Kruegel, C. & Vigna, G. Bullseye polytope: A scalable clean-label poisoning attack with improved transferability , 159–178 (2021)

    2021

  87. [95]

    Liu, Y. et al. Backdoor Defense with Machine Unlearning , 280–289 (IEEE Press, London, United Kingdom, 2022)

    2022

  88. [96]

    & Hein, M

    Meinke, A. & Hein, M. Towards neural networks that provably know when they don’t know (2020). 1909.12180

    2020 arXiv

  89. [97]

    Abdelfattah, S., Kasmarik, K. & Hu, J. A robust policy bootstrapping algo- rithm for multi-objective reinforcement learning in non-stationary environments. Adaptive Behavior 28, 273–292 (2020)

    2020

  90. [98]

    & Topcu, U

    Djeumou, F., Cubuktepe, M., Lennon, C. & Topcu, U. Task-Guided Inverse Reinforcement Learning Under Partial Information (2021). 2105.14073. 23

    2021 arXiv

  91. [99]

    Yampolskiy, R. V. Leakproofing the Singularity: Artificial intelligence confinement problem. Journal of Consciousness Studies 19, 194–214 (2012)

    2012

  92. [100]

    Yampolskiy, R. V. in Artificial Intelligence Safety Engineering: Why Machine Ethics Is a Wrong Approach (ed.M¨ uller, V. C.)Philosophy and Theory of Artificial Intelligence Studies in Applied Philosophy, Epistemology and Rational Ethics, 389–396 (Springer, Berlin, Heidelberg, 2013)

    2013

  93. [101]

    Yampolskiy, R. V. Taxonomy of Pathways to Dangerous AI (2015). 1511.03246

    2015 arXiv

  94. [102]

    Wheatley, S., Sovacool, B. K. & Sornette, D. Reassessing the safety of nuclear power. Energy Research & Social Science 15, 96–100 (2016)

    2016

  95. [103]

    & Ho, WK

    Tay, EB., Gan, OP. & Ho, WK. Rauch, HE. (ed.) A study on real-time artificial intelligence. (ed.Rauch, HE.) Artificial Intelligence in Real-Time Control 1997 , 109–114 (1998)

    1998

  96. [104]

    Bach, J., Goertzel, B

    Hibbard, B. Bach, J., Goertzel, B. & Ikl´ e, M. (eds) Avoiding Unintended AI Behaviors. (eds Bach, J., Goertzel, B. & Ikl´ e, M.) Artificial General Intelligence , Lecture Notes in Computer Science, 107–116 (Springer, Berlin, Heidelberg, 2012)

    2012

  97. [105]

    Bieger, J., Goertzel, B

    Sezener, CE. Bieger, J., Goertzel, B. & Potapov, A. (eds) Inferring Human Values for Safe AGI Design. (eds Bieger, J., Goertzel, B. & Potapov, A.) Artificial General Intelligence (AGI 2015) , Vol. 9205, 152–155 (2015)

    2015

  98. [106]

    & Guerraoui, R

    El Mhamdi, E. & Guerraoui, R. When Neurons Fail, 1028–1037 (2017)

    2017

  99. [107]

    & Grote, T

    Freiesleben, T. & Grote, T. Beyond generalization: A theory of robustness in machine learning. Synthese 202 (2023)

    2023

  100. [108]

    & Shah, J

    Sanneman, L. & Shah, J. Transparent Value Alignment , 557–560 (ACM, Stockholm Sweden, 2023)

    2023

  101. [109]

    & Mitchell, T

    Murphy, B., Talukdar, P. & Mitchell, T. Kay, M. & Boitet, C. (eds) Learning Effective and Interpretable Semantic Models using Non-Negative Sparse Embed- ding. (eds Kay, M. & Boitet, C.) Proceedings of COLING 2012, 1933–1950 (The COLING 2012 Organizing Committee, Mumbai, India, 2012)

    2012

  102. [110]

    & Hovy, E

    Subramanian, A., Pruthi, D., Jhamtani, H., Berg-Kirkpatrick, T. & Hovy, E. SPINE: SParse Interpretable Neural Embeddings (2017). 1711.08792

    2017 arXiv

  103. [111]

    & Negahban, S

    Shaham, U., Yamada, Y. & Negahban, S. Understanding adversarial training: Increasing local stability of supervised models through robust optimization. Neurocomputing 307, 195–204 (2018)

    2018

  104. [112]

    & Srinivasa, C

    Wu, G., Hashemi, M. & Srinivasa, C. PUMA: Performance Unchanged Model Augmentation for Training Data Removal (2022). 2203.00846. 24

    2022 arXiv

  105. [113]

    & Yang, L

    Jing, S. & Yang, L. A robust extreme learning machine framework for uncertain data classification. Journal of Supercomputing 76, 2390–2416 (2020)

    2020

  106. [114]

    & Luo, ZZ

    Gan, HT., Li, ZH., Fan, YL. & Luo, ZZ. Dual Learning-Based Safe Semi- Supervised Learning. IEEE Access 6, 2615–2621 (2018)

    2018

  107. [115]

    Engstrom, L. et al. Adversarial Robustness as a Prior for Learned Representations. arXiv (2019)

    2019

  108. [116]

    & Lowd, D

    Brophy, J. & Lowd, D. Machine Unlearning for Random Forests , 1092–1104 (PMLR, 2021)

    2021

  109. [117]

    S., Tarun, A

    Chundawat, V. S., Tarun, A. K., Mandal, M. & Kankanhalli, M. Zero-Shot Machine Unlearning. IEEE Transactions on Information Forensics and Security 18, 2345–2354 (2023)

    2023

  110. [118]

    & Jha, S

    Chen, J., Li, Y., Wu, X., Liang, Y. & Jha, S. Oliver, N., P´ erez-Cruz, F., Kramer, S., Read, J. & Lozano, J. A. (eds) ATOM: Robustifying Out-of-Distribution Detection Using Outlier Mining . (eds Oliver, N., P´ erez-Cruz, F., Kramer, S., Read, J. & Lozano, J. A.) Machine Learn...

    2021

  111. [119]

    & Horvitz, E

    Lakkaraju, H., Kamar, E., Caruana, R. & Horvitz, E. Identifying unknown unknowns in the open world: Representations and policies for guided exploration , Vol. 31 (2017)

    2017

  112. [120]

    & Huang, QM

    Zhuo, JB., Wang, SH., Zhang, WG. & Huang, QM. Deep Unsupervised Convolutional Domain Adaptation , 261–269 (2017)

    2017

  113. [121]

    & Bishop, N

    Bossens, DM. & Bishop, N. Explicit Explore, Exploit, or Escape (E4): Near- optimal safety-constrained reinforcement learning in polynomial time. Machine Learning 112, 817–858 (2023)

    2023

  114. [122]

    & Trimpe, S

    Massiani, PF., Heim, S., Solowjow, F. & Trimpe, S. Safe Value Functions. IEEE Transactions on Automatic Control 68, 2743–2757 (2023)

    2023

  115. [123]

    & Shroff, N

    Shi, M., Liang, Y. & Shroff, N. A Near-Optimal Algorithm for Safe Reinforcement Learning Under Instantaneous Hard Constraints , Vol. 202, 31243–31268 (2023)

    2023

  116. [124]

    Hunt, N. et al. Verifiably Safe Exploration for End-to-End Reinforcement Learning (2021)

    2021

  117. [125]

    J., Shen, A., Bastani, O

    Ma, Y. J., Shen, A., Bastani, O. & Dinesh, J. Conservative and Adaptive Penalty for Model-Based Safe Reinforcement Learning, Vol. 36, 5404–5412 (2022)

    2022

  118. [126]

    Zwane, S. et al. Safe Trajectory Sampling in Model-Based Reinforcement Learning, Vol. 2023-August (2023). 25

    2023

  119. [127]

    & Lauer, M

    Fischer, J., Eyberg, C., Werling, M. & Lauer, M. Sampling-based Inverse Reinforcement Learning Algorithms with Safety Constraints , 791–798 (2021)

    2021

  120. [128]

    & Zhou, M

    Zhou, Z., Liu, G. & Zhou, M. A Robust Mean-Field Actor-Critic Reinforcement Learning Against Adversarial Perturbations on Agent States. IEEE Transactions on Neural Networks and Learning Systems 1–12 (2023)

    2023

  121. [129]

    Aligning individual and collective welfare in complex socio- technical systems by combining metaheuristics and reinforcement learning

    Bazzan, ALC. Aligning individual and collective welfare in complex socio- technical systems by combining metaheuristics and reinforcement learning. Engineering Applications of Artificial Intelligence 79, 23–33 (2019)

    2019

  122. [130]

    J., Haupt, A

    Christoffersen, P. J., Haupt, A. A. & Hadfield-Menell, D. Get It in Writing: Formal Contracts Mitigate Social Dilemmas in Multi-Agent RL, AAMAS ’23, 448– 456 (International Foundation for Autonomous Agents and Multiagent Systems, Richland, SC, 2023)

    2023

  123. [131]

    Christiano, P. F. et al. Deep Reinforcement Learning from Human Preferences , Vol. 30 (Curran Associates, Inc., 2017)

    2017

  124. [132]

    & Lipton, Z

    Kaushik, D., Hovy, E. & Lipton, Z. C. Learning the Difference that Makes a Difference with Counterfactually-Augmented Data (arXiv, 2020). 1909.12434

    2020 arXiv

  125. [133]

    & Sreenath, K

    Li, ZY., Zeng, J., Thirugnanam, A. & Sreenath, K. Bridging Model-based Safety and Model-free Reinforcement Learning through System Identification of Low Dimensional Linear Models. Robotics: Science and System 18 (2022)

    2022

  126. [134]

    & IEEE A Contact-Safe Reinforcement Learning Framework for Contact-Rich Robot Manipulation , 2476–2482 (2022)

    Zhu, X., Kang, SC., Chen, JY. & IEEE A Contact-Safe Reinforcement Learning Framework for Contact-Rich Robot Manipulation , 2476–2482 (2022)

    2022

  127. [135]

    & Inam, R

    Terra, A., Riaz, H., Raizer, K., Hata, A. & Inam, R. Safety vs. Efficiency: AI-Based Risk Mitigation in Collaborative Robotics , 151–160 (2020)

    2020

  128. [136]

    & Wagner, D

    Carlini, N. & Wagner, D. Towards Evaluating the Robustness of Neural Networks , 39–57 (IEEE Computer Society, 2017)

    2017

  129. [137]

    & Clune, J

    Nguyen, A., Yosinski, J. & Clune, J. Deep neural networks are easily fooled: High confidence predictions for unrecognizable images, 427–436 (IEEE, Boston, MA, USA, 2015)

    2015

  130. [138]

    Kaufmann, M. et al. Testing Robustness Against Unforeseen Adversaries. arXiv (2019)

    2019

  131. [139]

    Hendrycks, D. et al. What Would Jiminy Cricket Do? Towards Agents That Behave Morally (2021)

    2021

  132. [140]

    Gardner, M. et al. Cohn, T., He, Y. & Liu, Y. (eds) Evaluating Models’ Local Decision Boundaries via Contrast Sets . (eds Cohn, T., He, Y. & Liu, Y.) Findings of the Association for Computational Linguistics: EMNLP 2020 , 1307–1323 26 (Association for Computational Linguistics...

    2020

  133. [141]

    Spears, D. F. in Assuring the Behavior of Adaptive Agents (eds Rouff, C. A., Hinchey, M., Rash, J., Truszkowski, W. & Gordon-Spears, D.) Agent Technol- ogy from a Formal Perspective NASA Monographs in Systems and Software Engineering, 227–257 (Springer, London, 2006)

    2006

  134. [142]

    & Putzer, H.A Safety Case Pattern for Systems with Machine Learning Components , Vol

    Wozniak, E., Cˆ arlan, C., Acar-Celik, E. & Putzer, H.A Safety Case Pattern for Systems with Machine Learning Components , Vol. 12235 LNCS, 370–382 (2020)

    2020

  135. [143]

    & Verma, R

    Gholampour, P. & Verma, R. Adversarial Robustness of Phishing Email Detection Models, 67–76 (2023)

    2023

  136. [144]

    & Steinhardt, J

    Nanda, N., Chan, L., Lieberum, T., Smith, J. & Steinhardt, J. Progress measures for grokking via mechanistic interpretability. arXiv (2023)

    2023

  137. [145]

    Olsson, C. et al. In-context learning and induction heads. Transformer Circuits Thread (2022). OpenAI

    2022

  138. [146]

    & Fei-Fei, L

    Karpathy, A., Johnson, J. & Fei-Fei, L. Visualizing and Understanding Recurrent Networks (arXiv, 2016). 1506.02078

    2016 arXiv

  139. [147]

    S., Barrett, D

    Morcos, A. S., Barrett, D. G. T., Rabinowitz, N. C. & Botvinick, M. On the importance of single directions for generalization. arXiv (2018)

    2018

  140. [148]

    G., Cohen, S

    Gyevnar, B., Wang, C., Lucas, C. G., Cohen, S. B. & Albrecht, S. V. Causal Explanations for Sequential Decision-Making in Multi-Agent Systems , AAMAS ’24, 771–779 (International Foundation for Autonomous Agents and Multiagent Systems, Richland, SC, 2024)

    2024

  141. [149]

    & Iwane, H

    Okawa, Y., Sasaki, T. & Iwane, H. Automatic Exploration Process Adjustment for Safe Reinforcement Learning with Joint Chance Constraint Satisfaction , Vol. 53, 1588–1595 (2020)

    2020

  142. [150]

    & She, QS

    Gan, HT., Luo, ZZ., Meng, M., Ma, YL. & She, QS. A risk degree-based safe semi-supervised learning algorithm. International Journal of Machine Learning and Cybernetics 7, 85–94 (2016)

    2016

  143. [151]

    Zhang, YX. et al. Barrier Lyapunov Function-Based Safe Reinforcement Learning for Autonomous Vehicles With Optimized Backstepping. IEEE Transactions on Neural Networks and Learning Systems (2022)

    2022

  144. [152]

    & Amini, M.-R

    Nicolae, M.-I., Sebban, M., Habrard, A., Gaussier, E. & Amini, M.-R. Algorithmic robustness for semi-supervised ( ϵ, γ, τ )-good metric learning, Vol. 9489, 253–263 (2015)

    2015

  145. [153]

    & Seto, E

    Khan, WU. & Seto, E. A ”Do No Harm” Novel Safety Checklist and Research Approach to Determine Whether to Launch an Artificial Intelligence-Based 27 Medical Technology: Introducing the Biological-Psychological, Economic, and Social (BPES) Framework. Journal of Medical Internet ...

    2023

  146. [154]

    & Werner, B

    Schumeg, B., Marotta, F. & Werner, B. Proposed V-Model for Verification, Validation, and Safety Activities for Artificial Intelligence , 61–66 (2023)

    2023

  147. [155]

    & Weyrich, M

    Kamm, S., Sahlab, N., Jazdi, N. & Weyrich, M. A Concept for Dynamic and Robust Machine Learning with Context Modeling for Heterogeneous Manufactur- ing Data , Vol. 118, 354–359 (2023)

    2023

  148. [156]

    & Nogueira, I

    Costa, E., Rebello, C., Fontana, M., Schnitman, L. & Nogueira, I. A Robust Learning Methodology for Uncertainty-Aware Scientific Machine Learning Models. Mathematics 11 (2023)

    2023

  149. [157]

    & Kyrki, V

    Aksjonov, A. & Kyrki, V. A Safety-Critical Decision-Making and Control Framework Combining Machine-Learning-Based and Rule-Based Algorithms. SAE International Journal of Vehicle Dynamics Stability And NVH 7, 287–299 (2023)

    2023

  150. [158]

    Antikainen, J. et al. Yue, T. & Mirakhorli, M. (eds) A Deployment Model to Extend Ethically Aligned AI Implementation Method ECCOLA . (eds Yue, T. & Mirakhorli, M.) 29th IEEE International Requirements Engineering Conference Workshops (REW 2021) , 230–235 (2021)

    2021

  151. [159]

    & Abrahamsson, P

    Vakkuri, V., Kemell, KK., Jantunen, M., Halme, E. & Abrahamsson, P. ECCOLA - A method for implementing ethically aligned AI systems. Journal of Systems and Software 182 (2021)

    2021

  152. [160]

    & Asudeh, A

    Zhang, H., Shahbazi, N., Chu, X. & Asudeh, A. FairRover: Explorative model building for fair and responsible machine learning (2021)

    2021

  153. [161]

    Coston, A. et al. A Validity Perspective on Evaluating the Justified Use of Data-driven Decision-making Algorithms , 690–704 (2023)

    2023

  154. [162]

    & Yung, M

    Gittens, A., Yener, B. & Yung, M. An Adversarial Perspective on Accuracy, Robustness, Fairness, and Privacy: Multilateral-Tradeoffs in Trustworthy ML. IEEE Access 10, 120850–120865 (2022)

    2022

  155. [163]

    & Critch, A

    Taylor, J., Yudkowsky, E., LaVictoire, P. & Critch, A. Alignment for advanced machine learning systems. Ethics of Artificial Intelligence 342–382 (2016)

    2016

  156. [164]

    & Yampolskiy, R

    Sotala, K. & Yampolskiy, R. V. Responses to catastrophic AGI risk: A survey. Physica Scripta 90, 018001 (2014)

    2014

  157. [165]

    Metacognition for artificial intelligence system safety-An approach to safe and desired behavior

    Johnson, B. Metacognition for artificial intelligence system safety-An approach to safe and desired behavior. Safety Science 151 (2022)

    2022

  158. [166]

    Hatherall, L. et al. Responsible Agency Through Answerability (2022). 28

    2022

  159. [167]

    Embedding responsibility in intelligent systems: From AI ethics to responsible AI ecosystems

    Stahl, BC. Embedding responsibility in intelligent systems: From AI ethics to responsible AI ecosystems. Scientific Reports 13 (2023)

    2023

  160. [168]

    Counterfactual learning in enhancing resilience in autonomous agent systems

    Samarasinghe, D. Counterfactual learning in enhancing resilience in autonomous agent systems. Frontiers in Artificial Intelligence 6 (2023)

    2023

  161. [169]

    & Joyce, J

    Diemert, S., Millet, L., Groves, J. & Joyce, J. Safety Integrity Levels for Artificial Intelligence, Vol. 14182 LNCS, 397–409 (2023)

    2023

  162. [170]

    & Jia, R

    Wang, J. & Jia, R. Data Banzhaf: A Robust Data Valuation Framework for Machine Learning, Vol. 206, 6388–6421 (2023)

    2023

  163. [171]

    & Hutter, M

    Everitt, T., Filan, D., Daswani, M. & Hutter, M. Self-Modification of Policy and Utility Function in Rational Agents. arXiv (2016)

    2016

  164. [172]

    & Artus, G

    Badea, C. & Artus, G. Bramer, M. & Stahl, F. (eds) Morality, Machines, and the Interpretation Problem: A Value-based, Wittgensteinian Approach to Building Moral Agents. (eds Bramer, M. & Stahl, F.) Artificial Intelligence, AI 2022 , Vol. 39, 124–137 (2022)

    2022

  165. [173]

    Beneficial Artificial Intelligence Coordination by Means of a Value Sensitive Design Approach

    Umbrello, S. Beneficial Artificial Intelligence Coordination by Means of a Value Sensitive Design Approach. Big Data and Cognitive Computing 3 (2019)

    2019

  166. [174]

    & Fox, J

    Yampolskiy, R. & Fox, J. Safety Engineering for Artificial General Intelligence. Topoi (2012)

    2012

  167. [175]

    & Etzioni, O

    Weld, D. & Etzioni, O. Barley, M. et al. (eds) The First Law of Robotics . (eds Barley, M. et al. ) Safety and Security in Multiagent Systems , Lecture Notes in Computer Science, 90–100 (Springer, Berlin, Heidelberg, 2009)

    2009

  168. [176]

    The responsibility gap: Ascribing responsibility for the actions of learning automata

    Matthias, A. The responsibility gap: Ascribing responsibility for the actions of learning automata. Ethics and Information Technology 6, 175–183 (2004)

    2004

  169. [177]

    Muller, VC

    Farina, L. Muller, VC. (ed.) Artificial Intelligence Systems, Responsibility and Agential Self-Awareness. (ed.Muller, VC.) Philosophy and Theory of Artificial Intelligence 2021 , Vol. 63, 15–25 (2022)

    2022

  170. [178]

    Lee, A. T. Flight simulation: virtual environments in aviation (Routledge, 2017)

    2017

  171. [179]

    Weidinger, L. et al. Sociotechnical safety evaluation of generative ai systems. arXiv preprint arXiv:2310.11986 (2023). 29

    2023 arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.