Pith. sign in

REVIEW 4 major objections 7 minor 65 references

AI Harmonics: a human-centric and harms severity-adaptive AI risk assessment framework

T0 review · 4 major / 7 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read The paper proposes AI Harmonics, a human-centric framework whose AIH metric measures AI harm concentration from ordinal severity labels alone, and reports that Political & economic and Physical harms are the most concentrated and most…

desk verdict The framework is usable and the empirical content is new, but AIH is a normalized mean severity rank, not a concentration measure, and the validation against CI is circular. read the letter →

arxiv 2509.10104 v1 pith:AJA7QIK3 submitted 2025-09-12 cs.AI stat.ME

Sofia Vei , Paolo Giudici , Pavlos Sermpezis , Athena Vakali , Adelaide Emma Bernardelli This is my paper · ORCID
classification cs.AIstat.ME
keywords AIharmsriskprioritizationordinalseverityharmconcentrationLorenzcurveCriticalityIndexstakeholderimpactsgovernance
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

The paper aims to turn AI risk assessment away from internal compliance checks and toward the people who actually experience AI harms, and it argues this can be done from rank-only severity judgments. Its central proposal is the AIH metric, defined as the area under a derivative Lorenz curve, which measures how concentrated a harm category's impact is across stakeholder groups without assuming numeric severity values. Applied to a corpus of expert-annotated incidents, the metric ranks Political & economic harms (AIH 0.85) and Physical harms (0.73) as the most concentrated and thus most urgent, with Financial & business (0.51) and Autonomy (0.53) at the bottom. The paper further claims the pipeline is dataset-agnostic and robust: harm-category rankings stay nearly unchanged under random permutations of the severity ordering and under removal of up to 80 percent of annotations.

What carries the argument

The load-bearing object is the derivative Lorenz curve $\ell_i(x)$. For harm category $c_i$, order the $M$ stakeholder groups by severity rank and plot cumulative frequencies $\sum_{j=0}^k f_{ij}$ against normalized severity ranks $k/M$; the metric is $AIH_i = \int_0^1 \ell_i(x)\,dx$. Because the y-axis uses ranks rather than cumulative severities, the curve needs only a total ordering of severity and is invariant to the numeric values attached to ranks. This is what lets the framework operate where numerical severity scores are unavailable or unreliable. AIH collapses to the standard Gini when numeric severities exist and relates to the Criticality Index by $AIH = CI\cdot(M-1)/M + 1/(2M)$, which is the formal link used to validate it.

What would settle it

Elicit category-specific severity orderings from a stakeholder panel and recompute AIH; if the top-ranked harm categories change from the paper's headline ranking, the central prioritization result depends on the assumed severity ordering rather than on the data itself.

Watch

Extended reading notes

Core claim

The discovery is that a Gini-style concentration measure can be rebuilt for ordinal severity data by swapping the Lorenz curve for its derivative. For each harm category, stakeholders are ranked from least to most severely affected, and the derivative Lorenz curve plots the cumulative share of stakeholders against the normalized severity rank; AIH is the integral of that curve. The paper shows that AIH is a linear transform of the Criticality Index, so the two ordinal metrics carry the same information, and reports that on its benchmark data Political & economic harms are by far the most concentrated (AIH 0.85), followed by Physical and Psychological harms at 0.73, while Financial & business and Autonomy are the most evenly spread. The conclusion is that harm concentration, measured this way, identifies where mitigation is most urgent.

Load-bearing premise

The ranking assumes one fixed ordering of how severely different stakeholder groups are harmed applies to every harm category; if severity order varies by category, the AIH values and resulting priorities can change.

Editorial extensions

If this is right

  • If AIH is correct, any structured incident dataset with ordinal severity labels can be converted into a prioritized list of harm categories without requiring numerical loss estimates.
  • On the benchmark data, mitigation resources should be directed first to Political & economic and Physical harms, since these show the sharpest concentration of severe harm.
  • The reported stability (Spearman correlation at least 0.97 under up to five random swaps of the severity ordering) means that reasonable disagreement about which stakeholder suffers more should not change the harm ranking.
  • Because the pipeline recomputes the metric from annotations, revised or newly collected severity judgments can be absorbed without redesigning the assessment.
  • The framework extends to datasets without stakeholder annotations by treating incidents or categories as the ranked units, so the same procedure applies across a range of incident repositories.

Reading between the lines

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

  • Inference: the fixed severity ordering across stakeholder groups is an input assumption, not a data product; if severity is category-dependent, per-category orderings could produce different priorities. The paper's own boundary analysis shows Political & economic AIH moving from 0.13 to 0.86 under extreme reorderings, so the assumption is load-bearing.
  • Inference: a decisive extension would be to run the pipeline with severity orderings elicited separately for each harm category from stakeholder panels and compare the resulting rankings to the headline ones.
  • Inference: the framework could be evaluated predictively instead of descriptively, by testing whether the harm categories flagged as most concentrated are also those where interventions reduce the frequency or severity of reported incidents.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 7 minor

Summary. The paper proposes AI Harmonics, a pipeline for prioritizing AI harm categories from ordinal severity annotations. The central contribution is a new metric AIH, defined as the area under a "derivative Lorenz curve" whose x-coordinates are cumulative stakeholder frequencies and whose y-coordinates are normalized severity ranks. The authors apply AIH and the Criticality Index (CI) to 816 AIAAIC annotations across nine harm categories and nine stakeholder groups, report that Political & Economic and Physical harms have the highest concentration, and support the findings with permutation, boundary, and annotation-removal sensitivity analyses plus an open-source implementation.

Significance. The framework's strengths are its dataset-agnostic design, use of real expert-annotated incidents, extensive sensitivity analyses, and public code and dashboard. If the metric were valid, the paper would offer a practical ordinal alternative to numeric risk scoring. However, the central mathematical claim is not correct: AIH is an affine function of the mean severity rank, so it is a location statistic rather than a concentration/inequality measure. The CI validation is tautological given Eq. (4), the printed CI formula contradicts Table 4, and the claimed convergence to the classic Gini is unsupported. These issues bear directly on the paper's stated contribution and on the headline empirical ranking.

major comments (4)
  1. [Section 4, Eqs. (1) and (3); Section 8] AIH is an affine function of the mean severity rank, not a measure of inequality or concentration. With M ordered ranks and frequencies f_i1,...,f_iM, the integral of the linear interpolation of the points (sum_{j=0}^k f_ij, k/M) equals (sum_j j f_ij - 1/2)/M, i.e., (mean rank - 1/2)/M. For M=9, both the perfectly even distribution f_j=1/9 and the highly uneven two-point distribution f_1=f_9=1/2 have mean rank 5 and therefore both give AIH=0.5. This directly contradicts the Section 8 claim that AIH "measures how unevenly harms are distributed" and the Section 4 claim that lower values mean severity is "more evenly distributed." The metric is a location statistic and cannot identify unevenness in the harm distribution.
  2. [Section 4, CI definition and Eq. (4); Table 4] The printed formula F_i^k = sum_{j=1}^k f_ij is an ascending cumulative frequency, which yields CI = (M - mean rank)/(M-1). For the Political & Economic category, using the heatmap frequencies and the Table 3 ordering gives a mean rank of about 8.20, so this ascending CI would be about 0.10, not the 0.89 reported in Table 4. The Table 4 values are reproduced only by a descending-cumulative version, CI = (mean rank - 1)/(M-1). Thus Eq. (4) relies on an unstated convention and the paper's benchmark table is inconsistent with the defining formula.
  3. [Section 2, Section 6.2, Figure 5] The validation of AIH against CI is tautological. Eq. (4) states AIH = CI*(M-1)/M + 1/(2M), so the two metrics are deterministically affine-equivalent; any scatter plot is perfectly collinear up to rounding. The paper itself acknowledges this equivalence, so the claim that "a strong empirical correlation ... confirms that our method faithfully captures ordinal concentration" is circular. The accompanying assertion that "AIH emphasizes inequality at the extremes while CI captures average severity rank" is false: both are affine transforms of the mean severity rank.
  4. [Section 4 and Section 8] The claim that AIH "converges to standard Gini under numerical inputs" is unsupported and, as stated, false. Numeric severities never enter the AIH formula, which uses only the ranks k/M. The area under the classic Lorenz curve is (1 - Gini)/2, not Gini, and AIH is not that area. For M=9, the uniform distribution has Gini=0 but AIH=0.5, while the two-point distribution f_1=f_9=1/2 has Gini=0.8 but AIH=0.5. The proposed metric therefore does not unify the ordinal and numerical cases in the claimed way, and no proof or formal statement of any limiting result is provided.
minor comments (7)
  1. [Section 6.2] The text says "its highest value, equal to 0.85, is reached for the Political & Economic harm category," but Table 4 lists CI=0.89 and AIH=0.85; please clarify which metric is being described.
  2. [Figure 5] The caption refers to a 45-degree line, but the axes are restricted to the range 0.55-0.9; please state whether the line is y=x over the displayed range or a different reference line.
  3. [Section 4] The notation for cumulative frequencies is inconsistent: F_i^k and F_ij are used interchangeably, and the relationship between them should be stated explicitly.
  4. [Table 9] The Pros/Cons table contains a duplicated row ("Supports pure ordinal severity scales") and the checkmark placement appears misaligned; please proofread the table.
  5. [Section 6.3.1] The sentence saying that the wide best/worst-case range in Table 5 is "confirming the findings of our previous experiments" is confusing, because a large range indicates sensitivity to the severity ordering rather than robustness; please clarify the intended interpretation.
  6. [Section 4 and Section 8] The claim that AIH "seamlessly collapses to the classic numerical Gini" needs a precise theorem or should be removed, since no numerical severity input appears in the AIH formula.
  7. [Appendix A, Table 9] The table attributes "Supports pure ordinal severity scales" to the Gini Index, but the classic Gini requires numeric values; it is the AIH variant that is ordinal, and the table should be corrected.
Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

No new physical or ontological entities are introduced. The main user-supplied input is the severity ordering of stakeholder groups, and the load-bearing domain axiom is that severity is a property of stakeholder groups independent of harm category.

free parameters (1)
  • Stakeholder severity ordering = Artists=1, Subjects=2, Business=3, Investors=4, Workers=5, Users=6, Vulnerable groups=7, Government=8, General public=9
    Ordinal ranks assigned by the authors as an input to the framework; all category-level AIH and CI values and rankings depend on this ordering. The paper labels it illustrative and tests local permutations, but the ordering is not estimated from data.
assumptions (3)
  • domain assumption AIAAIC expert annotations from 2024-03-14 to 2024-04-11 are representative of real-world AI harms.
    The framework and conclusions are demonstrated only on this dataset and annotation window; no external validation set is used.
  • domain assumption Harm severity can be represented by a single total order over stakeholder groups that is valid across all harm categories.
    The framework assigns one severity rank per stakeholder group in Table 3 and applies it to every harm category; category-dependent severities would change the results.
  • standard math Standard Riemann and trapezoidal integration plus the algebraic manipulations in the Corollary are accepted.
    Eq. 4 relies on simple algebra, but the printed CI formula in Section 4 does not match the definition needed for the Corollary, creating an internal inconsistency.

how reviews work

0 comments
Cite this review

Pith. "Pith review of AI Harmonics: a human-centric and harms severity-adaptive AI risk assessment framework." pith.science (2026). https://pith.science/paper/AJA7QIK3

@misc{pith2026250910104,
  author       = {Pith},
  title        = {Pith review of: AI Harmonics: a human-centric and harms severity-adaptive AI risk assessment framework},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AJA7QIK3}},
  note         = {Machine review of arXiv:2509.10104}
}
read the original abstract

The absolute dominance of Artificial Intelligence (AI) introduces unprecedented societal harms and risks. Existing AI risk assessment models focus on internal compliance, often neglecting diverse stakeholder perspectives and real-world consequences. We propose a paradigm shift to a human-centric, harm-severity adaptive approach grounded in empirical incident data. We present AI Harmonics, which includes a novel AI harm assessment metric (AIH) that leverages ordinal severity data to capture relative impact without requiring precise numerical estimates. AI Harmonics combines a robust, generalized methodology with a data-driven, stakeholder-aware framework for exploring and prioritizing AI harms. Experiments on annotated incident data confirm that political and physical harms exhibit the highest concentration and thus warrant urgent mitigation: political harms erode public trust, while physical harms pose serious, even life-threatening risks, underscoring the real-world relevance of our approach. Finally, we demonstrate that AI Harmonics consistently identifies uneven harm distributions, enabling policymakers and organizations to target their mitigation efforts effectively.

Figures

Figures reproduced from arXiv: 2509.10104 by the authors.

Figure 1
Figure 1. Incidents by Harm Category and Stakeholder Group Each cell shows the number of annotated incidents (freq) for a given harm category (rows) and stakeholder group (columns). The color intensity (legend on the right) is also scaled to Frequency. those two groups. In the Autonomy category, users (29) and artists/content creators (15) are most impacted, while in Physical harms, vulnerable groups (22) and users (14) again… view at source ↗
Figure 2
Figure 2. Left: Classic Lorenz curve 𝐿(𝑥) that can be used when numerical severity is available. Right: “Derivative” Lorenz curve 𝓁(𝑥) that can be used when only ordinal severity is available. 5. The AI Harmonics Framework and Methodology This section presents the core methodological framework proposed for assessing the severity of AI-related harms in a systematic, stakeholder-oriented, and data-driven manner. At its foundati… view at source ↗
Figure 3
Figure 3. Overview of the proposed stakeholder-oriented AI harm assessment framework, illustrating the main steps from dataset input to harm prioritization [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: Lorenz Curves for Harm Categories based on Severity [PITH_FULL_IMAGE:figures/full_fig_p016_4.png]
Figure 5
Figure 5. Figure 5: Scatter plot comparing mean the AIH for different harm categories using AIH and the CI. The dashed line indicates the 45-degree line. 6.3. Sensitivity analysis The values of the AIH metric depend on the underlying severity values, and the results in Section 6.1 have be…
Figure 6
Figure 6. Figure 6: Mean AIH per scenario for variations in ordinal values of severity for AIH. example, a permutation can be to reverse the order of Business (from severity 3 to severity 4) and Investors (from severity 4 to severity 3), which means that we change the ordering. Experiment…
Figure 7
Figure 7. Figure 7: Spearman’s rank correlation (𝜌) between scenarios of ordinal-severity perturbations for AIH. are all greater than or equal to 0.98, which implies a very strong agreement in harm prioritization regardless of minor perturbations. These results confirm that the concentrat…
Figure 8
Figure 8. Figure 8: Mean AIH for each harm category under each progressive random annotation removal. Points mark the mean AIH; vertical bars show 95% confidence intervals. policymakers can confidently prioritize the same harm categories even when working with incomplete or evolving incid…
Figure 9
Figure 9. Figure 9: Hierarchical clustering of harm categories by stakeholder impact. Linkage distance on the horizontal axis measures dissimilarity in the distribution of incident counts across stakeholder groups. Categories that merge at low distance share very similar stakeholder-impac…
Figure 10
Figure 10. Figure 10: Scatter plot comparing mean the AIH metric and the Criticality Index for a sample of 20 harm subcategories. 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 Best Case AUC=0.131 Worst Case AUC=0.869 Equality (AUC=0.500) Cumulative Share of Stakeholders Severity [PITH_FULL_IMAG…
Figure 11
Figure 11. Figure 11: AIH boundaries as derivative Lorenz Curves for the category "Political & economic", which appears to have the highest harm concentration among the experiments. D. Appendix: Additional experimental results for sensitivity analysis [PITH_FULL_IMAGE:figures/full_fig_p03…
Figure 12
Figure 12. Figure 12: Boxplots of the AIH for a selected Political & Economic subcategory, across scenarios. Political instability Institutional trust loss Critical infrastructure damage Political manipulation Economic instability Electoral interference Economic/political power 0.82 0.84 0…
Figure 13
Figure 13. Figure 13: Scatter of mean AIH vs. Criticality Index for each Political & Economic subcategory (original ordering). The dashed 45° line denotes perfect agreement between the two measures. To illustrate that robustness holds even within the most unequal category, we present “Poli…
Figure 14
Figure 14. Figure 14: AIH Boxplots for each Political & Economic subcategory. and Political instability, all lie between these extremes and fall close to the 45 degree reference line, confirming that both indices produce consistent subcategory rankings within this top-ranked category [PIT…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

65 extracted references · 56 canonical work pages

  1. [1]

    J. F. Avila-Rieger, P. B. Adkins-Jackson, T. G. Hill-Jarrett, W. R. Robinson, K. M. Keyes, N. Schupf, A. M. Brickman, R. P. Mayeux, J. J. Manly, Early life exposure to structural sexism and late-life memory trajectories among black and white women and men in the united states, Alzheimer’s & Dementia (2024)

    work page 2024

  2. [2]

    J. Jones, Don’t fear artificial intelligence, question the business model: How surveillance capitalists use media to invade privacy, disrupt moral autonomy, and harm democracy, Journal of Communication Inquiry 49 (1) (2025) 6–26

    work page 2025

  3. [3]

    Van Bekkum, F

    M. Van Bekkum, F. Z. Borgesius, Digital welfare fraud detection and the dutch syri judgment, European Journal of Social Security 23 (4) (2021) 323 –340

    work page 2021

  4. [4]

    N. A. Smuha, K. Yeung, The european union’s ai act: beyond motherhood and apple pie?, Available at SSRN (2024)

    work page 2024

  5. [5]

    M. Jariwala, A comparative analysis of the eu ai act and the colorado ai act: Regulatory approaches to artificial intelligence governance, International Journal of Computer Applications 975 (2024) 8887

    work page 2024

  6. [6]

    Novelli, F

    C. Novelli, F. Casolari, A. Rotolo, M. Taddeo, L. Floridi, Taking ai risks seriously: a new assessment model for the ai act, AI & SOCIETY 39 (5) (2024) 2493–2497

    work page 2024

  7. [7]

    I. D. Raji, A. Smart, R. N. White, M. Mitchell, T. Gebru, B. Hutchinson, J. Smith-Loud, D. Theron, P. Barnes, Closing the ai accountability gap: Defining an end-to-end framework for internal algorithmic auditing, in: Proceedings of the 2020 conference on fairness, accountability, and transparency, 2020, pp. 33–44

    work page 2020

  8. [8]

    Thomas, H

    C. Thomas, H. Roberts, J. Mökander, A. Tsamados, M. Taddeo, L. Floridi, The case for a broader approach to ai assurance: addressing “hidden” harms in the development of artificial intelligence, AI & SOCIETY (2024) 1–16

    work page 2024

Show all 65 references
  1. [9]

    it’s everybody’s role to speak up... but not everyone will

    C. M. Lancaster, K. Schulenberg, C. Flathmann, N. J. McNeese, G. Freeman, “it’s everybody’s role to speak up... but not everyone will”: Understanding ai professionals’ perceptions of accountability for ai bias mitigation, ACM Journal on Responsible Computing 1 (1) (2024) 1–30

    2024

  2. [10]

    Braun, I beg to differ: how disagreement is handled in the annotation of legal machine learning data sets, Artificial intelligence and law 32 (3) (2024) 839–862

    D. Braun, I beg to differ: how disagreement is handled in the annotation of legal machine learning data sets, Artificial intelligence and law 32 (3) (2024) 839–862

    2024

  3. [11]

    Paeth, D

    K. Paeth, D. Atherton, N. Pittaras, H. Frase, S. McGregor, Lessons for editors of ai incidents from the ai incident database, in: Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 39, 2025, pp. 28946–28953

    2025

  4. [12]

    Davidson, D

    T. Davidson, D. Warmsley, M. Macy, I. Weber, Automated hate speech detection and the problem of offensive language, in: Proceedings of the international AAAI conference on web and social media, Vol. 11, 2017, pp. 512–515

    2017

  5. [13]

    Raman, N

    D. Raman, N. Madkour, E. R. Murphy, K. Jackson, J. Newman, Intolerable risk threshold recommendations for artificial intelligence, arXiv preprint arXiv:2503.05812 (2025)

    2025 arXiv

  6. [14]

    T. Aven, R. Flage, Risk assessment with broad uncertainty and knowledge characterisation: an illustrating case study, Knowledge in risk assessment and management (2018) 1–26

    2018

  7. [15]

    Flage, T

    R. Flage, T. Aven, E. Zio, P. Baraldi, Concerns, challenges, and directions of development for the issue of representing uncertainty in risk assessment, Risk analysis 34 (7) (2014) 1196–1207

    2014

  8. [16]

    Robinson, Likert scale, in: Encyclopedia of quality of life and well-being research, Springer, 2024, pp

    J. Robinson, Likert scale, in: Encyclopedia of quality of life and well-being research, Springer, 2024, pp. 3917–3918

    2024

  9. [17]

    AI, Algorithmic and Automation Incident and Controversy (AIAAIC), AIAAIC Repository, https://www.aiaaic.org/ aiaaic-repository, licensed under CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/) (2022)

    2022

  10. [18]

    Giudici, E

    P. Giudici, E. Raffinetti, Explainable ai methods in cyber risk measurement, Quality and Reliability Engineering International (2023)

    2023

  11. [19]

    T.LeQuy,A.Roy,V.Iosifidis,W.Zhang,E.Ntoutsi,Asurveyondatasetsforfairness-awaremachinelearning,WileyInterdisciplinaryReviews: Data Mining and Knowledge Discovery 12 (3) (2022) e1452. S. Vei et al.:Preprint submitted to ElsevierPage 24 of 32 AI Harmonics

    2022

  12. [20]

    C. A. Groenewald, A. Garg, S. S. Yerasuri, et al., Smart supply chain management optimization and risk mitigation with artificial intelligence, Naturalista Campano 28 (1) (2024) 261–270

    2024

  13. [21]

    Ashir, K

    F. Ashir, K. Mekonen, The impact of artificial intelligence on auditing: Navigating ethical challenges (2024)

    2024

  14. [22]

    Novelli, F

    C. Novelli, F. Casolari, A. Rotolo, M. Taddeo, L. Floridi, Ai risk assessment: A scenario-based, proportional methodology for the ai act, Digital Society 3 (1) (2024) 13

    2024

  15. [23]

    AI, Artificial intelligence risk management framework (ai rmf 1.0) (2023)

    N. AI, Artificial intelligence risk management framework (ai rmf 1.0) (2023)

    2023

  16. [24]

    X. Qi, Y. Huang, Y. Zeng, E. Debenedetti, J. Geiping, L. He, K. Huang, U. Madhushani, V. Sehwag, W. Shi, et al., Ai risk management should incorporate both safety and security, arXiv preprint arXiv:2405.19524 (2024)

    2024 arXiv

  17. [25]

    A.-K.Becker,O.Dumitrasc,K.Broelemann,Standardizedinterpretablefairnessmeasuresforcontinuousriskscores,in:Forty-firstInternational Conference on Machine Learning

  18. [26]

    Lim, Determinants of socially responsible ai governance, Duke Law & Technology Review 25 (1) (2025) 183–232

    D. Lim, Determinants of socially responsible ai governance, Duke Law & Technology Review 25 (1) (2025) 183–232

    2025

  19. [27]

    A. G. Alliance, Briefing paper series (2024)

    2024

  20. [28]

    Dehankar, S

    P. Dehankar, S. Das, Ethics in ai: Balancing innovation with responsibility, in: Smart Systems: Engineering and Managing Information for Future Success: Navigating the Landscape of Intelligent Technologies, Springer, 2025, pp. 125–136

    2025

  21. [29]

    Y. Zeng, Y. Yang, A. Zhou, J. Z. Tan, Y. Tu, Y. Mai, K. Klyman, M. Pan, R. Jia, D. Song, et al., Air-bench 2024: A safety benchmark based on risk categories from regulations and policies, arXiv preprint arXiv:2407.17436 (2024)

    2024 arXiv

  22. [30]

    Hoffman, R

    F. Hoffman, R. Kesharwani, J. Maynard, How an outside perspective can help an organization enhance their supply chain risk mitigation strategy., Issues in Information Systems 24 (1) (2023)

    2023

  23. [31]

    Karnofsky, Developing ai risk management with the same ambition and urgency as ai products (2024)

    H. Karnofsky, Developing ai risk management with the same ambition and urgency as ai products (2024)

    2024

  24. [32]

    Babaei, P

    G. Babaei, P. Giudici, E. Raffinetti, Explainable fintech lending, Journal of Economics and Business 125-126 (2023)

    2023

  25. [33]

    Giudici, M

    P. Giudici, M. Centurelli, S. Turchetta, Artificial intelligence risk measurement, Expert Systems with Applications 235 (2024) 121220

    2024

  26. [34]

    B. Xia, Q. Lu, H. Perera, L. Zhu, Z. Xing, Y. Liu, J. Whittle, Towards concrete and connected ai risk assessment (c 2 aira): A systematic mapping study, in: 2023 IEEE/ACM 2nd International Conference on AI Engineering–Software Engineering for AI (CAIN), IEEE, 2023, pp. 104–116

    2023

  27. [35]

    Afzal, S

    F. Afzal, S. Yunfei, M. Nazir, S. M. Bhatti, A review of artificial intelligence based risk assessment methods for capturing complexity-risk interdependencies: Cost overrun in construction projects, International Journal of Managing Projects in Business 14 (2) (2021) 300–328

    2021

  28. [36]

    Golpayegani, H

    D. Golpayegani, H. J. Pandit, D. Lewis, Airo: An ontology for representing ai risks based on the proposed eu ai act and iso risk management standards, in: Towards a Knowledge-Aware AI, IOS Press, 2022, pp. 51–65

    2022

  29. [37]

    Kereopa-Yorke, Safeguarding cognitive well-being in an ai-augmented world: A socio-technical systems perspective on neuroethics, technostress, and risk management (2023)

    B. Kereopa-Yorke, Safeguarding cognitive well-being in an ai-augmented world: A socio-technical systems perspective on neuroethics, technostress, and risk management (2023)

    2023

  30. [38]

    A. O. Ikudabo, P. Kumar, Ai-driven risk assessment and management in banking: balancing innovation and security, International Journal of Research Publication and Reviews 5 (10) (2024) 3573–88

    2024

  31. [39]

    Chowdhury, P

    S. Chowdhury, P. Budhwar, G. Wood, Generative artificial intelligence in business: towards a strategic human resource management framework, British Journal of Management (2024)

    2024

  32. [40]

    Caton, C

    S. Caton, C. Haas, Fairness in machine learning: A survey, ACM Computing Surveys 56 (7) (2024) 1–38

    2024

  33. [41]

    S. R. Jordan, Designing artificial intelligence review boards: creating risk metrics for review of ai, in: 2019 IEEE International Symposium on Technology and Society (ISTAS), IEEE, 2019, pp. 1–7

    2019

  34. [42]

    Elhoseny, A

    M. Elhoseny, A. Darwiesh, A. H. El-Baz, J. J. Rodrigues, Enhancing cryptocurrency security using ai risk management model, IEEE Consumer Electronics Magazine 13 (1) (2023) 48–53

    2023

  35. [43]

    Ricciardi Celsi, The dilemma of rapid ai advancements: Striking a balance between innovation and regulation by pursuing risk-aware value creation, Information 14 (12) (2023) 645

    L. Ricciardi Celsi, The dilemma of rapid ai advancements: Striking a balance between innovation and regulation by pursuing risk-aware value creation, Information 14 (12) (2023) 645

    2023

  36. [44]

    Aldasoro, L

    I. Aldasoro, L. Gambacorta, P. Giudici, T. Leach, The drivers of cyber risks, Journal of Financial Stability (2022)

    2022

  37. [45]

    X. Liu, F. Fan, L. Kong, Z. Diao, W. Xie, J. Lu, J. You, Unimodal regularized neuron stick-breaking for ordinal classification, Neurocomputing 388 (2020) 34–44

    2020

  38. [46]

    M. C. Calzarossa, P. Giudici, R. Zieni, How robust are ensemble machine learning explanations?, Neurocomputing 630 (2025) 129686

    2025

  39. [47]

    Facchinetti, P

    S. Facchinetti, P. Giudici, S. A. Osmetti, et al., How to measure cybersecurity risk, in: Book of Short Papers SIS 2018, Vol. 2018, Pearson, 2018, pp. 1643–1645

    2018

  40. [48]

    Salim, J

    S. Salim, J. Jayasudha, B. Soniya, Ensuring ethical ai: Unpacking the significance of risk analysis under the european union’s artificial intelligence act, in: 2024 IEEE Region 10 Symposium (TENSYMP), IEEE, 2024, pp. 1–6

    2024

  41. [49]

    Ballot Jones, J

    L. Ballot Jones, J. Thornton, D. De Silva, Limitations of risk-based artificial intelligence regulation: a structuration theory approach, Discover Artificial Intelligence 5 (1) (2025) 1–13

    2025

  42. [50]

    D. S. Schiff, S. Kelley, J. Camacho Ibáñez, The emergence of artificial intelligence ethics auditing, Big Data & Society 11 (4) (2024) 20539517241299732

    2024

  43. [51]

    A. M. Davani, M. Díaz, V. Prabhakaran, Dealing with disagreements: Looking beyond the majority vote in subjective annotations, Transactions of the Association for Computational Linguistics 10 (2022) 92–110

    2022

  44. [52]

    Abercrombie, D

    G. Abercrombie, D. Benbouzid, P. Giudici, D. Golpayegani, J. Hernandez, P. Noro, H. Pandit, E. Paraschou, C. Pownall, J. Prajapati, et al., A collaborative, human-centred taxonomy of ai, algorithmic, and automation harms, arXiv preprint arXiv:2407.01294 (2024)

    2024 arXiv

  45. [53]

    URLhttps://doi.org/10.1787/d1a8d965-en

    OECD, Defining ai incidents and related terms, OECD Artificial Intelligence Papers 16, OECD Publishing, Paris (2024).doi:10.1787/ d1a8d965-en. URLhttps://doi.org/10.1787/d1a8d965-en

    2024 doi

  46. [54]

    P. Ward, J. Stanley, R. Ferguson, C. M. Gladding, K. J. Burns, Risk discovery protocol for ai assurance (v1.0), Tech. Rep. PR-24-2963, The MITRE Corporation, Bedford, MA (2024). URL https://www.mitre.org/sites/default/files/2025-04/PR-24-2963-Risk-Discovery-Protocol-AI-Assuran...

    2024

  47. [55]

    Giudici, E

    P. Giudici, E. Raffinetti, Rga: a unified measure of predictive accuracy, Advanced in data analysis and classification 19 (2025) 67–93

    2025

  48. [56]

    Facchinetti, P

    S. Facchinetti, P. Giudici, S. Osmetti, Cyber risk measurement with ordinal data, Statistical methods with applications 29 (2019) 173–185

    2019

  49. [57]

    B. E. Strom, A. Applebaum, D. P. Miller, K. C. Nickels, A. G. Pennington, C. B. Thomas, Mitre att&ck: Design and philosophy, in: Technical report, The MITRE Corporation, 2018

    2018

  50. [58]

    MITRE | ATLAS™, Mitre atlas: Adversarial threat landscape for artificial-intelligence systems,https://atlas.mitre.org/, retrieved fromhttps://atlas.mitre.org/(2023)

    2023

  51. [59]

    McGregor, Preventing repeated real world ai failures by cataloging incidents: The ai incident database, in: Proceedings of the AAAI Conference on Artificial Intelligence, Vol

    S. McGregor, Preventing repeated real world ai failures by cataloging incidents: The ai incident database, in: Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 35, 2021, pp. 15458–15463

    2021

  52. [60]

    Slattery, A

    P. Slattery, A. K. Saeri, E. A. C. Grundy, J. Graham, M. Noetel, R. Uuk, J. Dao, S. Pour, S. Casper, N. Thompson, The ai risk repository: A comprehensive meta-review, database, and taxonomy of risks from artificial intelligence (2024).arXiv:2408.12622. URLhttps://arxiv.org/abs...

    2024 arXiv

  53. [61]

    AVID, Ai vulnerability database,https://avidml.org/(2024)

    2024

  54. [62]

    George Washington University Law School, DAIL – Database of AI Litigation, https://blogs.gwu.edu/law-eti/ ai-litigation-database/

  55. [63]

    doi:10.1787/f326d4ac-en

    OECD,Towardsacommonreportingframeworkforaiincidents,Tech.Rep.34,OECDPublishing,Paris(2024). doi:10.1787/f326d4ac-en. URLhttps://doi.org/10.1787/f326d4ac-en

    2024 doi

  56. [64]

    AI Harmonics Project, AI Harmonics Interactive Dashboard,https://aiaaic.csd.auth.gr/(2025)

    2025

  57. [65]

    victim profiles

    AI Harmonics Codebase, AI Harmonics: Code & Data,https://github.com/Datalab-AUTH/AIH(2025). S. Vei et al.:Preprint submitted to ElsevierPage 26 of 32 AI Harmonics A. Appendix:AIHvsCI Table 9: Pros and Cons of Gini Index vs. Criticality Index for AI Harm Prioritization. Gini In...

    2025

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.