Pith. sign in

REVIEW 2 major objections 6 minor 47 references

Visual uncertainty explanations can calibrate people's trust in AI models.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

People who saw uncertainty visualizations trusted a more certain model more, but the study does not demonstrate that this explanation calibrates trust better than numeric accuracy.

T0 review reviewed 2026-08-04 challenge →

load-bearing objection A transparent first usability study of Unsupervised DeepView whose main 'trust calibration' claim outruns the design: it shows users' trust tracks the uncertainty display, not measured trustworthiness. the 2 major comments →

arxiv 2509.08989 v1 pith:AIV2QIG4 submitted 2025-09-10 cs.AI

Uncertainty Awareness and Trust in Explainable AI- On Trust Calibration using Local and Global Explanations

Carina Newen , Daniel Bodemer , Sonja Glantz , Emmanuel M\"uller , Magdalena Wischnewski , Lenka Schnaubert This is my paper
classification cs.AI
keywords trust calibrationexplainable AIuncertainty visualizationglobal explanationslocal explanationsuser studyUnsupervised DeepViewhuman-AI trust
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

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 reports a user study testing whether Unsupervised DeepView, an explainable-AI method that visualizes where a machine-learning model is uncertain, helps lay users calibrate their trust. The central finding is that participants who saw the visual explanation rated a regularized, less uncertain model as significantly more trustworthy than an unregularized, more uncertain model, even though both were presented with the same numerical accuracy. The paper also found that the visual explanation did not outperform plain numeric accuracy percentages on explanation satisfaction or on a trust-in-automation scale, and it uses the open-ended responses to derive practical guidelines for designing XAI algorithms, such as simplifying color schemes and combining global and local views. If the result holds, it suggests that global uncertainty visualizations can serve as a trust-calibration tool for non-experts, complementing rather than replacing numeric metrics.

Core claim

The paper's core claim is that Unsupervised DeepView, an algorithm that produces a global scatter-plot view of a dataset plus local examples labeled certain or uncertain, is capable of trust calibration. In an online experiment with 196 participants, the visual group saw two learners with identical test and validation accuracies but different uncertainty visualizations; the learner with fewer uncertain pixels and sensible uncertainty labels was rated significantly more trustworthy across all items (paired t-test t(99) = -5.17, p < 0.001, Cohen's d = -0.517). The authors take this as evidence that the visual explanation allowed users to distinguish a better-generalizing model from a worse one

What carries the argument

Unsupervised DeepView is an explainable AI algorithm that visualizes uncertainty and robustness for high-dimensional data: it projects data points into a global scatter plot where darker blue regions indicate higher model certainty and light blue or white regions indicate uncertainty, alongside local example images labeled certain or uncertain. The study uses this visualization as the experimental stimulus, and the argument rests on the contrast between two learners that share numerical accuracy but differ in the amount and plausibility of the uncertainty pixels shown. The Explanation Satisfaction Scale and the Trust in Automation Scale are the measurement instruments used to quantify user r

Load-bearing premise

The load-bearing premise is that the higher trust ratings for the less uncertain model were produced by the visual explanation itself, rather than by a simpler cue such as the visible number of uncertain pixels, and that those ratings reflect trustworthiness rather than surface features.

What would settle it

Present the same two learners (same numerical accuracies, same uncertainty counts) to a new group using only plain text—'Model A estimates 5% of images as uncertain, Model B estimates 20%'—without the DeepView visualization; if the trust difference is just as large, the visual explanation is not the active ingredient. Alternatively, eye-tracking showing that participants who rated correctly ignored the global image would weaken the causal claim.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • If Unsupervised DeepView calibrates trust, then global uncertainty visualizations can be used as a practical tool for helping non-experts differentiate better from worse models, even when numeric accuracy alone is ambiguous.
  • Combining global and local explanations appears to support interpretability: participants based decisions on both the global image and the local uncertain examples, roughly evenly.
  • The absence of a satisfaction or trust advantage over numeric values suggests that complex XAI methods may not need to outperform simple numbers on subjective liking to still be useful for calibration.
  • Design guidelines derived from open answers: keep the amount of information minimal, use fewer and more clearly distinguished colors, and consider separate explanations for expert and lay audiences.
  • The mistrust pattern (p = 0.070) suggests a trade-off: a 'complicated' visual explanation may be seen as more obscure even when it supports calibration, pointing to a need for explanatory onboarding.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • A natural extension not tested here is that the observed effect could be driven by a simple numeric cue: the visual condition differed in the number of uncertain pixels, and participants might have used that count as an easy heuristic; a follow-up presenting the same uncertainty counts as plain numbers could separate the visual contribution.
  • The paper's trust-calibration test only compared two models within the visual condition; without a control condition receiving the same scenario with only a textual summary of uncertainty, the causal role of the visualization remains an inference.
  • The finding suggests a general evaluation protocol for XAI algorithms: before testing subjective satisfaction, check whether the explanation moves trust in the direction of objective model quality; this protocol could be applied to other uncertainty-aware explanation methods.
  • If trust calibration via global uncertainty visualizations replicates, it would motivate integrating such visualizations into high-stakes decision support systems, but only after testing with domain experts and real tasks rather than crowdsourced participants.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 6 minor

Summary. The paper reports a user study (N=196) comparing a purely numeric explanation (test/validation accuracy) with Unsupervised DeepView, a visual global-and-local uncertainty explanation, on trust calibration and explanation satisfaction. Participants rated two machine learners either from accuracy numbers (control) or from accuracy numbers plus DeepView visualizations (experimental). A baseline scenario showed that participants trusted a model with higher validation accuracy. In the visual condition, participants trusted a regularized learner more than an unregularized one (t(99)=-5.17, p<0.001, d=-0.517), which the authors interpret as evidence that Unsupervised DeepView supports trust calibration. Secondary analyses found no significant difference in explanation satisfaction, a marginally non-significant mistrust difference, and no trust difference. Open-ended responses yielded design guidelines for XAI algorithms.

Significance. If the central claim were sound, the paper would be a valuable empirical evaluation of an understudied class of XAI (global uncertainty explanations), with a preregistered power analysis, well-established scales, and transparent reporting of both significant and non-significant results. However, the central construct—trust calibration—is operationalized without an objective ground-truth measure of model trustworthiness, and the key comparison lacks a matched numeric-only control. The contribution is therefore currently suggestive rather than conclusive. The paper does offer useful exploratory observations and design recommendations that could inform future, better-controlled studies.

major comments (2)
  1. [§IV.B and Table III; §II] The central claim that 'Unsupervised DeepView is capable of some form of trust calibration' rests on a significant paired t-test between trust ratings for two learners. However, trust calibration is defined in §II as 'the balance between human trust and the actual trustworthiness of the application.' In Scenario 2 (§III.B), the two learners are given identical test and validation accuracy, and no independent measure of their actual trustworthiness (e.g., held-out accuracy, expected calibration error, adversarial robustness) is reported. The observed difference shows only that participants' trust ratings were sensitive to the uncertainty display, not that the ratings were calibrated to any external criterion. If the two models are in fact equally trustworthy, the result could represent miscalibration. This gap directly undermines the interpretation of the main result.
  2. [§III.B Scenario 2; Fig. 4] The two learners in Scenario 2 differ in a simple quantitative feature: 'Machine learner one had much fewer global predicted uncertainties as compared to machine learner two.' This count is a numeric cue that could drive trust ratings independently of the richer visual explanation (global image and example instances). Scenario 2 was presented only to the visual group, so there is no analogous numeric-only comparison for this exact choice, and the visual pattern is confounded with the uncertainty count. Consequently, the t-test cannot attribute the trust difference to the visual explanation per se. An additional control condition presenting the uncertainty counts as numbers, or an analysis covarying the count, is needed.
minor comments (6)
  1. [Table III caption] The caption reads 'DO PERCENTAGE ESTIMATIONS LEAD TO USEFUL TRUST CALIBRATION?' but the table reports the visual-explanation Scenario 2. The caption should be corrected to describe the visual condition.
  2. [§IV.D and Table V] The text says the mistrust-scale comparison 'resulting in a statistically significant result,' but Table V reports p=0.070, which is not significant at the conventional α=0.05 level. This is a factual error that should be corrected. Also, throughout, 'no difference' should be phrased as 'no statistically significant difference' (e.g., §IV.C).
  3. [Fig. 4] The figure labels are in German ('unsicher', 'sicher'). Provide English translations or at least a figure caption that defines them, since the paper is written in English.
  4. [§IV.E] The text refers to 'the following Table 6,' but no table appears in the manuscript at that point. The quantitative breakdown of what participants based their decisions on is missing or mis-referenced; please insert the table or correct the reference.
  5. [Related Work] There is a typo: 'Shapely values' should be 'Shapley values' (also in Table I line 'Shapley values' is correct). Also, decimal commas in §III.A ('Cronbach’s alpha = 0,684') are inconsistent with the rest of the manuscript.
  6. [§V] The limitation about 'different amounts of information in the two conditions' is acknowledged, but the missing ground-truth trustworthiness measure is not listed. Please add this to the limitations, as it directly qualifies the calibration claim.

Circularity Check

1 steps flagged

Central trust-calibration result reduces to a check that human trust follows Unsupervised DeepView's own uncertainty output.

specific steps
  1. self definitional [Section III.B (Scenario 2) and Section IV.B (Evaluation 2)]
    "This is an expected outcome since we designed learner two to be more uncertain. We conclude that Unsupervised DeepView is capable of some form of trust calibration."

    Trust calibration is defined as 'the balance between human trust and the actual trustworthiness of the application' (Section II). The experiment operationalizes 'objectively good' learners by Unsupervised DeepView's own uncertainty output: the two learners have identical test/validation accuracy, and the only difference is that 'one learner labelled instances that looked badly written as uncertain, therefore providing a trustworthy estimate of uncertain and certain, and the other did the opposite.' Learner one is declared more trustworthy solely because DeepView reports fewer uncertain pixels. The observed t-test therefore shows only that participants' trust tracked the uncertainty cue displayed by the algorithm under evaluation. Calling this 'trust calibration' reduces the target (actual

full rationale

The paper is primarily a user study with independent behavioral measurements, and the authors' self-citation to Unsupervised DeepView [30] is not in itself circular: the algorithm is the intervention, not a hidden premise. However, the central claim—that Unsupervised DeepView supports trust calibration—rests on an operationalization that equates 'actual trustworthiness' with the algorithm's own uncertainty estimate. In Scenario 2, both learners had the same test and validation accuracy; the only distinguishing information was the DeepView visualization's uncertainty counts and labels. The significant trust difference (t(99)=-5.17) thus demonstrates sensitivity to the visual explanation's uncertainty display, but it does not demonstrate calibration to an independent ground truth. The paper explicitly acknowledges that 'verifying this alignment is practically unattainable' and substitutes a comparison between 'objectively good' and 'bad' learners, yet the objective status is never externally validated—it is defined by the algorithm under test. This makes the central inference partially self-definitional. Other results (satisfaction, trust-in-automation) are not circular, and no equation-level reduction or fitted parameter is involved. The circularity is therefore concentrated in the interpretation of the headline result, warranting a score of 6 rather than a higher score.

Axiom & Free-Parameter Ledger

0 free parameters · 4 axioms · 0 invented entities

This is an empirical user study, not a derivation; no free parameters are fitted. The central claim rests on the assumption that the relative trust difference is a valid proxy for trust calibration, that the two learners genuinely differ in trustworthiness, and that participants understood the visual explanation as intended. The evaluated algorithm is the authors' own prior work (ref [30]), so the evaluation is not independent in that sense.

axioms (4)
  • domain assumption Trust calibration can be operationalized as a statistically significant difference in trust ratings between an objectively better and worse model.
    Stated in Section III as the foundational test; direct alignment with actual trustworthiness is declared practically unattainable.
  • domain assumption The less regularized learner in Scenario 2 is objectively less trustworthy.
    Section III.B states the algorithm without regularization has more uncertain predictions; no external measure of trustworthiness is provided.
  • domain assumption The numeric accuracy percentages are a fair baseline explanation.
    Section III.A/B; the numeric condition sees only test and validation accuracies.
  • domain assumption The German translations of the Explanation Satisfaction Scale and Trust in Automation Scale retain their psychometric properties.
    Section III.C/D; Cronbach's alphas are reported but no validation of the translations is given.

reviewed 2026-08-04 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Uncertainty Awareness and Trust in Explainable AI- On Trust Calibration using Local and Global Explanations." pith.science (2026). https://pith.science/paper/AIV2QIG4

@misc{pith2026250908989,
  author       = {Pith},
  title        = {Pith review of: Uncertainty Awareness and Trust in Explainable AI- On Trust Calibration using Local and Global Explanations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AIV2QIG4}},
  note         = {Machine review of arXiv:2509.08989}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Explainable AI has become a common term in the literature, scrutinized by computer scientists and statisticians and highlighted by psychological or philosophical researchers. One major effort many researchers tackle is constructing general guidelines for XAI schemes, which we derived from our study. While some areas of XAI are well studied, we focus on uncertainty explanations and consider global explanations, which are often left out. We chose an algorithm that covers various concepts simultaneously, such as uncertainty, robustness, and global XAI, and tested its ability to calibrate trust. We then checked whether an algorithm that aims to provide more of an intuitive visual understanding, despite being complicated to understand, can provide higher user satisfaction and human interpretability.

Figures

Figures reproduced from arXiv: 2509.08989 by Carina Newen, Daniel Bodemer, Emmanuel M\"uller, Lenka Schnaubert, Magdalena Wischnewski, Sonja Glantz.

Figure 1
Figure 1. Figure 1: Overview of the key components of this paper. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: The Trust in Automation Scale C. Open-Format Questions Finally, we asked open questions for both groups of par￾ticipants for a deeper understanding of how people relate to AI, what they currently want from AI, and how difficult/easy each concept is to understand, especially for lay people. The questions were: Which aspects of the explanation did you understand best? Which aspects were most obscure? What wo… view at source ↗
Figure 4
Figure 4. Figure 4: Snapshot of the original study: ”unsicher” means insecure and ”sicher” translates to ”certain”. On the left-hand side, we see machine learner one, the [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Evaluation of Scenario 1 In this Table, the SE difference denotes the standard error difference, and the effect size was measured using Cohen’s d effect size. The mean of machine learner one was M= 2.947 compared to M=2.366 for machine learner two. The standard deviation of machine learner one was 0.8442. Note that a low value indicates more trust, so the second learner was statis￾tically recognized as mor… view at source ↗
Figure 6
Figure 6. Figure 6: Evaluation of Scenario 2 significantly different from the much less uncertain model. This means that our hypothesis that Unsupervised DeepView supports trust calibration was confirmed. C. Evaluation 3 – Does Unsupervised DeepView lead to more satisfactory explanations than simple numerical values? The reliability analysis of the Explanation Satisfaction Scale using Cronbach’s alpha was 0.746, which means i… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

47 extracted references · 6 linked inside Pith

  1. [1]

    In: 18th International Conference on Autonomous Agents and Multiagent Sys- tems (AAMAS 2019), Montreal, Canada, May 13–17, 2019

    Anjomshoae, S., Najjar, A., Calvaresi, D., Fr ¨amling, K.: Explainable agents and robots: Results from a systematic literature review. In: 18th International Conference on Autonomous Agents and Multiagent Sys- tems (AAMAS 2019), Montreal, Canada, May 13–17, 2019. pp. 1078–

    2019

  2. [2]

    arXiv preprint arXiv:2006.06848 (2020)

    Antor ´an, J., Bhatt, U., Adel, T., Weller, A., Hern ´andez-Lobato, J.M.: Getting a clue: A method for explaining uncertainty estimates. arXiv preprint arXiv:2006.06848 (2020)

    Pith/arXiv arXiv 2006

  3. [3]

    Information fusion 58, 82–115 (2020)

    Arrieta, A.B., D ´ıaz-Rodr´ıguez, N., Del Ser, J., Bennetot, A., Tabik, S., Barbado, A., Garc ´ıa, S., Gil-L ´opez, S., Molina, D., Benjamins, R., et al.: Explainable artificial intelligence (xai): Concepts, taxonomies, opportunities and challenges toward responsible ai. Information fusion 58, 82–115 (2020)

    2020

  4. [4]

    International Journal of Social Robotics3, 41–52 (2011)

    Bainbridge, W.A., Hart, J.W., Kim, E.S., Scassellati, B.: The benefits of interactions with physically present robots over video-displayed agents. International Journal of Social Robotics3, 41–52 (2011)

    2011

  5. [5]

    In: Machine Learning and Knowledge Discovery in Databases: European Conference, ECML PKDD 2018, Dublin, Ireland, September 10–14, 2018, Proceedings, Part I 18

    Casalicchio, G., Molnar, C., Bischl, B.: Visualizing the feature impor- tance for black box models. In: Machine Learning and Knowledge Discovery in Databases: European Conference, ECML PKDD 2018, Dublin, Ireland, September 10–14, 2018, Proceedings, Part I 18. pp. 655–670. Springer (2019)

    2018

  6. [6]

    Applied Soft Computing91, 106263 (2020)

    Dastile, X., Celik, T., Potsane, M.: Statistical and machine learning models in credit scoring: A systematic literature survey. Applied Soft Computing91, 106263 (2020)

    2020

  7. [7]

    IEEE signal processing magazine 29(6), 141–142 (2012)

    Deng, L.: The mnist database of handwritten digit images for machine learning research [best of the web]. IEEE signal processing magazine 29(6), 141–142 (2012)

    2012

  8. [8]

    Information Fusion81, 143–153 (2022)

    Dieber, J., Kirrane, S.: A novel model usability evaluation framework (muse) for explainable artificial intelligence. Information Fusion81, 143–153 (2022)

    2022

  9. [9]

    arXiv preprint arXiv:1502.03519 (2015)

    Dong, X.L., Gabrilovich, E., Murphy, K., Dang, V ., Horn, W., Lugaresi, C., Sun, S., Zhang, W.: Knowledge-based trust: Estimating the trustwor- thiness of web sources. arXiv preprint arXiv:1502.03519 (2015)

    Pith/arXiv arXiv 2015

  10. [10]

    International journal of human-computer studies58(6), 697–718 (2003)

    Dzindolet, M.T., Peterson, S.A., Pomranky, R.A., Pierce, L.G., Beck, H.P.: The role of trust in automation reliance. International journal of human-computer studies58(6), 697–718 (2003)

    2003

  11. [11]

    Computers in Human Behavior115, 106607 (2021)

    Gillath, O., Ai, T., Branicky, M.S., Keshmiri, S., Davison, R.B., Spauld- ing, R.: Attachment and trust in artificial intelligence. Computers in Human Behavior115, 106607 (2021)

    2021

  12. [12]

    Academy of Management Annals14(2), 627–660 (2020)

    Glikson, E., Woolley, A.W.: Human trust in artificial intelligence: Review of empirical research. Academy of Management Annals14(2), 627–660 (2020)

    2020

  13. [13]

    Goodfellow, I.J., Shlens, J., Szegedy, C.: Explain- ing and harnessing adversarial examples (2015), https://arxiv.org/abs/1412.6572

    Pith/arXiv arXiv 2015

  14. [14]

    arXiv preprint arXiv:1812.04608 (2018)

    Hoffman, R.R., Mueller, S.T., Klein, G., Litman, J.: Metrics for ex- plainable ai: Challenges and prospects. arXiv preprint arXiv:1812.04608 (2018)

    Pith/arXiv arXiv 2018

  15. [15]

    In: Proceedings of the 2021 ACM conference on fairness, accountability, and transparency

    Jacovi, A., Marasovi ´c, A., Miller, T., Goldberg, Y .: Formalizing trust in artificial intelligence: Prerequisites, causes and goals of human trust in ai. In: Proceedings of the 2021 ACM conference on fairness, accountability, and transparency. pp. 624–635 (2021)

    2021

  16. [16]

    Kirsch, A.: Explain to whom? putting the user in the center of ex- plainable ai. In: Proceedings of the First International Workshop on Comprehensibility and Explanation in AI and ML 2017 co-located with 16th International Conference of the Italian Association for Artificial Intelligence (AI* IA 2017) (2017)

    2017

  17. [17]

    Artificial Intelligence in medicine23(1), 89–109 (2001)

    Kononenko, I.: Machine learning for medical diagnosis: history, state of the art and perspective. Artificial Intelligence in medicine23(1), 89–109 (2001)

    2001

  18. [18]

    In: International conference on machine learning

    Kumar, I.E., Venkatasubramanian, S., Scheidegger, C., Friedler, S.: Problems with shapley-value-based explanations as feature importance measures. In: International conference on machine learning. pp. 5491–

  19. [19]

    Journal of communication 56(4), 754–772 (2006)

    Lee, K.M., Peng, W., Jin, S.A., Yan, C.: Can robots manifest personal- ity?: An empirical test of personality recognition, social responses, and social presence in human–robot interaction. Journal of communication 56(4), 754–772 (2006)

    2006

  20. [20]

    In: Proceedings of the 2020 CHI conference on human factors in computing systems

    Liao, Q.V ., Gruen, D., Miller, S.: Questioning the ai: informing design practices for explainable ai user experiences. In: Proceedings of the 2020 CHI conference on human factors in computing systems. pp. 1–15 (2020)

    2020

  21. [21]

    In: Proceedings of the SIGCHI conference on human factors in computing systems

    Lim, B.Y ., Dey, A.K., Avrahami, D.: Why and why not explanations improve the intelligibility of context-aware intelligent systems. In: Proceedings of the SIGCHI conference on human factors in computing systems. pp. 2119–2128 (2009)

    2009

  22. [22]

    Nature machine intelligence2(1), 56–67 (2020)

    Lundberg, S.M., Erion, G., Chen, H., DeGrave, A., Prutkin, J.M., Nair, B., Katz, R., Himmelfarb, J., Bansal, N., Lee, S.I.: From local explanations to global understanding with explainable ai for trees. Nature machine intelligence2(1), 56–67 (2020)

    2020

  23. [23]

    Advances in neural information processing systems30 (2017)

    Lundberg, S.M., Lee, S.I.: A unified approach to interpreting model predictions. Advances in neural information processing systems30 (2017)

    2017

  24. [24]

    In: International Conference on Artificial Intelligence and Statistics

    Marx, C., Park, Y ., Hasson, H., Wang, Y ., Ermon, S., Huan, L.: But are you sure? an uncertainty-aware perspective on explainable ai. In: International Conference on Artificial Intelligence and Statistics. pp. 7375–7391. PMLR (2023)

    2023

  25. [25]

    In: MILCOM 2015-2015 IEEE Military Communications Conference

    Mayhew, M., Atighetchi, M., Adler, A., Greenstadt, R.: Use of machine learning in big data analytics for insider threat detection. In: MILCOM 2015-2015 IEEE Military Communications Conference. pp. 915–922. IEEE (2015)

    2015

  26. [26]

    Academy of management journal38(1), 24–59 (1995)

    McAllister, D.J.: Affect-and cognition-based trust as foundations for interpersonal cooperation in organizations. Academy of management journal38(1), 24–59 (1995)

    1995

  27. [27]

    In: Proceedings of the Human Factors and Ergonomics Society Annual Meeting

    McDermott, P.L., Brink, R.N.t.: Practical guidance for evaluating cal- ibrated trust. In: Proceedings of the Human Factors and Ergonomics Society Annual Meeting. vol. 63, pp. 362–366. SAGE Publications Sage CA: Los Angeles, CA (2019)

    2019

  28. [28]

    ACM Transactions on Interactive Intelligent Systems (TiiS)11(3-4), 1–45 (2021)

    Mohseni, S., Zarei, N., Ragan, E.D.: A multidisciplinary survey and framework for design and evaluation of explainable ai systems. ACM Transactions on Interactive Intelligent Systems (TiiS)11(3-4), 1–45 (2021)

    2021

  29. [29]

    In: Proceedings of the 3rd ACM CoNEXT Workshop on Big DAta, Machine Learning and Artificial Intelligence for Data Communication Networks

    Morichetta, A., Casas, P., Mellia, M.: Explain-it: Towards explainable ai for unsupervised network traffic analysis. In: Proceedings of the 3rd ACM CoNEXT Workshop on Big DAta, Machine Learning and Artificial Intelligence for Data Communication Networks. pp. 22–28 (2019)

    2019

  30. [30]

    In: 2022 IEEE International Conference on Knowledge Graph (ICKG)

    Newen, C., M ¨uller, E.: Unsupervised deepview: Global explainability of uncertainties for high dimensional data. In: 2022 IEEE International Conference on Knowledge Graph (ICKG). pp. 196–202. IEEE (2022)

    2022

  31. [31]

    Zeitschrift f ¨ur Arbeitswissenschaft3(70), 151–160 (2016)

    P ¨ohler, G., Heine, T., Deml, B.: Itemanalyse und faktorstruktur eines fragebogens zur messung von vertrauen im umgang mit automatischen systemen. Zeitschrift f ¨ur Arbeitswissenschaft3(70), 151–160 (2016)

    2016

  32. [32]

    In: Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining

    Ribeiro, M.T., Singh, S., Guestrin, C.: ” why should i trust you?” explaining the predictions of any classifier. In: Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining. pp. 1135–1144 (2016)

    2016

  33. [33]

    In: Proceedings of the AAAI conference on artificial intelligence

    Ribeiro, M.T., Singh, S., Guestrin, C.: Anchors: High-precision model- agnostic explanations. In: Proceedings of the AAAI conference on artificial intelligence. vol. 32 (2018)

    2018

  34. [34]

    CEUR Workshop Proceedings (2019)

    Ribera, M., Lapedriza Garc ´ıa, `A.: Can we do better explanations? a proposal of user-centered explainable ai. CEUR Workshop Proceedings (2019)

    2019

  35. [35]

    Autonomous Agents and Multi-Agent Systems33, 673–705 (2019)

    Rosenfeld, A., Richardson, A.: Explainability in human–agent systems. Autonomous Agents and Multi-Agent Systems33, 673–705 (2019)

    2019

  36. [36]

    In: Proceedings of the tenth annual ACM/IEEE international conference on human-robot interaction

    Salem, M., Lakatos, G., Amirabdollahian, F., Dautenhahn, K.: Would you trust a (faulty) robot? effects of error, task type and personality on human-robot cooperation and trust. In: Proceedings of the tenth annual ACM/IEEE international conference on human-robot interaction. pp. 141–148 (2015)

    2015

  37. [37]

    arXiv preprint arXiv:1909.09154 (2019)

    Schulz, A., Hinder, F., Hammer, B.: Deepview: Visualizing classification boundaries of deep neural networks as scatter plots using discriminative dimensionality reduction. arXiv preprint arXiv:1909.09154 (2019)

    Pith/arXiv arXiv 1909

  38. [38]

    Artificial Intelligence294, 103457 (2021)

    Setzu, M., Guidotti, R., Monreale, A., Turini, F., Pedreschi, D., Gi- annotti, F.: Glocalx-from local to global explanations of black box ai models. Artificial Intelligence294, 103457 (2021)

    2021

  39. [39]

    arXiv preprint arXiv:2105.11828 (2021)

    Seuß, D.: Bridging the gap between explainable ai and uncertainty quantification to enhance trustability. arXiv preprint arXiv:2105.11828 (2021)

    Pith/arXiv arXiv 2021

  40. [40]

    International Journal of Human-Computer Studies146, 102551 (2021)

    Shin, D.: The effects of explainability and causability on perception, trust, and acceptance: Implications for explainable ai. International Journal of Human-Computer Studies146, 102551 (2021)

    2021

  41. [41]

    Cutter business technology journal31(2), 47–53 (2018)

    Siau, K., Wang, W.: Building trust in artificial intelligence, machine learning, and robotics. Cutter business technology journal31(2), 47–53 (2018)

    2018

  42. [42]

    International Journal of Control62(6), 1391–1407 (1995)

    Sj ¨oberg, J., Ljung, L.: Overtraining, regularization and searching for a minimum, with application to neural networks. International Journal of Control62(6), 1391–1407 (1995)

    1995

  43. [43]

    Advances in neural information processing systems34, 9391–9404 (2021)

    Slack, D., Hilgard, A., Singh, S., Lakkaraju, H.: Reliable post hoc explanations: Modeling uncertainty in explainability. Advances in neural information processing systems34, 9391–9404 (2021)

    2021

  44. [44]

    Sundararajan, M., Taly, A., Yan, Q.: Axiomatic attribution for deep networks (2017)

    2017

  45. [45]

    Journal of cognitive engineering and decision making13(4), 279–294 (2019)

    Tenhundfeld, N.L., De Visser, E.J., Haring, K.S., Ries, A.J., Finomore, V .S., Tossell, C.C.: Calibrating trust in automation through familiarity with the autoparking feature of a tesla model x. Journal of cognitive engineering and decision making13(4), 279–294 (2019)

    2019

  46. [46]

    In: Proceedings of the 2020 conference on fairness, accountability, and transparency

    Toreini, E., Aitken, M., Coopamootoo, K., Elliott, K., Zelaya, C.G., Van Moorsel, A.: The relationship between trust in ai and trustworthy machine learning technologies. In: Proceedings of the 2020 conference on fairness, accountability, and transparency. pp. 272–283 (2020)

    2020

  47. [1088]

    International Foundation for Autonomous Agents and Multiagent Systems (2019)

    2019

This paper was first reviewed by deepseek-v4-flash on August 4, 2026.