Pith. sign in

REVIEW 3 major objections 7 minor 97 references

Towards Uncertainty Aware Task Delegation and Human-AI Collaborative Decision-Making

T0 review · 3 major / 7 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 Showing clinicians where a case sits relative to class centers in embedding space, rather than giving them a bare probability, improves how they use AI predictions for stroke rehabilitation assessment.

desk verdict The paper's headline claim about distance-based confidence scores is confounded by bundled interactive explanations; the UQ comparison and system building are solid, but the user-study result needs heavy revision. read the letter →

arxiv 2505.18066 v1 pith:U6CUIVVX submitted 2025-05-23 cs.HC cs.AIcs.LG

Min Hun Lee , Martyn Zhe Yu Tok This is my paper · ORCID
classification cs.HCcs.AIcs.LG
keywords human-AIcollaborationuncertaintyquantificationconfidencescorestaskdelegationexplainableAIstrokerehabilitationclinicaldecisionsupportappropriatereliance
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 claims that how AI confidence is shown to a human decision-maker changes how well the person uses the AI: distance-based scores, computed by measuring how far a case sits from each class's center in the model's learned embedding space and then visualized alongside similar cases, improved decision accuracy relative to numerical probability scores in a stroke rehabilitation assessment task. In a user study with 29 health professionals and medical/health students (27 retained for analysis), participants using the distance-based interface arrived at 74.34% correct decisions versus 66.14% for those seeing numerical scores, switched wrong decisions into correct ones 7.15% more often, and made 7.14% fewer switches into error (p<0.01). The same distance measure, tested alone, flagged AI's uncertain cases more reliably than probability-based methods, reaching a 93.63% average F1-score in a technical evaluation. The paper positions this as a path toward helping non-expert users understand when to lean on AI rather than trusting or dismissing it wholesale.

What carries the argument

The central object is the distance-based confidence score: for a $K$-class model, the system computes the distance $d$ between the input's representation and the centroid of each class in the model's processed layers, normalizes it by the maximum observed distance ($\hat{d} = 1 - d/d_{\max}$), and applies a softmax to obtain confidence values $e^{\hat{d}} / \sum_{j=0}^{K} e^{\hat{d}_j}$. Around this score the paper builds an interface with two interactive features: a threshold-exploration step where users adjust a confidence cutoff and see AI performance on delegated cases from a held-out dataset, and an embedding visualization that shows the case as a point among its $k$-nearest neighbors and class centroids (using t-SNE and Euclidean distance), with hover and click access to neighbor images, AI performance, and therapist agreement. The distance score does the work of flagging which cases the model is unsure about, while the visualization and examples are what the paper argues make that uncertainty legible and actionable to users without machine-learning backgrounds.

What would settle it

Run a 2x2 user study that crosses the two interface elements independently, numerical scores with and without interactive example explanations and distance-based scores with and without them, and check whether the roughly 8% accuracy advantage of the distance condition survives when explanations are held constant; if it does not, the headline attribution to distance-based confidence scores fails. On the technical side, an independent evaluation on out-of-distribution data where distance-based scores do not rank errors above maximum class probability would also refute the claim.

Watch

Extended reading notes

Core claim

The paper's core claim is that replacing a bare numerical confidence value with a distance-based score shown as an embedding visualization, accompanied by interactive example-based explanations, produces more accurate and better-calibrated human reliance on AI than the same decision-support system without those features. Two results carry this claim: a technical comparison in which a feed-forward neural network using distance-to-centroid uncertainty achieved a 93.63% average F1-score at identifying cases whose AI prediction should be replaced by an expert label, versus 82.86% for maximum class probability (the strongest probability-based alternative), and a 2x2 user study in which the distance-based condition raised correct decisions from 66.14% to 74.34%, raised the rate at which participants changed wrong decisions into right ones, and lowered the rate of changes into wrong answers.

Load-bearing premise

The paper assumes the user-study comparison isolates the effect of distance-based confidence scores, but the experimental condition adds interactive example-based explanations on top of the distance visualization, so the measured gains could come from the explanations or the extra effort they demand rather than from the distance measure itself.

Editorial extensions

If this is right

  • In clinical decision support, distance-to-centroid confidence could serve as the screening criterion for which cases need human review, since it separated uncertain from certain cases more cleanly than probability outputs in the paper's evaluation.
  • Allowing users to set and inspect a delegation threshold before reviewing AI outputs reduced overreliance on wrong AI outputs (by 3.58% with numerical scores and 1.85% with distance-based scores), suggesting that threshold exploration acts as a cognitive forcing function.
  • The distance-based condition converted a small performance decline under numerical scores (from 67% to 66%) into a complementary gain (from 71% to 74%), indicating that human and AI errors overlapped less when uncertainty was visualized.
  • Both expert therapists and novices improved their correct-decision rate (7.94% and 8.33% respectively), so the benefit did not depend on prior domain expertise.

Reading between the lines

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

  • The experimental condition bundles the distance-based visualization together with interactive example-based explanations, so the paper's attribution of the 8.20% gain to distance-based confidence scores is not isolated; a 2x2 ablation crossing the two features independently would be needed to separate their contributions.
  • If the improvement comes largely from the extra analytical engagement the interactive display forces, then simpler interventions that slow users down, such as requiring a written justification before agreeing with AI, might reproduce part of the effect without any embedding visualization.
  • The distance score's advantage on data from healthy participants, inputs the model was not trained to classify, points toward out-of-distribution flagging as its most natural use case; testing the same interface in other domains where models meet unfamiliar inputs would clarify how far the result generalizes.
  • Preference responses hint at a practical tension: several participants found numerical scores faster and easier to read, so real-world adoption may require offering both presentations rather than replacing one with the other.
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 / 7 minor

Summary. The paper presents an AI-based decision support system for physical stroke rehabilitation assessment and investigates two interactive features: (1) exploration of confidence-score thresholds for task delegation to AI, and (2) distance-based visualizations of confidence scores with interactive example-based explanations, compared against a numerical confidence-score baseline. The authors first evaluate several uncertainty quantification methods offline, finding that a feed-forward NN distance-based method best identifies uncertain cases. They then report a 2x2 mixed-design user study with 29 recruited participants (27 analyzed) including therapists and novices. The headline results are that, compared to numerical confidence scores, the distance-based condition yields an 8.20% higher rate of correct decisions, a 7.15% higher rate of changing decisions to correct ones, and a 7.14% lower rate of incorrect changes, with p<0.01 claimed in the abstract.

Significance. If the central claim were supported, the paper would make a useful contribution to human-AI collaborative decision-making by showing how uncertainty visualization can improve appropriate reliance, particularly for non-technical users in a clinical context. The offline comparison of multiple uncertainty quantification techniques on a real clinical dataset is a strength, as is the inclusion of both domain experts and novices in the user study. However, the main causal attribution is currently undermined by a confound in the experimental design, so the significance is contingent on the authors resolving that confound and correcting the statistical overstatements.

major comments (3)
  1. [§3.3.1, Abstract] The two experimental conditions do not isolate distance-based confidence scores. The 'Distance-based Confidence Scores' condition includes all baseline features plus the distance-based visualization and interactive example-based explanations, while the 'Numerical Confidence Scores' condition includes neither. Consequently, the observed improvements (8.20% higher correct decisions, 7.15% higher ChangedRight, 7.14% lower ChangedWrong in some subgroups) could be driven entirely by the interactive example-based explanations (showing neighbor images, therapist agreement, and embedding context) rather than by the distance-based uncertainty score itself. This confound directly affects the headline claim in the abstract and the framing in the title. Section 5.3 does not acknowledge this issue. The authors should either add an experimental condition that presents numerical scores with the same interactive explanations, or reanalyze the data to separate the effect of the confidence-score representation from the effect of the additional interactive explanation content, and revise the claims accordingly.
  2. [Abstract vs. Table 8, §4.3] The abstract reports p<0.01 for all three improvements, but the data in Table 8 do not support this. The overall ChangedRight comparison has p=0.03, and the overall ChangedWrong comparison has p=0.24; only the Exp subgroup shows a significant reduction in ChangedWrong (p=0.0006). The 7.14% lower incorrect-changes figure is not an overall result but a subgroup result. The abstract and Section 5.2 should be corrected to report the exact p-values and to attribute the significance to the appropriate subgroups rather than claiming uniform p<0.01 for all three overall comparisons.
  3. [§5.1, Table 7] The text states that Exp participants had a 3.58% lower overreliance rate (agreeing with wrong AI outputs) with numerical confidence scores and a 1.85% lower rate with distance-based confidence scores, but no p-values are given. Table 7 shows that the overall 'Agree with Wrong AI outputs' comparison has p=0.27 and all subgroup p-values are greater than 0.05. These claims should be reported as non-significant trends with the corresponding p-values, or the analysis should be clarified if a different test or subgroup was used.
minor comments (7)
  1. [§5.2] The text refers to 'Table 4.3' which does not exist; the intended reference is likely Table 2.
  2. [§3.2.3] The formula for distance-based confidence scores uses 'e^{d_j}' in the softmax, but the normalized quantity is defined as ʄd = 1 − d/dmax; the notation should be made consistent (e.g., e^{ʄd_j}) and the summation index should be clearly defined (e.g., over the K classes).
  3. [§4.2.1] The text references 'Appendix. Figure 3' for the complementary performance gain, but the main text earlier refers to 'Figure 3'; please make the cross-referencing consistent.
  4. [§3.3.3] The procedure for determining the default threshold of 40% for the NoExp group is described only as a 'grid-search over confidence thresholds ranging from 0 to 1'; it is unclear whether this was based on the offline evaluation, pilot data, or a separate criterion, and this should be clarified.
  5. [Table 6] Several rows for student participants (e.g., NV06, NV07, NV08, NV09, NV10, NV11, NV12, NV13, NV14, NV18, NV19) have missing entries for 'Setting' or '# of yrs'; consider using 'N/A' or a dash for clarity.
  6. [§4.1.1] The term 'NonExp' is used in this section while the rest of the paper uses 'NoExp'; please unify the terminology.
  7. [§4.5] The post-study questionnaire preference results (11 vs. 7 vs. 9) are reported without any statistical test; at a minimum, state whether the distribution differs from chance or report this as a descriptive trend.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central result is an empirical user-study comparison, not a derivation that reduces to its own inputs.

full rationale

This paper's headline claim is an empirical finding from a 2x2 mixed-design user study, measuring participants' decision correctness and decision changes after interacting with two interface conditions. The outcome metrics are behavioral measurements, not quantities derived from the uncertainty-score formula, so the result does not reduce to the inputs by construction. The distance-based confidence score is a fixed transformation of distances to class centroids with a maximum-distance normalization and softmax; the normalization is computed from the model's embedding data and is not fitted to the user-study labels or to participants' decisions, so this is not a fitted-input-called-prediction case. The Section 3.2.3 comparison of uncertainty quantification techniques uses a standard threshold-replacement evaluation with therapist ground truth; selecting the best-performing UQ method on held-out data and then using it in a subsequent user study is a normal model-selection pipeline, not circularity. The paper includes several self-citations (e.g., [50], [51], [52]) for domain context, dataset construction, and design preferences, but none is load-bearing for the central user-study result, and none is invoked as a uniqueness theorem or as the sole justification for the claimed improvement. The abstract's attribution of the improvement specifically to distance-based uncertainty scores is weakened by the experimental confound that the 'Distance-based Confidence Scores' condition also adds interactive example-based explanations; however, a confound is an internal-validity limitation rather than a circular reduction, and the instructions restrict circularity findings to cases where a specific derivation or fit is equivalent to its own inputs. No such equivalence is present here, so the appropriate finding is no significant circularity.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central claims rest on model hyperparameters tuned by grid search, a data-dependent distance normalization, and domain assumptions about therapist labels and confidence-correctness correlation. No new theoretical entities are introduced.

free parameters (6)
  • ROM model architecture and learning rate = 3x256 hidden units, lr 0.005
    Selected by grid-search with LOSO CV; supplies AI outputs for ROM decisions.
  • Compensation model architecture and learning rate = 3 layers (32/64), lr 0.0005
    Selected by grid-search; supplies AI outputs for Compensation decisions.
  • MCDropout rate = 0.3
    Selected from 0.1 to 0.5 for the Bayesian UQ baseline.
  • NoExp default delegation threshold = 40%
    Empirically determined by grid-search over confidence thresholds in Section 3.3.3.
  • t-SNE dimensionality reduction and k-NN metric = t-SNE, Euclidean, k=5-30
    Chosen from Appendix Table 5; controls the interactive embedding visualization.
  • Distance normalization scale and softmax = 1 - d/d_max, softmax temp 1
    Uses the maximum observed distance as a data-dependent calibration statistic for distance-based confidence.
assumptions (5)
  • domain assumption Therapist annotations from a clinically validated assessment tool are correct ground truth.
    Used to define right/wrong decisions and to evaluate UQ; Section 3.2.1 and Section 3.3.4.
  • domain assumption Correct AI predictions tend to have higher confidence scores than incorrect ones.
    Stated explicitly in Section 3.2.3: 'we assume that successful, right AI predictions tend to be associated with high confidence scores...' This underpins the threshold-based task delegation.
  • standard math Leave-one-subject-out cross-validation gives unbiased estimates of model and UQ performance.
    Used for all model evaluations in Section 3.2.2 and Appendix A.2.
  • domain assumption The held-out dataset shown during threshold exploration is representative of real AI performance.
    Exp participants learn from this held-out performance when setting delegation thresholds; Section 3.2, Figure 1a.
  • domain assumption Within-subject counterbalancing of the two conditions removes carryover effects.
    Assumed by the mixed-design analysis; Section 3.3.3.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards Uncertainty Aware Task Delegation and Human-AI Collaborative Decision-Making." pith.science (2026). https://pith.science/paper/U6CUIVVX

@misc{pith2026250518066,
  author       = {Pith},
  title        = {Pith review of: Towards Uncertainty Aware Task Delegation and Human-AI Collaborative Decision-Making},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/U6CUIVVX}},
  note         = {Machine review of arXiv:2505.18066}
}
abstract

Despite the growing promise of artificial intelligence (AI) in supporting decision-making across domains, fostering appropriate human reliance on AI remains a critical challenge. In this paper, we investigate the utility of exploring distance-based uncertainty scores for task delegation to AI and describe how these scores can be visualized through embedding representations for human-AI decision-making. After developing an AI-based system for physical stroke rehabilitation assessment, we conducted a study with 19 health professionals and 10 students in medicine/health to understand the effect of exploring distance-based uncertainty scores on users' reliance on AI. Our findings showed that distance-based uncertainty scores outperformed traditional probability-based uncertainty scores in identifying uncertain cases. In addition, after exploring confidence scores for task delegation and reviewing embedding-based visualizations of distance-based uncertainty scores, participants achieved an 8.20% higher rate of correct decisions, a 7.15% higher rate of changing their decisions to correct ones, and a 7.14% lower rate of incorrect changes after reviewing AI outputs than those reviewing probability-based uncertainty scores ($p<0.01$). Our findings highlight the potential of distance-based uncertainty scores to enhance decision accuracy and appropriate reliance on AI while discussing ongoing challenges for human-AI collaborative decision-making.

Figures

Figures reproduced from arXiv: 2505.18066 by the authors.

Figure 1
Figure 1. (a) Users can explore different thresholds of confidence scores, review AI performance on delegated cases from a [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Interface for AI-assisted decision-making. For each case, the system shows the video of a patient along with an [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Performance of AI-assisted decision-making on [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

97 extracted references · 59 canonical work pages

  1. [1]

    Hervé Abdi and Lynne J Williams. 2010. Principal component analysis. Wiley interdisciplinary reviews: computational statistics 2, 4 (2010), 433–459

    2010

  2. [2]

    Ashraf Abdul, Jo Vermeulen, Danding Wang, Brian Y Lim, and Mohan Kankan- halli. 2018. Trends and trajectories for explainable, accountable and intelligible systems: An hci research agenda. In Proceedings of the 2018 CHI conference on human factors in computing systems . 1–18

    2018

  3. [3]

    Saleema Amershi, Dan Weld, Mihaela Vorvoreanu, Adam Fourney, Besmira Nushi, Penny Collisson, Jina Suh, Shamsi Iqbal, Paul N Bennett, Kori Inkpen, et al. 2019. Guidelines for human-AI interaction. In Proceedings of the 2019 chi conference on human factors in computing systems . 1–13

    2019

  4. [4]

    Theo Araujo, Natali Helberger, Sanne Kruikemeier, and Claes H De Vreese

  5. [5]

    Vijay Arya, Rachel KE Bellamy, Pin-Yu Chen, Amit Dhurandhar, Michael Hind, Samuel C Hoffman, Stephanie Houde, Q Vera Liao, Ronny Luss, Aleksandra Mojsilović, et al. 2019. One explanation does not fit all: A toolkit and taxonomy of ai explainability techniques. arXiv preprint arXiv:1909.03012 (2019)

    arXiv 2019

  6. [6]

    Gagan Bansal, Tongshuang Wu, Joyce Zhou, Raymond Fok, Besmira Nushi, Ece Kamar, Marco Tulio Ribeiro, and Daniel Weld. 2021. Does the whole exceed its parts? the effect of ai explanations on complementary team performance. In Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems . 1–16

    2021

  7. [7]

    Etienne Becht, Leland McInnes, John Healy, Charles-Antoine Dutertre, Im- manuel WH Kwok, Lai Guan Ng, Florent Ginhoux, and Evan W Newell. 2019. Dimensionality reduction for visualizing single-cell data using UMAP. Nature biotechnology 37, 1 (2019), 38–44

    2019

  8. [8]

    Emma Beede, Elizabeth Baylor, Fred Hersch, Anna Iurchenko, Lauren Wilcox, Paisan Ruamviboonsuk, and Laura M Vardoulakis. 2020. A human-centered evaluation of a deep learning system deployed in clinics for the detection of diabetic retinopathy. In Proceedings of the 2020 CHI conference on human factors in computing systems. 1–12

    2020

Show all 97 references
  1. [9]

    Angie Boggust, Brandon Carter, and Arvind Satyanarayan. 2022. Embedding comparator: Visualizing differences in global structure and local neighborhoods via small multiples. In 27th international conference on intelligent user interfaces . 746–766

    2022

  2. [10]

    Zana Buçinca, Maja Barbara Malaya, and Krzysztof Z Gajos. 2021. To trust or to think: cognitive forcing functions can reduce overreliance on AI in AI-assisted decision-making. Proceedings of the ACM on Human-computer Interaction 5, CSCW1 (2021), 1–21

    2021

  3. [11]

    Adrian Bussone, Simone Stumpf, and Dympna O’Sullivan. 2015. The role of explanations on trust and reliance in clinical decision support systems. In 2015 international conference on healthcare informatics . IEEE, 160–169

    2015

  4. [12]

    Carrie J Cai, Jonas Jongejan, and Jess Holbrook. 2019. The effects of example- based explanations in a machine learning interface. In Proceedings of the 24th international conference on intelligent user interfaces . 258–262

    2019

  5. [13]

    Carrie J Cai, Emily Reif, Narayan Hegde, Jason Hipp, Been Kim, Daniel Smilkov, Martin Wattenberg, Fernanda Viegas, Greg S Corrado, Martin C Stumpe, et al

  6. [15]

    Carrie J Cai, Samantha Winter, David Steiner, Lauren Wilcox, and Michael Terry

  7. [16]

    João Caldeira and Brian Nord. 2020. Deeply Uncertain: Comparing Methods of Uncertainty Quantification in Deep Learning Algorithms. arXiv preprint arXiv:2004.10710 (2020)

    2020 arXiv

  8. [17]

    Hello AI

    " Hello AI": uncovering the onboarding needs of medical practitioners for human-AI collaborative decision-making. Proceedings of the ACM on Human- computer Interaction 3, CSCW (2019), 1–24

    2019

  9. [18]

    Valerie Chen, Q Vera Liao, Jennifer Wortman Vaughan, and Gagan Bansal. 2023. Understanding the role of human intuition on reliance in human-AI decision- making with explanations.Proceedings of the ACM on Human-computer Interaction 7, CSCW2 (2023), 1–32

    2023

  10. [19]

    Hao-Fei Cheng, Ruotong Wang, Zheng Zhang, Fiona O’Connell, Terrance Gray, F Maxwell Harper, and Haiyi Zhu. 2019. Explaining decision-making algorithms through UI: Strategies to help non-expert stakeholders. In Proceedings of the 2019 chi conference on human factors in computin...

    2019

  11. [20]

    Charles Corbière, Nicolas Thome, Avner Bar-Hen, Matthieu Cord, and Patrick Pérez. 2019. Addressing failure prediction by learning model confidence.Advances in Neural Information Processing Systems 32 (2019)

    2019

  12. [21]

    Rich Caruana, Yin Lou, Johannes Gehrke, Paul Koch, Marc Sturm, and Noemie Elhadad. 2015. Intelligible models for healthcare: Predicting pneumonia risk and hospital 30-day readmission. InProceedings of the 21th ACM SIGKDD international conference on knowledge discovery and data...

    2015

  13. [22]

    Maria De-Arteaga, Riccardo Fogliato, and Alexandra Chouldechova. 2020. A case for humans-in-the-loop: Decisions in the presence of erroneous algorithmic scores. In Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems. 1–12

    2020

  14. [23]

    Andre Esteva, Brett Kuprel, Roberto A Novoa, Justin Ko, Susan M Swetter, He- len M Blau, and Sebastian Thrun. 2017. Dermatologist-level classification of skin cancer with deep neural networks. nature 542, 7639 (2017), 115–118

    2017

  15. [24]

    Luciano Floridi. 2019. Establishing the rules for building trustworthy AI. Nature Machine Intelligence 1, 6 (2019), 261–262

    2019

  16. [25]

    Jack Cuzick. 1985. A Wilcoxon-type test for trend. Statistics in medicine 4, 1 (1985), 87–90

    1985

  17. [26]

    Felix Gille, Anna Jobin, and Marcello Ienca. 2020. What we talk about when we talk about trust: Theory of trust for AI in healthcare. Intelligence-Based Medicine 1 (2020), 100001

    2020

  18. [27]

    Leilani H Gilpin, David Bau, Ben Z Yuan, Ayesha Bajwa, Michael Specter, and Lalana Kagal. 2018. Explaining explanations: An overview of interpretability of machine learning. In 2018 IEEE 5th International Conference on data science and advanced analytics (DSAA). IEEE, 80–89

    2018

  19. [28]

    David J Gladstone, Cynthia J Danells, and Sandra E Black. 2002. The Fugl-Meyer assessment of motor recovery after stroke: a critical review of its measurement properties. Neurorehabilitation and neural repair 16, 3 (2002), 232–240

    2002

  20. [29]

    Yarin Gal and Zoubin Ghahramani. 2016. A theoretically grounded application of dropout in recurrent neural networks. Advances in neural information processing systems 29 (2016)

    2016

  21. [30]

    Patrick Hemmer, Monika Westphal, Max Schemmer, Sebastian Vetter, Michael Vössing, and Gerhard Satzger. 2023. Human-AI collaboration: the effect of AI delegation on human task performance and task satisfaction. In Proceedings of the 28th International Conference on Intelligent ...

    2023

  22. [31]

    Dan Hendrycks and Kevin Gimpel. 2016. A baseline for detecting misclassified and out-of-distribution examples in neural networks. arXiv preprint arXiv:1610.02136 (2016)

    2016 arXiv

  23. [32]

    Kenneth Holstein, Maria De-Arteaga, Lakshmi Tumati, and Yanghuidi Cheng

  24. [33]

    O’Reilly Media, Inc

    Miguel Grinberg. 2018. Flask web development: developing web applications with python. " O’Reilly Media, Inc. "

    2018

  25. [34]

    Min Hun Lee, Daniel P Siewiorek, Asim Smailagic, Alexandre Bernardino, and Sergi Bermudez i Badia. 2023. Design, development, and evaluation of an inter- active personalized social robot to monitor and coach post-stroke rehabilitation exercises. User Modeling and User-Adapted ...

    2023

  26. [35]

    Alon Jacovi, Ana Marasović, Tim Miller, and Yoav Goldberg. 2021. 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. 624–635

    2021

  27. [36]

    Heinrich Jiang, Been Kim, Melody Guan, and Maya Gupta. 2018. To trust or not to trust a classifier. Advances in neural information processing systems 31 (2018)

    2018

  28. [37]

    Harmanpreet Kaur, Harsha Nori, Samuel Jenkins, Rich Caruana, Hanna Wal- lach, and Jennifer Wortman Vaughan. 2020. Interpreting interpretability: under- standing data scientists’ use of interpretability tools for machine learning. In Proceedings of the 2020 CHI conference on hu...

    2020

  29. [38]

    Jessica Hullman, Xiaoli Qiao, Michael Correll, Alex Kale, and Matthew Kay

  30. [39]

    Alex Kendall and Yarin Gal. 2017. What uncertainties do we need in bayesian deep learning for computer vision? Advances in neural information processing FAccT ’25, June 23–26, 2025, Athens, Greece Lee and Tok systems 30 (2017)

    2017

  31. [40]

    Saif Khairat, David Marc, William Crosby, Ali Al Sanousi, et al. 2018. Reasons for physicians not adopting clinical decision support systems: critical analysis. JMIR medical informatics 6, 2 (2018), e8912

    2018

  32. [41]

    Dajung Kim, Niko Vegt, Valentijn Visch, and Marina Bos-De Vos. 2024. How Much Decision Power Should (A) I Have?: Investigating Patients’ Preferences Towards AI Autonomy in Healthcare Decision Making. In Proceedings of the CHI Conference on Human Factors in Computing Systems . 1–17

    2024

  33. [42]

    Ajay Kohli and Saurabh Jha. 2018. Why CAD failed in mammography. Journal of the American College of Radiology 15, 3 (2018), 535–537

    2018

  34. [43]

    Tzu-Sheng Kuo, Hong Shen, Jisoo Geum, Nev Jones, Jason I Hong, Haiyi Zhu, and Kenneth Holstein. 2023. Understanding Frontline Workers’ and Unhoused Individuals’ Perspectives on AI Used in Homeless Services. In Proceedings of the 2023 CHI Conference on Human Factors in Computin...

    2023

  35. [44]

    Anna Kawakami, Venkatesh Sivaraman, Hao-Fei Cheng, Logan Stapleton, Yanghuidi Cheng, Diana Qing, Adam Perer, Zhiwei Steven Wu, Haiyi Zhu, and Kenneth Holstein. 2022. Improving human-AI partnerships in child welfare: understanding worker practices, challenges, and desires for a...

    2022

  36. [45]

    Vivian Lai and Chenhao Tan. 2019. On human predictions with explanations and predictions of machine learning models: A case study on deception detection. In Proceedings of the conference on fairness, accountability, and transparency . 29–38

    2019

  37. [46]

    Himabindu Lakkaraju, Julius Adebayo, and Sameer Singh. 2020. Explaining machine learning predictions: State-of-the-art, challenges, and opportunities. NeurIPS Tutorial (2020)

    2020

  38. [47]

    Min Hun Lee and Chong Jun Chew. 2023. Understanding the Effect of Counter- factual Explanations on Trust and Reliance on AI for Human-AI Collaborative Clinical Decision Making. Proceedings of the ACM on Human-Computer Interaction 7, CSCW2 (2023), 1–22

    2023

  39. [48]

    Min Hun Lee, Silvana Xin Yi Choo, Shamala D Thilarajah, et al. 2024. Improving Health Professionals’ Onboarding with AI and XAI for Trustworthy Human-AI Collaborative Decision Making. arXiv preprint arXiv:2405.16424 (2024)

    2024 arXiv

  40. [49]

    Min Hun Lee, Renee Bao Xuan Ng, Silvana Xinyi Choo, and Shamala Thilarajah

  41. [50]

    Vivian Lai, Chacha Chen, Q Vera Liao, Alison Smith-Renner, and Chenhao Tan

  42. [51]

    arXiv preprint arXiv:2112.11471 (2021)

    Towards a science of human-ai decision making: a survey of empirical studies. arXiv preprint arXiv:2112.11471 (2021)

    2021 arXiv

  43. [52]

    Min Hun Lee, Daniel P Siewiorek, Asim Smailagic, Alexandre Bernardino, and Sergi Bermúdez i Badia. 2021. A human-ai collaborative approach for clinical decision making on rehabilitation assessment. In Proceedings of the 2021 CHI conference on human factors in computing systems . 1–14

    2021

  44. [53]

    Weixin Liang, Girmaw Abebe Tadesse, Daniel Ho, L Fei-Fei, Matei Zaharia, Ce Zhang, and James Zou. 2022. Advances, challenges and opportunities in creating data for trustworthy AI. Nature Machine Intelligence 4, 8 (2022), 669–677

    2022

  45. [54]

    Gabriel Lima, Nina Grgić-Hlača, and Meeyoung Cha. 2021. Human perceptions on moral responsibility of AI: A case study in AI-assisted bail decision-making. In Proceedings of the 2021 CHI conference on human factors in computing systems . 1–17

    2021

  46. [55]

    Alex John London. 2019. Artificial intelligence and black-box medical decisions: accuracy versus explainability. Hastings Center Report 49, 1 (2019), 15–21

    2019

  47. [56]

    Scott M Lundberg and Su-In Lee. 2017. A Unified Approach to Interpreting Model Predictions. In Advances in Neural Information Processing Systems 30 , I. Guyon, U. V. Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett (Eds.). Curran Associates, Inc., 476...

    2017

  48. [57]

    Shuai Ma, Ying Lei, Xinru Wang, Chengbo Zheng, Chuhan Shi, Ming Yin, and Xiaojuan Ma. 2023. Who should i trust: Ai or myself? leveraging human and ai correctness likelihood to promote appropriate trust in ai-assisted decision- making. In Proceedings of the 2023 CHI Conference ...

    2023

  49. [58]

    Min Hun Lee, Daniel P Siewiorek, Asim Smailagic, Alexandre Bernardino, and Sergi Bermúdez i Badia. 2019. Learning to assess the quality of stroke rehabilita- tion exercises. In Proceedings of the 24th International Conference on Intelligent User Interfaces. 218–228

    2019

  50. [59]

    Min Hun Lee, Daniel P Siewiorek, Asim Smailagic, Alexandre Bernardino, and Sergi Bermúdez i Badia. 2020. Co-design and evaluation of an intelligent decision support system for stroke rehabilitation assessment. Proceedings of the ACM on Human-Computer Interaction 4, CSCW2 (2020), 1–27

    2020

  51. [60]

    T Nathan Mundhenk, Barry Y Chen, and Gerald Friedland. 2019. Efficient saliency maps for explainable AI. arXiv preprint arXiv:1911.11293 (2019)

    2019 arXiv

  52. [61]

    Ju Gang Nam, Sunggyun Park, Eui Jin Hwang, Jong Hyuk Lee, Kwang-Nam Jin, Kun Young Lim, Thienkai Huy Vu, Jae Ho Sohn, Sangheum Hwang, Jin Mo Goo, et al. 2019. Development and validation of deep learning–based automatic detection algorithm for malignant pulmonary nodules on che...

    2019

  53. [62]

    Nicolas Papernot and Patrick McDaniel. 2018. Deep k-nearest neighbors: Towards confident, interpretable and robust deep learning. arXiv preprint arXiv:1803.04765 (2018)

    2018 arXiv

  54. [63]

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al. 2019. Pytorch: An imperative style, high-performance deep learning library. Advances in neural information processing system...

    2019

  55. [64]

    Snehal Prabhudesai, Leyao Yang, Sumit Asthana, Xun Huan, Q Vera Liao, and Nikola Banovic. 2023. Understanding uncertainty: how lay decision-makers perceive and interpret uncertainty in human-AI decision making. In Proceedings of the 28th international conference on intelligent...

    2023

  56. [65]

    Alun Preece. 2018. Asking ‘Why’in AI: Explainability of intelligent systems– perspectives and challenges. Intelligent Systems in Accounting, Finance and Management 25, 2 (2018), 63–72

    2018

  57. [66]

    Thomas M Maddox, John S Rumsfeld, and Philip RO Payne. 2019. Questions for artificial intelligence in health care. Jama 321, 1 (2019), 31–32

    2019

  58. [67]

    Frank J Massey Jr. 1951. The Kolmogorov-Smirnov test for goodness of fit.Journal of the American statistical Association 46, 253 (1951), 68–78

    1951

  59. [68]

    Cynthia Rudin. 2019. Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead. Nature machine intelligence 1, 5 (2019), 206–215

    2019

  60. [69]

    Sara Salimzadeh, Gaole He, and Ujwal Gadiraju. 2024. Dealing with Uncertainty: Understanding the Impact of Prognostic Versus Diagnostic Tasks on Trust and Reliance in Human-AI Decision Making. In Proceedings of the CHI Conference on Human Factors in Computing Systems . 1–17

    2024

  61. [70]

    Ramandeep Singh, Mannudeep K Kalra, Chayanin Nitiwarangkul, John A Patti, Fatemeh Homayounieh, Atul Padole, Pooja Rao, Preetham Putha, Victorine V Muse, Amita Sharma, et al. 2018. Deep learning in chest radiography: detection of findings and presence of change. PloS one 13, 10...

    2018

  62. [71]

    Ralph C Smith. 2013. Uncertainty quantification: theory, implementation, and applications. SIAM

    2013

  63. [72]

    Logan Stapleton, Min Hun Lee, Diana Qing, Marya Wright, Alexandra Choulde- chova, Ken Holstein, Zhiwei Steven Wu, and Haiyi Zhu. 2022. Imagining new futures beyond predictive systems in child welfare: A qualitative study with impacted stakeholders. In Proceedings of the 2022 A...

    2022

  64. [73]

    Harini Suresh, Steven R Gomez, Kevin K Nam, and Arvind Satyanarayan. 2021. Beyond expertise and roles: A framework to characterize the stakeholders of interpretable machine learning and their needs. In Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems . 1–16

    2021

  65. [74]

    Pranav Rajpurkar, Emma Chen, Oishi Banerjee, and Eric J Topol. 2022. AI in health and medicine. Nature medicine 28, 1 (2022), 31–38

    2022

  66. [75]

    Why should i trust you?

    Marco Tulio Ribeiro, Sameer Singh, and Carlos Guestrin. 2016. " 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 . 1135–1144

    2016

  67. [76]

    Ehsan Toreini, Mhairi Aitken, Kovila Coopamootoo, Karen Elliott, Carlos Gonza- lez Zelaya, and Aad Van Moorsel. 2020. The relationship between trust in AI and trustworthy machine learning technologies. In Proceedings of the 2020 conference on fairness, accountability, and tran...

    2020

  68. [77]

    Peter D Turney and Patrick Pantel. 2010. From frequency to meaning: Vector space models of semantics. Journal of artificial intelligence research 37 (2010), 141–188

    2010

  69. [78]

    Joost Van Amersfoort, Lewis Smith, Yee Whye Teh, and Yarin Gal. 2020. Uncer- tainty estimation using a single deep deterministic neural network. In Interna- tional conference on machine learning . PMLR, 9690–9700

    2020

  70. [79]

    Laurens Van der Maaten and Geoffrey Hinton. 2008. Visualizing data using t-SNE. Journal of machine learning research 9, 11 (2008)

    2008

  71. [80]

    Kush R Vashney. 2022. Trustworthy machine learning. Independently published

    2022

  72. [81]

    Brilliant AI doctor

    Dakuo Wang, Liuping Wang, Zhan Zhang, Ding Wang, Haiyi Zhu, Yvonne Gao, Xiangmin Fan, and Feng Tian. 2021. “Brilliant AI doctor” in rural clinics: Chal- lenges in AI-powered clinical decision support system deployment. InProceedings of the 2021 CHI conference on human factors ...

    2021

  73. [82]

    Reed T Sutton, David Pincock, Daniel C Baumgart, Daniel C Sadowski, Richard N Fedorak, and Karen I Kroeker. 2020. An overview of clinical decision support systems: benefits, risks, and strategies for success.NPJ digital medicine 3, 1 (2020), 17

    2020

  74. [83]

    Eric J Topol. 2019. High-performance medicine: the convergence of human and artificial intelligence. Nature medicine 25, 1 (2019), 44–56

    2019

  75. [84]

    Jeannette M Wing. 2021. Trustworthy ai. Commun. ACM 64, 10 (2021), 64–71

    2021

  76. [85]

    Aleš Završnik. 2020. Criminal justice, artificial intelligence systems, and human rights. In ERA forum, Vol. 20. Springer, 567–583

    2020

  77. [86]

    Dongping Zhang, Angelos Chatzimparmpas, Negar Kamali, and Jessica Hullman

  78. [87]

    Wencan Zhang and Brian Y Lim. 2022. Towards relatable explainable AI with the perceptual process. In Proceedings of the 2022 CHI Conference on Human Factors Towards Uncertainty Aware Task Delegation and Human-AI Collaborative Decision-Making FAccT ’25, June 23–26, 2025, Athens...

    2022

  79. [88]

    Yunfeng Zhang, Q Vera Liao, and Rachel KE Bellamy. 2020. Effect of confidence and explanation on accuracy and trust calibration in AI-assisted decision making. In Proceedings of the 2020 conference on fairness, accountability, and transparency . 295–305. A SYSTEM IMPLEMENTATIO...

    2020

  80. [90]

    Danding Wang, Qian Yang, Ashraf Abdul, and Brian Y Lim. 2019. Designing theory-driven user-centric explainable AI. In Proceedings of the 2019 CHI confer- ence on human factors in computing systems . 1–15

    2019

  81. [91]

    Xinru Wang and Ming Yin. 2021. Are explanations helpful? a comparative study of the effects of explanations in ai-assisted decision-making. In 26th international conference on intelligent user interfaces . 318–328

    2021

  82. [95]

    In Proceedings of the CHI Conference on Human Factors in Computing Systems

    Evaluating the utility of conformal prediction sets for ai-advised image labeling. In Proceedings of the CHI Conference on Human Factors in Computing Systems. 1–19

  83. [98]

    helped me validate the competence of the system

    to process the estimated joint positions of post-stroke sur- vivors’ exercises and extract various kinematic features. For the kinematic features of the ‘Range of Motion’ (ROM), we computed joint angles, such as elbow flexion, shoulder flexion, and elbow extension, and normali...

    2025

  84. [2018]

    IEEE transactions on visualization and computer graphics 25, 1 (2018), 903–913

    In pursuit of error: A survey of uncertainty visualization evaluation. IEEE transactions on visualization and computer graphics 25, 1 (2018), 903–913

    2018

  85. [2019]

    In Proceedings of the 2019 chi conference on human factors in computing systems

    Human-centered tools for coping with imperfect algorithms during medical decision-making. In Proceedings of the 2019 chi conference on human factors in computing systems. 1–14

    2019

  86. [2020]

    AI & society 35 (2020), 611–623

    In AI we trust? Perceptions about automated decision-making by artificial intelligence. AI & society 35 (2020), 611–623

    2020

  87. [2021]

    In Extended Abstracts of the 2021 CHI Conference on Human Factors in Computing Systems

    Onboarding Materials as Cross-functional Boundary Objects for Developing AI Assistants. In Extended Abstracts of the 2021 CHI Conference on Human Factors in Computing Systems. 1–7

    2021

  88. [2023]

    Proceedings of the ACM on Human-Computer Interaction 7, CSCW1 (2023), 1–20

    Toward supporting perceptual complementarity in human-AI collaboration via reflection on unobservables. Proceedings of the ACM on Human-Computer Interaction 7, CSCW1 (2023), 1–20

    2023

  89. [2024]

    Interactive Example-based Explanations to Improve Health Professionals’ Onboarding with AI for Human-AI Collaborative Decision Making.arXiv preprint arXiv:2409.15814 (2024)

    2024 arXiv

Pith tools

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