Exploring the effects of human-centered AI explanations on trust and reliance
Open accessThe DOI matches its Crossref record; 99 of the 99 references it cites were found in a registry or an index; 23 of the 59 sources its claims cite were checked for support.
Abstract
Transparency is widely regarded as crucial for the responsible real-world deployment of artificial intelligence (AI) and is considered an essential prerequisite to establishing trust in AI. There are several approaches to enabling transparency, with one promising attempt being human-centered explanations. However, there is little research into the effectiveness of human-centered explanations on end-users' trust. What complicates the comparison of existing empirical work is that trust is measured in different ways. Some researchers measure subjective trust using questionnaires, while others measure objective trust-related behavior such as reliance. To bridge these gaps, we investigated the effects of two promising human-centered post-hoc explanations, feature importance and counterfactuals , on trust and reliance. We compared these two explanations with a control condition in a decision-making experiment ( N = 380). Results showed that human-centered explanations can significantly increase reliance but the type of decision-making (increasing a price vs. decreasing a price) had an even greater influence. This challenges the presumed importance of transparency over other factors in human decision-making involving AI, such as potential heuristics and biases. We conclude that trust does not necessarily equate to reliance and emphasize the importance of appropriate, validated, and agreed-upon metrics to design and evaluate human-centered AI.
Overview
CoChat Summary
TL;DR
This online experiment (N = 380) tested feature importance and counterfactual explanations against a control for effects on trust and reliance in an apartment-pricing task. Counterfactual explanations increased reliance only when the AI recommended decreasing the price, while explanations did not affect trust and recommendation direction had a larger effect on reliance.
Key findings
- Human-centered explanations did not significantly affect trust: TPA overall H(2) = 1.54, p = 0.46; trust H(2) = 5.06, p = 0.08; distrust H(2) = 2.03, p = 0.36. H4–H6 were rejected.
- Reliance was significantly higher for counterfactual explanations vs. control only when the AI recommended decreasing the initial price: β = 0.04, 95% CI [0.01, 0.08], t(378) = 2.31, p = 0.02, an approximately 4% increase in WOA. Feature importance increased WOA by 2% vs. control but was not significant (p = 0.27).
- Recommendation type had a larger effect than explanations: increase vs. decrease recommendations was highly significant for WOA, β = 0.05, 95% CI [0.03, 0.07], t(1,899) = 3.62, p < 0.001, and improved model fit, χ²(1) = 13.05, p < 0.001. Participants relied more on recommendations to decrease prices.
- Overall WOA was positive (M = 0.69, SD = 0.36), and TPA ratings were average overall (M = 5.01, SD = 0.86), with higher trust (M = 4.98, SD = 1.05) than distrust (M = 2.95, SD = 1.62).
Methods
- Online decision-making experiment on Amazon Mechanical Turk; 913 recruited, 798 completed, and 380 remained after exclusions. Sample was predominantly male (61%), mean age 37.03 years (SD = 10.15), and 68% had a higher-education degree.
- Participants estimated subleasing prices for six real Zillow apartments, received an alleged AI price recommendation (random 10–20% deviation from their initial estimate), and then gave a final price. The design was 3 (explanation: feature importance vs. counterfactual vs. control) × 2 (recommendation: increase vs. decrease price), with explanation between-subjects and recommendation within-subjects.
- Reliance was measured by Weight of Advice (WOA); trust was measured by the Trust between People and Automation Scale (TPA, 12 items, 7-point). WOA was analyzed with linear mixed-effects models; TPA was analyzed with Kruskal-Wallis tests after exploratory factor analysis indicated a two-factor trust/distrust structure.
Limitations
- The paper states the study was domain-specific to real estate, used low-stakes MTurk decisions, and did not test untrustworthy recommendations or larger deviations from initial estimates; future work should broaden domains and examine high-stakes or poorer-performing AI.
- The paper states the design could not distinguish dispositional, situational, and learned trust, and trust was measured after task completion, so trial-level effects of bias and heuristics on trust could not be examined.
- Evident from the text: the final sample was predominantly male and highly educated, which may limit generalizability; the task deliberately had no “true price,” and WOA captures reliance behavior rather than trust attitudes.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
21 Valid · 0 Not valid · 19 Not checked · 68 Own results and interpretations
The sources with a verdict support this claim; the rest weren’t checked yet, have no full text we could read, or their verdicts aren’t rated yet.
In the paper
“However, research has shown that subjective trust can be a poor predictor of actual reliance (Dzindolet et al., 2003; Miller et al., 2016; Papenmeier et al., 2022).”
- Also stated in: § 1. Introduction · plain
“research has shown that subjective trust can be a poor predictor of actual reliance (Dzindolet et al., 2003; Miller et al., 2016; Papenmeier et al., 2022).”
What the sources say
- Not checked — no full text of the sourcecites Dzindolet et al., 2003 source
- Not checked — no full text of the sourcecites Miller et al., 2016 source
- Supportedcites Papenmeier et al., 2022 source
Section 4.4 reports a weak correlation between self-reported and behavioural trust (r = 0.2998) and concludes that neither is a good proxy for the other. The discussion characterizes the behavioural measure as a better proxy for reliance than for trust as a whole, supporting the claim that subjective trust can poorly predict actual reliance. (LLM, conf=0.91)
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § 1. Introduction · plain
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“To the best of our knowledge, there is little to no empirical research on the effectiveness of these two humancentered explanations derived from the literature (i.e., Adadi and Berrada, 2018) using Miller's criteria in fostering end-users' trust in AI.”
What the sources say
- Not checked — no full text of the sourcecites Adadi and Berrada, 2018 source
- Partly supportedcites Miller, 2019 sourceExperimental — accuracy not yet measured
Section 4.6.6 identifies criteria for evaluating explanations—simplicity, generality, and coherence—and discusses their potential role in generating trust. It does not establish the claimed lack of empirical research on feature-importance and counterfactual explanations evaluated using those criteria for end-users’ trust. (LLM, conf=0.84)
hedged- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Every source the paper cites for this claim was checked and supports it.
In the paper
“This deliberate choice by the participants to either rely on the AI recommendation or not makes our experiment an actual decision-making task rather than a proxy task (Buçinca et al., 2020).”
What the sources say
- Supportedcites Buçinca et al., 2020 source
The Discussion (p. 9) distinguishes the proxy task from the actual decision-making task: in the latter, users aim to make accurate nutritional decisions and choose whether and how deeply to engage with AI recommendations. It also reports that participants sometimes ignored incorrect recommendations, supporting the cited distinction. (LLM, conf=0.91)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The sources with a verdict support this claim; the rest weren’t checked yet, have no full text we could read, or their verdicts aren’t rated yet.
In the paper
“Even if an AI system is trusted, reliance must not necessarily follow (Kirlik, 1993; Körber, 2018)”
What the sources say
- Not checked — no full text of the sourcecites Kirlik, 1993 source
- Supportedcites Körber, 2018 source
In the theoretical discussion of trust and reliance, the paper states that even when trust is high and an automated system is perceived as capable, reliance does not necessarily follow. This directly supports the citing sentence’s paraphrase. (LLM, conf=0.95)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Every source the paper cites for this claim was checked and supports it.
In the paper
“Whether trust translates into reliance is more nuanced than often assumed and depends on an interaction between the operator, the automation, and the situation (Körber, 2018).”
What the sources say
- Supportedcites Körber, 2018 source
Körber's theoretical section states the claim almost verbatim: 'Whether trust translates into reliance behavior depends on a dynamic interaction of operator, automation, situational factors, and interface (Lee & See, 2004)', and surrounds it with the nuance the citing sentence asserts — trust and reliance are 'closely related' yet 'distinct constructs', attitude/intention/behaviour stand in a 'probabilistic' rather than deterministic relationship (Ajzen & Fishbein, 1980), and 'even if trust is at a high level ... reliance does not necessarily follow' (Kirlik, 1993), with environmental and cognitive constraints determining whether intention becomes actual reliance. Körber adopts and relies on this point to build his own model and justify a self-report questionnaire, so citing him for it is appropriate even though the four-factor formulation is credited onward to Lee & See (2004); the citing paraphrase drops 'interface' from the list of interacting factors but asserts nothing the source contradicts. (LLM, conf=0.82)
plainstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
53 more claims · 1 with their cited sources checked
Create a free account to see all 108 claims and every source verdict.
- Probably not a substantial claim. 29
- It may state two separate claims as one. 26
- The quote only partly states this claim. 1
Statistics recomputed
Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.
- Consistentt(378) = 0.64reported p = .53 · recomputed p = .523
“t (378) = 0.64, p = .53”
- Consistentt(378) = 1.33reported p = .19 · recomputed p = .184
“t (378) = 1.33, p = .19”
- Consistentt(378) = 0.63reported p = .53 · recomputed p = .529
“t (378) = 0.63, p = .53”
- Consistentχ²(1) = 13.05reported p < .001 · recomputed p < .001
“χ 2 (1) = 13.05, p < .001”
- Consistentt(378) = 2.31reported p = .02 · recomputed p = .021
“t (378) = 2.31, p = .02”
1 more recomputed test
Create a free account to see every test the Stats check recomputed.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ 6. Discussion
“A WOA of 0.70 signifies that participants adopted 70% of the AI recommendations when updating their prior beliefs to form their final estimate.”
§ 4.1. Measures
“WOA is equal to 1 if the participant's final prediction matches the AI recommendation and equal to 0.5 if they average their initial prediction with the AI recommendation.”
Citations
Cites
Fitting Linear Mixed-Effects Models Using lme4
Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015) · Journal of Statistical Software · doi:10.18637/jss.v067.i01 · Cited by 86,835
LeCun, Y., Bengio, Y., & Hinton, G. (2015) · Nature · doi:10.1038/nature14539 · Cited by 84,126
Perceived usefulness, perceived ease of use, and user acceptance of information technology
Davis, F. D. (1989) · MIS Quarterly · doi:10.2307/249008 · Cited by 68,483
Tversky, A., & Kahneman, D. (1974) · Science · doi:10.1126/science.185.4157.1124 · Cited by 28,810
(1980) · Medical Entomology and Zoology · Cited by 25,810
The Framing of Decisions and the Psychology of Choice
Tversky, A., & Kahneman, D. (1981) · Science · doi:10.1126/science.7455683 · Cited by 18,497
Use of Ranks in One-Criterion Variance Analysis
Kruskal, W. H., & Wallis, W. A. (1952) · Journal of the American Statistical Association · doi:10.1080/01621459.1952.10483441 · Cited by 13,912
Arrieta, A. B., Díaz-Rodríguez, N., Del Ser, J., Bennetot, A., Tabik, S., Barbado, A., García, S., Gil-López, S., Molina, D., Benjamins, R., et al. (2019) · arXiv · doi:10.48550/arxiv.1910.10045 · Cited by 11,321
An integrative model of organizational trust
Mayer, R. C., Davis, J. H., & Schoorman, F. D. (1995) · The Academy of Management Review · doi:10.2307/258792 · Cited by 10,736
Barredo Arrieta, A., Díaz-Rodríguez, N., Del Ser, J., Bennetot, A., Tabik, S., Barbado, A., Garcia, S., Gil-Lopez, S., Molina, D., Benjamins, R., et al. (2020) · Information Fusion · doi:10.1016/j.inffus.2019.12.012 · Cited by 10,423
Research works printed but not linked: 5
- 2022, “ACM conference on fairness, accountability, and transparency 2022 (ACM FAccT 2022) call for papers,”
- Biran, 2017, “Explanation and justification in machine learning: a survey,”, 8
- 2016, Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the Protection of Natural Persons With Regard to the Processing of Personal Data and on the Free Movement of Such Data, and Repealing Directive 95/46/EC (General Data Protection Regulation)
- Hartmann, 2020, Vertrauen - Die unsichtbare Macht
- Solso, 2005, Cognitive Psychology
Cited by
Patient Consent and The Right to Notice and Explanation of AI Systems Used in Health Care
Hurley, M. E., Lang, B. H., Kostick-Quenet, K. M., Smith, J. N., & Blumenthal-Barby, J. (2025) · The American Journal of Bioethics · doi:10.1080/15265161.2024.2399828 · Cited by 63
“Patient trust in an AI system may also differ due to its particular application and context, regardless of the quality of the explanation given (Scharowski et al. 2023).”
AIdeation: Designing a Human-AI Collaborative Ideation System for Concept Designers
Wang, W.-F., Lu, C.-T., Ponsa i Campanyà, N., Chen, B.-Y., & Chen, M. Y. (2025) · Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3714148 · Cited by 50
“Through effective communication, iterative feedback, and user control, these systems create dynamic collaborations to enhance workflows [45, 91, 105].”
Baron, S. (2025) · Philosophy & Technology · doi:10.1007/s13347-024-00837-6 · Cited by 29
“As in the more general case of reliance, explainability can promote reliance on AI (see, e.g., Scharowski et al., 2023; Schemmer et al., 2023; Schoeffer et al., 2024).”
Torkamaan, H., Steinert, S., Pera, M. S., Kudina, O., Freire, S. K., Verma, H., Kelly, S., Sekwenz, M.-T., Yang, J., van Nunen, K., et al. (2024) · Behaviour and Information Technology · doi:10.1080/0144929x.2024.2431068 · Cited by 43
Eckhardt, S., Kühl, N., Dolata, M., & Schwabe, G. (2025) · ACM Computing Surveys · doi:10.1145/3776528 · Cited by 44
A System to Ensure Information Trustworthiness in Artificial Intelligence Enhanced Higher Education
Khan, U. A., Kauttonen, J., Aunimo, L., & Alamäki, A. V. (2024) · Journal of Information Technology Education Research · doi:10.28945/5295 · Cited by 13
“The integration of transparency and explainability features into its operation ensures users have clear insight into its decision-making processes, fostering trust and promoting wider adoption, the issues mentioned by Scharowski et al. (2023) and Szczepański et al. (2021).”
BackgroundIncreasing transparency of computer-aided detection impairs decision-making in visual search
Kunar, M. A., Watson, D., & Montana, G. (2024) · Psychonomic Bulletin & Review · doi:10.3758/s13423-024-02601-5 · Cited by 10
“Given that explainable AI (XAI) is a complex and difficult field of research (Biran & Cotton, 2017; Scharowski et al., 2023), the above results question whether the need to produce transparent AI is always necessary.”
The Influence of Curiosity Traits and On-Demand Explanations in AI-Assisted Decision-Making
Cau, F. M., & Spano, L. D. (2025) · ACM International Conference on Intelligent User Interfaces (IUI) · doi:10.1145/3708359.3712165 · Cited by 15
“Explainable AI (XAI) methods have emerged to address this by providing users with insights into how AI decisions are made, aiming to improve users’ accuracy [6, 13, 45] and foster appropriate reliance on AI [86, 96].”
Cultural Tightness Increases Individuals’ Preference for Control in Human–AI Collaboration
Sui, J., Zhang, Y., & Zhao, Y. (2025) · International Journal of Human-Computer Interaction · doi:10.1080/10447318.2025.2530079 · Cited by 6
“…systems that adapt to users ’ needs (Hong et al., 2009; Jiang et al., 2023), as well as improvements in relationship dynamics, such as increasing transparency (V € ossing et al., 2022) and explainability (Scharowski et al., 2023), and clearly defined collaborative roles (Seeber et al., 2020).”
InfluentialRaees, M., Khan, V.-J., Lykourentzou, I., & Papangelis, K. (2026) · Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3772318.3791467 · Cited by 5
“Initial works [65] mostly studied users’ trust in automation; however, trust measures do not necessarily enhance reliance [131], which is gaining attention in recent work [81, 130].”
Diverse Counterfactual Explanations (DiCE) Role in Improving Sales and e-Commerce Strategies
Oprea, S., & Bârã, A. (2025) · Journal of theoretical and applied electronic commerce research · doi:10.3390/jtaer20020096 · Cited by 9
Lv, Z., Tang, C., Zheng, Y., & Yang, X. (2025) · International Journal of Human-Computer Interaction · doi:10.1080/10447318.2025.2560518 · Cited by 12
“…highlight that dynamic interpretability is central to proactive decision support: by annotating decision rationales in real time, AI Agents enable users to rapidly comprehend complex contexts, thereby enhancing information processing efficiency and decision confidence (Scharowski et al., 2023).”
Kuhl, U., & Bush, A. (2025) · Communications in computer and information science · doi:10.1007/978-3-032-08333-3_12 · Cited by 12
“A number of user-based-evaluations fail to demonstrate significant improvements in trust driven by CEs [41, 34].”
From Trust to Appropriate Reliance: Measurement Constructs in Human-AI Decision-Making
Raees, M., & Papangelis, K. (2026) · arXiv · doi:10.48550/arxiv.2604.23896 · Cited by 5
“AI-assisted decision-making research operationalizes trust and reliance measures, where trust generally reflects the users’ subjective perception (i.e., affective or emotional attitude) towards others or systems [45, 54], while reliance pertains to their (observable) objective behavior [25, 36, 70].”
Exploring the impact of explainable AI and cognitive capabilities on users’ decisions
Cau, F. M., & Spano, L. D. (2025) · User Modeling and User-Adapted Interaction · doi:10.1007/s11257-025-09438-0 · Cited by 20
“…have shown that explanations may lead to increased user accuracy (Lai and Tan 2019; Buçinca et al. 2020; Bansal et al. 2021; Herm 2023) and appropriate reliance on AI (Wang and Yin 2022; Scharowski et al. 2023; Chen et al. 2023) when compared to AI prediction alone or not showing any assistance.”
Effects of AI explanations on trust and reliance: a study in job shop scheduling
Saßmannshausen, T. M., Burggräf, P., Hassenzahl, M., & Sauer, C. R. (2026) · Ergonomics · doi:10.1080/00140139.2026.2634115 · Cited by 4
Dong, A., Li, P., Chen, Y., Gibson, S., Zhao, L., & He, M. (2026) · Sustainability · doi:10.3390/su18115313 · Cited by 9
Wang, W., Gao, S., Chen, Q., & Ma, S. (2025) · Behavioral Sciences · doi:10.3390/bs15121646 · Cited by 5
“This transparency enhances users’ trust in and willingness to use the information provided (Scharowski et al., 2023).”
How explanatory features of AI and time frame reshape adolescents’ decision-making
Shen, Z., Chen, Y., Zhang, J., & Chen, H. (2026) · Computers & Education · doi:10.1016/j.compedu.2026.105563 · Cited by 4
Marquardt, M., Graf, P., Jansen, E. M., Hillmann, S., & Voigt‐Antons, J. (2024) · TATuP - Zeitschrift für Technikfolgenabschätzung in Theorie und Praxis · doi:10.14512/tatup.33.1.41 · Cited by 4
Ethical AI governance, trust and deepfake regulation in the UAE’s media landscape
Heluey, N. (2025) · Journal of Digital Media & Policy · doi:10.1386/jdmp_00173_1 · Cited by 9
Explainability, Safety Cues, and Trust in GenAI Advisors: A SEM–ANN Hybrid Study
Balaskas, S., Stamatiou, I., & Androulakis, G. (2025) · Future Internet · doi:10.3390/fi17120566 · Cited by 4
“First, research combines the constructs of safety and transparency: “transparency” is operationalized as a general disclosure or explanation cue, and guardrails (visible boundaries, refusals, and policy cues) are excluded or absorbed into the same construct [12].”
InfluentialPersuasive explanations for recommender systems: how explanations can influence users’ choices?
Rahman, S. M. T., Siemon, D., & Ruotsalo, T. (2025) · International Journal of Human-Computer Studies · doi:10.1016/j.ijhcs.2025.103720 · Cited by 6
Automatically Engineering Trusted Software: A Research Roadmap
Brun, Y., Chakraborty, S., Goues, C. L., Păsăreanu, C. S., & Singla, A. K. (2026) · ACM Transactions on Software Engineering and Methodology · doi:10.1145/3779132 · Cited by 5
“Measuring trust presents additional challenges: self-reported measures [49, 70, 176] often disagree with behavioral observations [44, 100, 157].”
Schneider, D., & Weber, K. (2024) · TATuP - Zeitschrift für Technikfolgenabschätzung in Theorie und Praxis · doi:10.14512/tatup.33.1.08 · Cited by 6
“Dieser Befund steht in Einklang mit der Unterscheidung von Vertrauen als Einstellung und ‚reliance‘ als Verhalten „[…] that follows the level of trust“ (Scharowski et al. 2023, S. 4).”
BackgroundA Human-Centric Evaluation of XAI Methods for Sentiment Analysis on Amazon Reviews
Abhiram, Amritha, C. N., Gokul, R., & Keerthimon, P. (2025) · International Conference on Advances in Computing, Communication, Embedded and Secure Systems (ACCESS) · doi:10.1109/access65134.2025.11135705 · Cited by 3
“Scharowski et al. [5] demonstrated that well-designed XAI outputs improve both user reliance and decision-making performance.”
Chu, S. L., Karcz, M., Hashky, A., Rani, N., Chaspari, T., Arthur, W., & Ragan, E. D. (2025) · International Journal of Human-Computer Studies · doi:10.1016/j.ijhcs.2025.103691 · Cited by 2
AGI-Enabled Solutions for Service Explosion in IoX: A Cyber-Physical-Social-Thinking Perspective
Wei, Q., Ning, H., Shi, F., Zhu, T., Ding, J., Derhab, A., & Daneshmand, M. (2025) · ACM Computing Surveys · doi:10.1145/3785669 · Cited by 6
“At the same time, the highly dynamic system states exacerbate the difficulty of fault detection and recovery, presenting a significant challenge for ASI to maintain stable operation in complex environments [127].”
The epistemologies of trust: conflicting worldviews in the “Trustworthy AI” discourse
Broniatowski, D. A., Joubin, A. A., Aaronson, S. A., Yang, Y. T., Sikka, N. K., Abroms, L. C., Zirikly, A., & Szajnfarber, Z. (2026) · AI and Ethics · doi:10.1007/s43681-026-01087-2 · Cited by 1
Using It but Not Trusting It: Technology Acceptance and Generative AI in European Fact-Checking
Dierickx, L., & van Dalen, A. (2026) · Emerging Media · doi:10.1177/27523543261449974 · Cited by 3
“Rather than disclosing every technical detail, explainability aims to enhance interpretability and perceived trust-worthiness by clarifying how results are reached (Scharowski et al., 2023).”
User Reliance on AI Support for Collaborative Partner Selection
Hrkalović, T. M., Dudzik, B., Hao, C., & Willemsen, M. C. (2026) · ACM International Conference on Intelligent User Interfaces (IUI) · doi:10.1145/3742413.3789082
“Explanations are the most studied approach [12, 25, 32, 54].”
Varatharaj, M. (2025) · International Journal of Scientific Research in Computer Science Engineering and Information Technology · doi:10.32628/cseit25111234
Transition to Organisation 5.0 - Barriers and Enablers of AI Adoption in Accounting and Finance
Kortesalmi, H., Aunimo, L., Rauhala, M., & Schlögl, S. (2025) · IFIP advances in information and communication technology · doi:10.1007/978-3-032-05681-8_12
Silvy, Suryanto, E. S., & Yustian, O. R. (2025) · International Conference on Information Management and Technology (ICIMTech) · doi:10.1109/icimtech67074.2025.11265294
“Adaptive AI systems in smartphones satisfy by personalizing user experiences, creating trust and engagement [26].”
Bias Correction and Explainability Framework for Large Language Models: A Knowledge-Driven Approach
Yang, X., Li, Q., Qian, C., Wang, H., Wu, Y., & Wang, W. (2026) · Big Data and Cognitive Computing · doi:10.3390/bdcc10020058 · Cited by 1
Zhang, D., Luo, S., Liu, Y., Zhang, X., Hu, Y., Shen, J., Yi, P., Zhao, K., & Liu, W. (2026) · Humanities and Social Sciences Communications · doi:10.1057/s41599-026-06738-7 · Cited by 5
Lin, J., & Powell, N. L. (2026) · International Journal of Human-Computer Interaction · doi:10.1080/10447318.2026.2636791 · Cited by 1
Data-Informed, Not Data-Driven: Leadership Practices for Responsible AI Integration in K-12 Schools
Ragab, K. (2026) · Advances in computational intelligence and robotics book series · doi:10.4018/979-8-3373-4908-4.ch011
Sun, X., Wei, Z., Ding, X., Guo, P., Xiao, H., Duan, Z., & Siek, H. L. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6539729
From Workslop to Work: How AI Explanations Reconfigure Human Judgment in Creative Tasks
Lim, S., & Jun, S. (2026) · CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI EA) · doi:10.1145/3772363.3798328 · Cited by 1
“In conditions with AI feedback (C2 and C3), trust and reliance were assessed using adapted XAI measures that capture tendencies toward deference and verification [20].”
Streamlining risk assessment for predictive maintenance: Can AI be trusted as an assistant?
Shobanke, M., Roberts, C., & Shittu, E. (2026) · Advanced Engineering Informatics · doi:10.1016/j.aei.2026.104803
Selvam, M. (2026) · IEEE International Conference on Secure IoT, Assured and Trusted Computing (SATC) · doi:10.1109/satc69565.2026.11542273 · Cited by 1
“Frontiers’ research on human-centered AI explanations reveals that post-hoc explanations (e.g., feature importance, counterfactual) can increase reliance but not reported trust, and effects vary by task type [18].”
A capability centric framework for delivering human centred AI services in healthcare
Atariani, M. (2026) · Discover Artificial Intelligence · doi:10.1007/s44163-026-01572-x
“• Implementing controls to counter the influence of potentially misleading or detrimental outputs [5, 7, 8].”
InfluentialWhen Explanations Deceive: Understanding Unintentional and Intentional Deception in XAI
Tekkesinoglu, S. (2026) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3805689.3812351
Möhle, M. H., Zall, M., Thorwart-Gumpert, T., Elbert, N., Thiesse, F., & Flath, C. M. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-30860-3_30 · Cited by 1
Cui, W., Yang, C.-L., & Narumi, T. (2026) · ACM Conference on Conversational User Interfaces (CUI) · doi:10.1145/3816046.3816274 · Cited by 1
Truth According to Machine: AI’s Influence on Human Credibility AssessmentPreprint
Glowacki, Z. R., Raffloer, G., & Frank, M. G. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7202488
A Systematic Validation of the Jian et al. (2000) “Trust in Automated Systems Scale”
Pithayarungsarit, P., Newberry, A., Lee, J. R., Roesler, E., & Gutzwiller, R. S. (2026) · Proceedings of the Human Factors and Ergonomics Society Annual Meeting · doi:10.1177/10711813261475227
Andoniou, C. (2026) · International Conference on Intelligent and Sustainable AI Systems (ICOSAAS) · doi:10.1109/icosaas68663.2026.11649248
“Recent research found that exposing the reasoning process of a model can increase user trust and reliance [5], [7].”
Tran, H. M. N., & Que, A. P. N. (2026) · International Conference on Emerging Trends in Information, Communication & Systems (ICETICS) · doi:10.1109/icetics69505.2026.11651282
“There is associated evidence that human centered explanations are capable of raising reliance, and trust and reliance are not always positively correlated, which implies that outcomes oriented towards adoption cannot be obtained by inference based on trust only [5].”
SSS Framework: A Think-First, AI-Second Model for AI-Assisted Writing.
Giray, L. (2026) · Annals of Biomedical Engineering · doi:10.1007/s10439-026-04358-5
Taylor, A. D., Goodloe, R., Rosenman, M. B., & Fulton, C. R. (2026) · Applied Clinical Informatics · doi:10.1055/a-2927-0698
Hein, I., Ullrich, D., Brunner, C., & Diefenbach, S. (2026) · Frontiers in Computer Science · doi:10.3389/fcomp.2026.1869586
The Role of Multilevel Trust in eXplainable Artificial Intelligence for Higher Education
Khaustova, V. Y., Reshetnyak, O. I., & Shapoval, O. (2026) · Lecture notes in networks and systems · doi:10.1007/978-3-032-27691-9_26
A hybrid fuzzy-LLM decision support system for lifecycle-oriented supply-chain strategy
Molina, A., Méndez, J. I., & Bárcenas-Cortés, A. L. (2026) · Expert Systems with Applications · doi:10.1016/j.eswa.2026.134176
Huali, F. A. N., Wang, Y., Pan, S., & Shao, S. (2026) · Frontiers in Psychology · doi:10.3389/fpsyg.2026.1924924
Metadata
- Authors
- Scharowski, N., Perrig, S. A. C., Svab, M., Opwis, K., & Brühlmann, F.
- Affiliations
- University of Basel (CH)
- Corresponding author
- Scharowski, Nicolas
- Venue
- Frontiers in Computer Science
- Publisher
- Frontiers Media SA
- Published
- 2023-07-17
- Volume
- 5
- ISSN
- 2624-9898 (ISSN-L)
- OpenAlex
- W4384524844
- Cited by
- 57
- References
- 97
- Citations by year
- 2026: 32 · 2025: 19 · 2024: 6
- FWCI
- 5.04
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Frontiers in Computer Science (publishedVersion, cc-by)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- OSF Preprints (OSF Preprints) (submittedVersion)
- Funders
- Universität Basel
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Explainable Artificial Intelligence (XAI)
- Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
- Social Sciences › Human Factors and Ergonomics › Human-Automation Interaction and Safety
- Keywords
- trust in AI, AI reliance, AI transparency, post-hoc explanation, feature importance, counterfactual explanations, subjective trust, heuristics and biases, trust metrics
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-09-30, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-09-30 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.