Jul 2026· Medical decision making· pp.
272989X261467775
· 0 citations· 53 references
Medicine
TL;DR
The findings support the AI aversion hypothesis and suggest that there needs to be more work to identify how best to explain AI-based calculators to foster trust and comfort more specifically.
Abstract
INTRODUCTION
Although medical risk calculators are increasingly being used for risk prediction in various contexts, prior research in this area suggests that people may be averse to recommendations from these tools. Moreover, given the rise of artificial intelligence in health care, aversion toward AI may manifest toward risk calculators that use AI/machine learning models. Recent work has suggested that one potential way to combat AI aversion is through explainable AI (XAI), which can make underlying models more transparent.
Methods
The current study investigated whether these factors would affect public trust, acceptance, and comfort related to recommendations from a risk calculator. Participants were randomized into a 2 (calculator type: statistical vs AI) × 2(model explanation: explained vs not explained) × 2(evaluability: risk reference table vs no table) between-subjects experimental design. They read a hypothetical scenario and received a calculator output with a risk estimate and recommendation before completing measures of trust, comfort, and acceptance.
Results
Analyses revealed main effects of calculator type and explanation: participants who received the AI-based calculator outputs were less likely to trust and be comfortable with the calculator recommendation. Participants who received an explanation of the underlying model were more likely to trust and accept the calculator recommendation. However, there was no effect of XAI on trust, acceptance, and comfort or any other significant interactions.
Conclusion
The current study serves as a starting point for research on trust and acceptance of AI-based risk calculators. Our findings support the AI aversion hypothesis and suggest that there needs to be more work to identify how best to explain AI-based calculators to foster trust and comfort more specifically.
Analysis of consumers' trust in AI-generated recommendations under conditions of AI-assisted decision-making shows that emotional trust may be a mediator in the intention to delegate decision-making to AI agents and proposes strategies to build more transparent and trustworthy AI recommendation systems that can improve the user experience.
Yayi Liu· Frontiers in Humanities and...· 0 citations
While participants rated hedged and unhedged AI as equally trustworthy and likely to be correct, they were significantly less likely to follow hedged advice in a binary choice, and how linguistic markers can be used to calibrate user reliance to model certainty is discussed.
Laura Spillner, Johanna Rockstroh, Nina Wenig et al.· International Conference on...· 0 citations
It is argued that calibration-aligned design (rather than trust maximization alone) should guide the development and assessment of high-stakes AI decision support, because reductions in reported trust do not consistently translate into commensurate changes in reliance behavior.
Algorithm aversion and anti-AI bias describe the human tendency to evaluate decisions made by human agents more favorably than those made by artificial intelligence (AI) algorithms. This study aimed to examine whether algorithm aversion influences students’ evaluations of multiple-choice questions (MCQs) as well as the psychometric quality of items generated by a large language model (LLM). Additionally, the authors explored the possible influence of AI literacy and attitudes toward AI. The research team administered a formative examination consisting of two blocks. Block 1 was labeled “authored by a medical educator,” although 50% of the 20 items were in fact generated by an LLM. Block 2 was labeled “generated by an LLM,” but likewise included 50% human-authored items. The human-authored and AI-generated questions were based on the same 20 learning objectives. Following each item, medical students rated its perceived relevance for exam preparation and its quality of wording. AI literacy and attitudes toward AI were assessed using validated instruments. Participants solved significantly more LLM-generated items (
M
= 77%) correctly than human-authored items (
M
= 48%),
U
= 34,
p
< .001,
r
= 0.71. No significant differences were observed in discriminatory power. Cumulative link mixed models revealed that students rated the relevance of actually AI-generated items more positively. In addition, they exhibited an anti-AI bias in their ratings of item wording quality, as items labeled as LLM-generated were rated significantly worse, irrespective of their actual origin. Although the overall findings statistically suggest an anti-AI bias, their practical implications remain unclear, as relevance and wording quality were rated very positively overall.
Matthias Carl Laupichler, Susanne Eigster, Seifollah Ahmadi et al.· The journal of the Internati...· 0 citations
Introduction Artificial intelligence (AI) is increasingly used in public agencies to route inquiries, screen eligibility, support caseworkers, and automate routine service encounters. Citizen acceptance of these services depends on their links to public authority, accountability, and visible opportunities for human recourse. This study examines a trust-based mechanism connecting institutional trust, risk perception, AI service trust, and behavioral intention in China's digital government context. Methods The study combined an LLM-driven agent simulation involving 936 agents across three independent seeds, a 3 × 3 factorial scenario experiment involving 900 simulated agents, and a human-validation pilot using the same questionnaire and scenario structure. The pilot generated 189 submitted records, of which 182 were retained after attention checking. Results In the synthetic calibration, institutional trust is positively associated with AI service trust (IT → AST β = 0.607) and negatively associated with risk perception (IT → RP β = −0.271); risk perception is negatively associated with AI service trust (RP → AST β = −0.459); and AI service trust is positively associated with behavioral intention (AST → BI β = 0.424). The same directional pattern appears in the human-validation pilot (IT → AST β = 0.357; IT → RP β = −0.240; RP → AST β = −0.513; AST → BI β = 0.650). Scenario means also align with the simulation pattern (Pearson r = 0.803 for AST and r = 0.875 for BI across the nine cells), with the lowest pilot AST (3.667) and BI (3.413) in the fully automated high-risk condition. Discussion The findings connect confidence in government institutions with service-specific trust and indicate that perceived risk constrains acceptance of AI-enabled public services. In high-stakes automated settings, visible arrangements for human review may be necessary for AI service trust to translate into intended use. Public-sector AI acceptance is therefore shaped jointly by institutional credibility, perceived risk, and service encounter design.
Huihui Wang, Shixin Zhu· Frontiers in Psychology· 0 citations
A replicable methodology is introduced, findings across two architecturally distinct LLMs from different developers are extended, and it is demonstrated that deliberate prompt design meaningfully reduces AI decision bias.
Jing-Jie Su, Yan Lang, Kay-Yut Chen· Review of Behavioral Economi...· 0 citations