Paper Type
ERF
Abstract
As AI-based decision-support systems increasingly shape decision-making, users often override algorithmic predictions when outputs conflict with their expectations. This study proposes a dynamic model of trust calibration in AI-driven sports analytics, conceptualizing trust as enacted reliance measured through accept-override decisions. We examine how expectation violation magnitude and explanation quality jointly influences reliance. Using a repeated-measures experimental design, participants generate their own forecasts, view AI predictions, and decide whether to accept or override them; expectation violation is modeled continuously as the discrepancy between user and AI forecasts. Integrating perspectives from NeuroIS and HCI, we incorporate eye tracking and galvanic skin response to capture pre-decisional cognitive effort and arousal preceding override behavior. Multilevel analyses will test how expectation violations predict reliance, how explanations moderate these effects, and whether trust predicts behavior beyond perceived surprise. The study advances research on algorithm aversion and explainable AI by positioning trust as a dynamic, process-level phenomenon.
Paper Number
1872
Recommended Citation
Finn, Norman; Mamudu, Favour; Vargas Diaz, Mariana; and Bačić, Dinko, "When Expectations Clash: Dynamic Trust Calibration and Expectation Violation in AI-Driven Sports Analytics DSS" (2026). AMCIS 2026 Proceedings. 27.
https://aisel.aisnet.org/amcis2026/sig_dsa/sig_dsa/27
When Expectations Clash: Dynamic Trust Calibration and Expectation Violation in AI-Driven Sports Analytics DSS
As AI-based decision-support systems increasingly shape decision-making, users often override algorithmic predictions when outputs conflict with their expectations. This study proposes a dynamic model of trust calibration in AI-driven sports analytics, conceptualizing trust as enacted reliance measured through accept-override decisions. We examine how expectation violation magnitude and explanation quality jointly influences reliance. Using a repeated-measures experimental design, participants generate their own forecasts, view AI predictions, and decide whether to accept or override them; expectation violation is modeled continuously as the discrepancy between user and AI forecasts. Integrating perspectives from NeuroIS and HCI, we incorporate eye tracking and galvanic skin response to capture pre-decisional cognitive effort and arousal preceding override behavior. Multilevel analyses will test how expectation violations predict reliance, how explanations moderate these effects, and whether trust predicts behavior beyond perceived surprise. The study advances research on algorithm aversion and explainable AI by positioning trust as a dynamic, process-level phenomenon.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.
Comments
SIG DSA