Paper Number
ICIS2026-4227
Paper Type
Short
Abstract
Explainable AI research increasingly advocates interactive natural-language explanations to support appropriate reliance, yet dialogical designs may also alter how users interpret an AI system's intent. Drawing on attribution theory, this short paper proposes that interactive explanations heighten perceived humanness, thereby strengthening intentionality attributions — more strongly for perceptions that the system seeks to persuade than to assist — ultimately impacting reliance behavior. Additionally, decision-makers’ Locus of Control conditions whether perceived humanness is read as assistance or persuasion. Using a between-subjects online experiment in a deception-detection task, we compare recommendation-only, static explanation, and interactive explanation conditions and measure behavioral reliance on correct and incorrect AI advice. The study contributes to explainable AI research by showing how explanation design affects assistive and persuasive attributions of intent, which shape reliance behavior.
Recommended Citation
Lomo, Marvin Adjei Kojo; Singh, Rahul; and Ray, Arindam, "How the Design of Explanations Affects Intentionality Attributions and Appropriate Reliance in AI-Augmented Decision-Making" (2026). ICIS 2026 Proceedings. 40.
https://aisel.aisnet.org/icis2026/user_engage/user_engage/40
How the Design of Explanations Affects Intentionality Attributions and Appropriate Reliance in AI-Augmented Decision-Making
Explainable AI research increasingly advocates interactive natural-language explanations to support appropriate reliance, yet dialogical designs may also alter how users interpret an AI system's intent. Drawing on attribution theory, this short paper proposes that interactive explanations heighten perceived humanness, thereby strengthening intentionality attributions — more strongly for perceptions that the system seeks to persuade than to assist — ultimately impacting reliance behavior. Additionally, decision-makers’ Locus of Control conditions whether perceived humanness is read as assistance or persuasion. Using a between-subjects online experiment in a deception-detection task, we compare recommendation-only, static explanation, and interactive explanation conditions and measure behavioral reliance on correct and incorrect AI advice. The study contributes to explainable AI research by showing how explanation design affects assistive and persuasive attributions of intent, which shape reliance behavior.
Comments
14_User_behavior