Paper Number

ECIS2026-1469

Paper Type

CRP

Abstract

Humans show more negative reactions to incorrect algorithmic advice compared to incorrect human advice, a phenomenon known as algorithm aversion. However, the cognitive underpinnings of this phenomenon are not fully explored, leaving questions of long-term implications of algorithm aversion open. Addressing this gap by conceptualizing algorithm aversion from the belief-updating lens, we conducted a multi-trial laboratory experiment in which participants (N = 50) made N = 900 decisions and used mouse trajectories to measure belief-updating. Results demonstrate that algorithm aversion could result from asymmetric belief-updating: individuals show more divergent mouse trajectories and delayed responses when relying on human advice, suggesting more extensive belief-updating. In contrast, individuals showed the opposite pattern when relying on algorithmic advice. We contribute a novel theoretical perspective on algorithm aversion by focusing on belief-updating and pre-decisional processes. Moreover, our insights can inform practitioners in the development of algorithmic decision-support-systems and the strategic timing of algorithm deployment.

Share

COinS
 
Jun 14th, 12:00 AM

“I Drift Away Therefore I Am In Decisional Conflict” – Investigating Algorithm Aversion As Asymmetric Belief Updating Using Cognitive Processes Data

Humans show more negative reactions to incorrect algorithmic advice compared to incorrect human advice, a phenomenon known as algorithm aversion. However, the cognitive underpinnings of this phenomenon are not fully explored, leaving questions of long-term implications of algorithm aversion open. Addressing this gap by conceptualizing algorithm aversion from the belief-updating lens, we conducted a multi-trial laboratory experiment in which participants (N = 50) made N = 900 decisions and used mouse trajectories to measure belief-updating. Results demonstrate that algorithm aversion could result from asymmetric belief-updating: individuals show more divergent mouse trajectories and delayed responses when relying on human advice, suggesting more extensive belief-updating. In contrast, individuals showed the opposite pattern when relying on algorithmic advice. We contribute a novel theoretical perspective on algorithm aversion by focusing on belief-updating and pre-decisional processes. Moreover, our insights can inform practitioners in the development of algorithmic decision-support-systems and the strategic timing of algorithm deployment.

When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.