Paper Type

Short

Paper Number

PACIS2026-1230

Description

Generative AI increasingly provides health explanations through chatbots and search-like AI summaries, yet users often encounter unfamiliar health questions without an established stance. This paper develops a message-evaluation model of how evidence presentation mode shapes perceived diagnosticity, perceived credibility, and intention to follow AI advice. Evidence presentation mode is operationalized as inline versus separated previews for the same verifiable source cues. Building on IS research on task interruptions in decision making, we argue that inline presentation reduces verification friction and facilitates claim-source comparison. We propose that perceived diagnosticity mediates the effects of evidence presentation on credibility and follow intention, and these indirect effects vary by users’ belief states. The paper contributes by introducing and operationalizing the no-belief state in AI-generated health explanations, clarifying when low-friction verification design most strongly shapes reliance, and offering guidance for presenting checkable sources to support informed and responsible use of AI health summaries.

Comments

14-Healthcare

Share

COinS
 
Jul 5th, 12:00 AM

No-Belief and Verification Friction: How Evidence Presentation Shapes Reliance on AI-Generated Health Explanations

Generative AI increasingly provides health explanations through chatbots and search-like AI summaries, yet users often encounter unfamiliar health questions without an established stance. This paper develops a message-evaluation model of how evidence presentation mode shapes perceived diagnosticity, perceived credibility, and intention to follow AI advice. Evidence presentation mode is operationalized as inline versus separated previews for the same verifiable source cues. Building on IS research on task interruptions in decision making, we argue that inline presentation reduces verification friction and facilitates claim-source comparison. We propose that perceived diagnosticity mediates the effects of evidence presentation on credibility and follow intention, and these indirect effects vary by users’ belief states. The paper contributes by introducing and operationalizing the no-belief state in AI-generated health explanations, clarifying when low-friction verification design most strongly shapes reliance, and offering guidance for presenting checkable sources to support informed and responsible use of AI health summaries.