Paper Type

Short

Paper Number

PACIS2026-2038

Description

As social media platforms deploy AI labels and provenance indicators, transparency intended to reduce uncertainty may instead heighten suspicion or suppress engagement when disclosure lacks verification or salience. This study examines this transparency paradox in the context of GenAI-generated images on social media. Drawing on signaling theory and the theory of consumption values, we theorize GenAI transparency as a configurational signaling system comprising AI info labeling, content credentials, and provenance information salience. We propose that these cues jointly shape simulated engagement behavior through two appraisal pathways: perceived misinformation risk and perceived epistemic value. The model will be tested using a 2 × 2 × 2 between-subjects online experiment manipulating AI labeling, credential verification, and provenance salience in a travel-discovery image context. This study advances IS research by explaining when transparency cues operate as warning signals versus interpretable provenance cues and how provenance design can mitigate risk and enable discovery-oriented engagement.

Comments

12-HCI

Share

COinS
 
Jul 5th, 12:00 AM

The GenAI Transparency Paradox: How AI Provenance Cues Mitigate Misinformation Risk and Enhance Epistemic Value on Social Media

As social media platforms deploy AI labels and provenance indicators, transparency intended to reduce uncertainty may instead heighten suspicion or suppress engagement when disclosure lacks verification or salience. This study examines this transparency paradox in the context of GenAI-generated images on social media. Drawing on signaling theory and the theory of consumption values, we theorize GenAI transparency as a configurational signaling system comprising AI info labeling, content credentials, and provenance information salience. We propose that these cues jointly shape simulated engagement behavior through two appraisal pathways: perceived misinformation risk and perceived epistemic value. The model will be tested using a 2 × 2 × 2 between-subjects online experiment manipulating AI labeling, credential verification, and provenance salience in a travel-discovery image context. This study advances IS research by explaining when transparency cues operate as warning signals versus interpretable provenance cues and how provenance design can mitigate risk and enable discovery-oriented engagement.