Paper Type
Complete
Paper Number
PACIS2026-1398
Description
As generative AI integrates into digital platforms, time-synchronized AI-generated video summaries (AIGS) increasingly help users navigate complex content. Yet their effects on micro-dynamics of in-consumption engagement remain unclear. This study conceptualizes AIGS as time-synchronized informational interventions and investigates their impact on in-consumption engagement quantity and semantic similarity during video consumption. Using a quasi-natural experiment on BiliBili, a major Chinese video platform characterized by real-time danmaku interaction, we apply a segment-level econometric design to separate AIGS effects from natural content climaxes and uncover a double-edged sword effect. We find that AIGS increase both in-consumption engagement quantity and semantic similarity. Mechanism analyses suggest that cognitive anchoring primarily explains the increase in engagement quantity, whereas semantic similarity is associated with both cognitive anchoring and cognitive miser. This study advances research on AI-assisted content consumption and offers governance insights for balancing engagement volume with discursive diversity on video-sharing platforms.
Recommended Citation
Pan, Zhao; Wang, Ling; and Li, Chenwei, "The Double-Edged Sword of Time-Synchronized AI-Generated Video Summaries on In-Consumption Engagement: Evidence from a Natural Experiment on Bilibili" (2026). PACIS 2026 Proceedings. 7.
https://aisel.aisnet.org/pacis2026/general_topic/general_topic/7
The Double-Edged Sword of Time-Synchronized AI-Generated Video Summaries on In-Consumption Engagement: Evidence from a Natural Experiment on Bilibili
As generative AI integrates into digital platforms, time-synchronized AI-generated video summaries (AIGS) increasingly help users navigate complex content. Yet their effects on micro-dynamics of in-consumption engagement remain unclear. This study conceptualizes AIGS as time-synchronized informational interventions and investigates their impact on in-consumption engagement quantity and semantic similarity during video consumption. Using a quasi-natural experiment on BiliBili, a major Chinese video platform characterized by real-time danmaku interaction, we apply a segment-level econometric design to separate AIGS effects from natural content climaxes and uncover a double-edged sword effect. We find that AIGS increase both in-consumption engagement quantity and semantic similarity. Mechanism analyses suggest that cognitive anchoring primarily explains the increase in engagement quantity, whereas semantic similarity is associated with both cognitive anchoring and cognitive miser. This study advances research on AI-assisted content consumption and offers governance insights for balancing engagement volume with discursive diversity on video-sharing platforms.
Comments
17-General