Evaluating multimodal emotion recognition in proactive conversational agents: A user study
This article discusses a study on multimodal emotion recognition in proactive conversational agents. The research highlights the discrepancies between users' facial expressions and their actual emotional states during interactions with the AI. It emphasizes the importance of refining these agents to better adapt to users' emotional changes through linguistic context.
- ▪The study involved 20 users engaging in unscripted dialogues with a conversational agent.
- ▪Users displayed serious facial expressions even when feeling positive emotions, indicating a 'poker face' effect.
- ▪The linguistic analysis of user interactions was found to be more reliable than visual cues in assessing emotions.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Human-Computer Interaction arXiv:2605.20200 (cs) [Submitted on 6 Apr 2026] Title:Evaluating multimodal emotion recognition in proactive conversational agents: A user study Authors:Adnana Dragut, Raquel Lacuesta, F. Xavier Gaya-Morey, Jose M. Buades-Rubio View a PDF of the paper titled Evaluating multimodal emotion recognition in proactive conversational agents: A user study, by Adnana Dragut and 3 other authors View PDF HTML (experimental) Abstract:This article presents a multimodal emotion recognition module integrated into a proactive Socially Interactive Agent (SIA) powered by generative artificial intelligence.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
