We presented a full paper titled “Do You (Dis)agree With Me? Modelling Implicit User Disagreement in Human–AI Interaction Using Gaze Data” at CHI 2026 in Barcelona during the Inferring Human State session. The ACM Conference on Human Factors in Computing Systems (CHI) is generally considered the most prestigious conference in the field of human–computer interaction and had an acceptance rate of 25.3% from a total of 6,730 completed submissions. The paper investigates whether implicit user disagreement with AI-generated outputs can be detected from passive gaze and facial signals during human–AI interaction. Using a controlled image-captioning task with 30 participants, we compared generalised and personalised machine learning models and found that personalised gaze-based models performed best, while facial and multimodal approaches were less effective. These findings suggest that disagreement with AI can be detected, but in a highly user-specific way, highlighting the importance of more adaptive and personalised AI systems. To support future research in this area, the dataset has also been made publicly available on GitHub.
Abdulrahman Mohamed from IML at the CHI conference 2026
Abdulrahman Mohamed presents the paper at the conference
