AffectRankTrace: A Tool for Continuous and Discrete Affective Annotation During Extended Usability Trials

This paper introduces AffectRankTrace, a tool for affect annotation that combines the strengths of discrete and continuous annotation techniques with rank-based principles. We hypothesize that by applying these principles, we can reliably target the three affective dimensions of pleasure, arousal, and dominance, and capture a broader spectrum of users' affective states. Using the tool, 25 participants provided annotations for the three dimensions following a 25-minute play session of a commercial horror game while wearing physiological sensors. They also offered feedback on task engagement and tool usability. Using both physiological and annotation data, we built a prediction model. Our results reported F1-scores of 0.630, 0.661 and 0.661 for arousal, pleasure and dominance, respectively. These findings align with baseline results from established multimodal datasets, suggesting the tool's reliability. Participants' feedback on usability and engagement also supports the tool's usability and suggests the potential for conducting annotation during extended usability trials.