Predicting emotional responses in interactive art using Random Forests: a model grounded in enactive aesthetics

Area of Science:

• Affective computing
• Human-computer interaction (HCI)
• Empirical aesthetics

Background:

• Interactive art elicits complex emotions, posing challenges for computational modeling due to their dynamic and subjective nature.
• Existing methods struggle to capture the full spectrum of emotional responses, from sensorimotor engagement to cognitive reflection.

Purpose of the Study:

• To introduce an interpretable machine learning framework for predicting emotional responses in interactive art.
• To evaluate the Random Forest algorithm's effectiveness in modeling multidimensional emotional experiences.
• To provide insights for designing emotionally adaptive interactive systems.

Main Methods:

• Utilized a Random Forest (RF) algorithm with 390 questionnaire responses.
• Operationalized emotions across five dimensions: bodily changes, sensory engagement, emotional connection, cognitive reflection, and active personalization.
• Employed cross-validation and test sets for model evaluation using classification and regression metrics.

Main Results:

• The RF model achieved high predictive accuracy for cognitive reflection (F1=0.746) and active personalization (F1=0.673).
• Bodily responses were less predictable (F1=0.379), likely due to their subjective and non-verbal nature.
• The model demonstrated consistent performance, validating its use as an exploratory tool.

Conclusions:

• Cognitively mediated emotional states are more tractable for computational modeling than sensorimotor or affective responses.
• The proposed framework offers actionable insights for designing emotionally adaptive interactive art and systems.
• Future research should incorporate multimodal data and address ethical considerations of affective adaptivity.