Modeling the Adoption of GenAI for Counter Speech: A Hybrid SEM-ANN Approach Uncovering the Interplay of AI Attributes and User Cognitive-Affective Load

Responding to online incivility threatening democratic discourse, counter speech emerges as a vital approach that is yet underutilized due to the intense cognitive and affective burden. While generative AI (GenAI) offers potential assistance in counter speech, its technical feasibility does not guarantee user adoption. This study addresses this gap by investigating the determinants of adopting GenAI in counter speech. It extends the UTAUT model by integrating GenAI’s transparency and anthropomorphism as antecedents and positioning cognitive and affective loads as moderators. A SEM-ANN approach was employed to analyze survey data from 441 participants. It revealed anthropomorphism predicted use intention, while transparency’s influence was fully mediated through performance and effort expectancies. Crucially, cognitive and affective loads amplified the relationships between GenAI attributes and expectancies. We reconceptualize GenAI as an adaptive scaffold and provide design implications for context-aware GenAI systems that dynamically support civic engagement, offering a sustainable path toward healthy online ecosystems.