As artificial intelligence (AI) agents become increasingly embedded in digital games and entertainment systems, anthropomorphism (the attribution of human-like traits to artificial agents) has emerged as a key factor shaping player engagement and experience. Despite growing empirical interest, existing findings remain fragmented across domains and lack a coherent synthesis focused on gaming-specific interaction dynamics. This study presents a critical literature review that systematically examines prior empirical and conceptual research on anthropomorphic AI agents in gaming and entertainment contexts. Drawing on Computers Are Social Actors (CASA), Cognitive Load Theory (CLT), and self–AI connection frameworks, the review classifies and compares six core anthropomorphic traits (emotions, social cues, individuality, autonomy, user connection, and appearance) and analyzes how each trait contributes to player engagement, attachment, immersion, and cognitive load. The analysis identifies recurring patterns, trade-offs, and design tensions, showing that while anthropomorphic features can enhance realism, trust, and emotional involvement, excessive or poorly aligned traits may increase extraneous cognitive load or foster over-attachment. By consolidating dispersed findings and highlighting underexplored areas, this study provides a theory-informed synthesis that advances conceptual understanding of anthropomorphism in games and offers actionable design implications for developers, as well as clear directions for future empirical research.
Hasan Tınmaz (Fri,) studied this question.