Believing the Machine: Human Trust and the Impact of AI Language Confidence and Elaboration

As artificial intelligence increasingly acts as a cognitive partner in decision-making, understanding how users evaluate and trust AI recommendations is critical. While prior research emphasizes algorithmic accuracy, less attention has been given to how presentation features, particularly linguistic confidence and elaboration, shape trust independent of correctness. This study examines how these cues influence trust, attention, and emotional engagement during AI-assisted decisions. Drawing on persuasion theory, trust in automation, and the Elaboration Likelihood Model, it proposes three within-subject experiments using ChatGPT-style outputs. Experiment 1 manipulates confidence framing and format, Experiment 2 varies elaboration depth and format, and Experiment 3 examines accuracy with confidence. Multimodal biometrics capture real-time responses alongside survey-based trust measures.