Supporting human adaptation in the age of Artificial Thinking, HAI/UX principles grounded in the TTA model

This research introduces the TTA (Think–Tool–Action) model to examine how humans adapt their cognition when Artificial Thinking Systems (ATS), like large language models, participate directly in the reasoning process. The author argues that while historical tools supported physical or computational tasks, modern AI operates within the human thinking stage, introducing risks such as automation bias, skill atrophy, and anthropomorphism. To address these issues, the thesis derives eight novel HAI(Human-AI Interaction)/UX principles organised into layers: orientation, relational, interaction, and longitudinal. These principles are transformed into specific UX levers, such as Transparency Drift Signals and Context Path Maps, which are designed to foster reflective and agency-preserving interactions. Empirical results from a user study indicate that these interface interventions successfully shift users toward more critical engagement and active self-monitoring. Ultimately, the work provides a framework for designing AI interfaces that protect human autonomy and sustain cognitive skills during collaborative reasoning.