A dual path cognitive behavioural model of artificial intelligence assisted decision making for environmentally sustainable behaviour

Artificial intelligence (AI) systems do more than alter choice architecture: they select, explain, personalise and repeatedly reinforce options. This conceptual paper develops a cognitive–behavioural pathway model explaining when AI-assisted decision-making may narrow the intention–behaviour gap in environmentally sustainable consumption, mobility, energy use and waste-related behaviour, and when the same assistance may become counterproductive. The model distinguishes three enabling mechanisms—diagnostic simplification, calibrated decision confidence and self-regulatory reinforcement—from three countervailing mechanisms—algorithmic and selection bias, miscalibrated reliance, and dependency, moral licensing or rebound. Calibrated reliance is positioned as the central boundary condition linking AI support to action: insufficient reliance produces underuse and algorithm aversion, whereas excessive reliance produces automation bias and reduced independent judgement. Task complexity and user experience are therefore expected to have asymmetric rather than uniformly beneficial moderating effects. The model also introduces bias migration, referring to the displacement of error from individual cognition to system data, filtering and human–AI interaction. The contribution lies in integrating positive and negative pathways within a temporally extended model that distinguishes immediate choice, repeated behaviour, net environmental impact and off-platform agency. The paper concludes with testable propositions and a three-stage empirical programme combining controlled experiments, longitudinal field evidence and algorithmic audits.