Human-Centric Reflective Architecture for Human-AI Collaborative Decision-Making

This paper formulates human-AI collaborative decision-making as a stochastic game between an AI agent and a human player. It introduces the Human-Centric Reflective Architecture (HCRA), combining human-calibrated models with reinforcement-learning agents that use linguistic feedback in an iterative reflective loop. The abstract reports improved decision-making effectiveness and high-quality recommendations. However, it does not disclose the evaluated tasks, baselines, numerical results, calibration procedure, or the specific reinforcement-learning setup, so the strength and generality of the reported gains cannot yet be independently assessed.