Thesis on Designing Human-AI Interaction Loops for Improvement and Balanced Objectives

A recent thesis available on arXiv (2608.05710) investigates the dynamics of human-AI interaction loops, emphasizing how people develop beliefs about AI systems and how they strategically express their preferences, behaviors, or characteristics. This interaction generates feedback or decisions from the AI, forming a continuous loop. The study aims to achieve three objectives: aiding individuals in forming accurate beliefs to enhance or secure beneficial outcomes at low costs, promoting improvement while deterring manipulative behaviors, and ensuring the AI system adheres to its core goals, including accuracy. The thesis, likely intended for a doctoral degree, is divided into three sections, analyzing interactions from both the perspectives of the individuals being evaluated and the AI systems. This new research highlights the critical need to understand and influence human-AI interactions in operational systems, with significant implications for fairness, transparency, and effectiveness.