A Dual-Process Perspective on Nudge Susceptibility in LLM-Based GUI Agents
This study examines how LLM-based GUI agents are influenced by nudges, highlighting potential vulnerabilities in their decision-making processes.
This study examines how LLM-based GUI agents are influenced by nudges, highlighting potential vulnerabilities in their decision-making processes.
Auto-scored: The paper identifies a minor risk of nudge susceptibility in GUI agents, which could lead to unintended actions, but no immediate severe harm is demonstrated.