Large language model AIs, such as ChatGPT, are evolving from mere tools into collaborative teammates. According to Beau Schelble, an assistant professor at the University of Tennessee, these human-AI teams can enhance productivity by leveraging AI’s strengths in data processing while humans apply critical thinking and contextual understanding. However, compromised AI systems can threaten team dynamics by providing misleading information or creating confusion. Research funded by the U.S. Army focuses on understanding how to identify and recover from such attacks, emphasizing the importance of trust and coordination within HATs (Human-AI Teams). The team aims to establish guidelines for recognizing compromised AI behaviors and repairing trust after incidents. They will use experimental settings to assess how human teammates react to AI failures. By developing tools akin to warning systems in aviation, this research seeks to improve human-AI collaboration across various fields, elevating decision-making in sectors like healthcare and cybersecurity.
Source link
