Do AI Chatbots Provide an Outside View?

Abstract

Given the rise of generative artificial intelligence (AI), we investigated to what extent AI chatbots provide an "outside view" relative to humans. Managers and individuals are often encouraged to beware of the "inside view," which can lead to problems with people focusing on their own experiences and falling prey to decision-making biases. We find that AI has an inside view on some tasks, such as falling prey to conjunction fallacy, overconfidence, and confirmation biases. However, AI can complement humans with an outside view relative to tasks like consideration of base rates, insensitivity to availability biases, and cognitive reflection. Here, we discuss policy implications, such as increasing AI literacy, identifying use case collaborations, balancing human-or AI-in-the-loop processes, and ongoing monitoring processes.

Keywords: Judgment and Decision Making, Generative AI, Behavioral Economics