How AI Chatbots Think Like Humans and Machines

In a landmark study by the Cornell SC Johnson College of Business, researchers found tantalizing evidence of AI chatbots reasoning in ways that are a special mixture of human-like cognitive biases and rational logic.

Decision Making

The dual nature of AI’s decision making was extensively investigated by Stephen Shu, Professor of Practice at the Charles H. Dyson School of Applied Economics and Management, in a study entitled “Do AI Chatbots Provide an Outside View?“. The human-like biases and rational decision-making that AI chatbots exhibit, which Shu and his team found, is something that overturns classic thinking concerning the purely logical nature of AI.

Key Findings:

1. Inside View Biases: AI chatbots will suffer from biases like the conjunction fallacy—in which they tend to believe that specified conditions are more likely than a general one—and the confirmation bias, in which they value information for its agreement with prevailing views.
2. Outside View Rationality: At the same time, AI chatbots are very good at maintaining rationality in areas where humans would tend to deviate from the right path. They weight base rates more effectively and are less susceptible to biases resulting from memory limitations and the endowment effect, which suggests people overvalue items that they own.

Business and Technology Implications

This is very important for business professionals. The use of AI chatbots can support decision-making, but it is necessary to understand their biases. Such techniques would be helpful in mitigating the risks of overconfidence and confirmation bias that may come with AI chatbots. In contrast, their “outside view” can be productively leveraged to support the process of decision-making from more rational standpoints.

Broader Impact and Future Research

As AI becomes integrated into fields beyond customer service and complex business analytics, an understanding of its decision-making nuances becomes increasingly important. This research not only illuminates the current capabilities and limitations of AI but also blazes a trail toward more informed and responsible uses of AI technologies in the future. The full study is published on the Cornell SC Johnson College of Business news site. (Cornell Business).