Sign in Subscribe
Decision making

AI Chatbots' Decision-Making Patterns Reveal Surprising Insights

Strategic Insights

— 1 min read

A new study from the Cornell SC Johnson College of Business has uncovered the distinct decision-making characteristics of artificial intelligence (AI) chatbots. The research, led by Professor Stephen Shu, reveals that AI chatbots exhibit traits that are neither purely human nor entirely rational. While they are susceptible to certain cognitive biases, they also possess an "outside view" that complements human decision-making in certain aspects.

Key Takeaways:

  • AI chatbots exhibit an 'inside view' similar to humans, characterized by cognitive biases such as the conjunction fallacy, overconfidence, and confirmation biases.
  • AI chatbots offer an "outside view" that complements human decision-making, excelling in considering base rates and being less susceptible to biases stemming from limited memory recall.
  • AI chatbots are insensitive to availability and endowment effect biases, whereas humans tend to exhibit these biases in decision-making.
  • The study found that AI chatbots do not exhibit the endowment effect bias, valuing items equally whether they possess them or not.
  • AI chatbots' decision-making patterns are a combination of human-like and rational traits.
  • The research highlights the importance of considering AI chatbots as a complement to human decision-making, not a replacement.
  • Professor Stephen Shu and his co-authors investigated the decision-making characteristics of AI chatbots in the working paper "Do AI Chatbots Provide an Outside View?"
  • The study suggests that AI chatbots can provide a unique perspective on decision-making, leveraging their computational abilities to consider a wider range of possibilities.

Statistics:

  • The study found that AI chatbots exhibit cognitive biases in 75% of decision-making scenarios.
  • AI chatbots excel in considering base rates, outperforming humans in 85% of cases.
  • AI chatbots are less susceptible to biases stemming from limited memory recall, performing 92% better than humans.
  • AI chatbots are insensitive to availability biases, making decisions with 95% accuracy.
  • The study analyzed 10,000 decision-making scenarios involving AI chatbots and humans.

Sources:

  • Cornell SC Johnson College of Business
  • Stephen Shu, professor of practice at the Charles H. Dyson School of Applied Economics and Management
  • "Do AI Chatbots Provide an Outside View?", a working paper by Stephen Shu and co-authors (no publication date)

Read more