Intelligence agents may be prone to irrational decision making

The study found that intelligence agents exhibited larger biases on 30 gain-loss framing decisions, and were also more confident in those decisions. Thirty-six agents were recruited for the study from an anonymous federal agency, and were presented with scenarios such as:

• The U.S. is preparing for the outbreak of an unusual disease, which is expected to kill 600 people. Do you: Save 200 people for sure, or choose the option with 1/3 probability that 600 will be saved and a 2/3 probability no one will be saved?
• In the same scenario, do you pick the option where 400 will surely die, or instead a 2/3 probability that all 600 will die and a 1/3 probability no one dies?

Notably, the different scenarios resulted in the same potential outcomes -- the first option in both scenarios, for example, has a net result of saving 200 people and losing 400.

But the results showed that agents treated these equivalent outcomes differently based on how the scenario was worded. Specifically, they were more willing than college students to take risks with human lives when outcomes were framed as losses. The researchers explain that the agents also "doubled-down" on their choices, expressing more confidence in their decisions.

These results shed light on the decision-making mechanisms of intelligence agents who identify and mitigate risks to national security, says Valerie Reyna, professor of human development and psychology at Cornell University and lead author of the study. Like some other laboratory gambling tasks, framing effects have been shown to predict real-world behavior, Reyna added.

Participants who had graduated college seemed to occupy a middle ground between college students and the intelligence agents, suggesting that people with more "advanced" reasoning skills are also more likely to show reasoning biases.

Ultimately, Reyna explains, meaning and context play a larger role in risky decision making as experts gain experience. That experience can enhance performance in some cases, but also has predictable pitfalls. Co-authors on this research include Christina Chick, Johnathan Corbin, and Andrew Hsia of Cornell University.