(Some) Humans do choose 5 dollars in the 5-and-10 problem

post by Akira Pyinya · 2025-04-07T22:56:56.606Z · LW · GW · 1 comments

Contents

1 comment

It's interesting to look at the 5-and-10 problem [? · GW] not as a thought experiment, but as a real-world problem. You may find that some people do indeed irrationally choose $5 in this problem, and we may say that they suffer from an anxiety disorder.

Let's dive into the details of how people think counterfactuals.

People think counterfactuals not by modeling an entire world including themselves, but by placing themselves in the counterfactual using imagination. When people imagine a counterfactual, they have feelings similar to what they actually feel in that counterfactual situation, like the mirror neurons that respond to both food and imaginary food. They are imagining a different environment without imagining a different self.

How do people learn to choose between counterfactuals? As we have learned from neuroscience and reinforcement learning, human behavior is not reinforced by rewards or utility, but by the Temporal Difference (TD) error, the difference between expected and actual rewards.

If you imagine yourself switching from one counterfactual to another, for example, from choosing $5 to choosing $10, you will get a positive TD. A positive TD error will reinforce this switch through the trial-and-error learning mechanism. You can learn to switch to $10 if you accidentally approach the $5 side the next time. You can also learn to avoid choosing $5, vice versa.

 

 

Unfortunately, this reinforcement mechanism does not always work perfectly. Sometimes the switching actions are not properly reinforced, and people may experience anxiety: feelings of dread about counterfactuals, as if they are really in bad situations. In the worst cases, they make bad decisions or even end their lives.

This mechanism of counterfactual thinking is slow and risky. Fortunately, we don't have to make decisions every minute. Most of our daily tasks are routine and planned, without much TD error, unless you are traveling without plans, on a chaotic battlefield, or living under a capricious ruler.

1 comments

Comments sorted by top scores.