Bayesian Learning

We know that probabilities are hard for us humans to understand and estimate.  Even seasoned professionals are fooled by seemingly easy problems (take, for example, Marylyn vos Savant’s Parade Magazine fiasco with the Monte Hall Puzzle), and sports fans around the world are fooled whenever they subscribe to the notion that their favorite player is “due” or is on a streak.  I hope I didn’t just alienate too many readers.

Anyway, today I learned about research that investigates some of the evolutionary origins of our troubles with probabilities, and seems to indicate that even purely rational thinkers would come to the same conclusions if their world view was similarly shaped (skewed?) as ours.  In other words, human decision making should be viewed as rational within a model-based framework, and that our decision making is not consistent with blank-slate / model-free learning.

What am I talking about?  Well, the researches looked at binary decision making where the choices have different probabilities of being correct.  Instead of always choosing the one with the higher probability, people essentially chose them in proportion to their chance of being correct (e.g., if the second choice had a probability of 30% of being correct, then close to 15 out of 50 people would go with it, instead of 0 out of 50).  To get a clearer picture, imagine flipping a coin and getting heads 5 times in a row.  You don’t know if the coin is true, or if it’s weighted one way or the other.  Given your experience, though, you might be tempted to think that the next flip has a higher chance of getting a heads.  In other words, it’s probably loaded.  The researchers showed that this is how humans interpret the information – we adopt a Bayesian approach to learning – and we naturally assume that 5 previous heads increases the chances of the next flip ending up heads.  Even if the coin is true, our experience shows that it’s more likely to have a pro-heads bias, so we assign that outcome a higher probability.  So far, so good – this is all consistent with what research has shown before, and I didn’t really see anything all that noteworthy.

What fascinated me was that the researchers showed that once people were told that the coins were in fact true (no bias whatsoever – equal chance of getting heads or tails), that they “did the right thing” and revised their expectations that the next flip had a 50/50 chance of landing on heads.  In other words, the people changed their world-model and made much better decisions – in contrast to what their “gut” was telling them.

From a risk management perspective, this is very positive news.  We may very well have the capacity to overcome our evolutionary biases and make better financial decisions when we are taught to do so.  Score one more for the value of a good education.

2 Responses to Bayesian Learning

  1. Sherry Wolfe says:

    Matt, I agree!

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s