For those interested in statistics or decision theory the “Bayesian Theorem” should be known material.
And then there is the rest of us.
All the same, I much enjoyed an article on Bayesian thinking presented by the Danish nudging expert Morten Munster. One of his strengths is that he explains complicated things in a language even I can follow. I am sure that is part of his nudging strategy. (After all, it is a little difficult to make people change their thinking if they have no clue what you are talking about.)
Morten’s explanation was filled with great examples; let me shamelessly steal his wedding couple:
During a wedding ceremony, the couple most likely assumes that the chance their union will end in a divorce is less than 50:50. All the same, that is the baseline chance of divorce based on the general statistic.
First step in Bayesian thinking is to neutrally be aware of the baseline for whatever you are engaged in, based on what similar activities that you were not involved in have shown. (Bayes referred to this as a Prior.)
The next step is to refine our thinking because we know a little more about our couple than about all the other couples.
Some factors make a divorce more likely down the line: Prior divorces, parents’ divorces, divorces among friends, history of infidelity, history of substance abuse, employment in jobs that have historically had higher than average divorce rates…
Some factors make divorce less likely: Add a “no” in front of all the above.
Notice that all this has absolutely nothing to do with how you feel about neither the factors, nor the couple getting married.
Finally, when you learn that the couple doesn’t live in USA, you adjust your prior or your other factors to reflect the data for their country.
Again, a piece of new information makes you adjust your expectations regardless how you feel about the higher/lower divorce rate of the other country.
This is not so hard, is it?
Well, in real life it is. We probably have opinions about divorce and/or know the people getting married, and our gut feeling is screaming to be allowed to have a say: This marriage is made in Heaven!!!
(Or alternatively: What does she/he have that I haven’t? / What a loser! / You just wait; I told you so… )
Even worse, we may earlier have expressed an opinion that is now contradicted by the somber facts. That means we must change our mind. And that can be difficult – if we are too invested in our opinions.
All that can lead down a discussion trail in which I don’t wish to go with this piece. Because what I really want to point out is that Bayesian thinking can also be useful when we meet people – online or in real life.
This thought came by way of Susan Rooks who talked with The Rabbi and the Shrink, a.k.a. Yonason Goldson, and Dr. Margarita Gurri. Susan described how she had grown up seeing pictures and hearing her parents’ enthusiastic praise for the people they had met abroad. Consequently, although she grew up in a rather homogeneous environment, she always assumed that people different from her were kind, interesting, and had something valuable to contribute.
How is that for a great Prior when it comes to building relationships?
When you meet people for the first time, do you think they are kind, interesting, and have something valuable to contribute before any of you have opened your mouth?
To some people, “strangers are friends you don’t know yet”. Even if they don’t look or sound like you.
To other people, “Stranger = Danger”.
One of these is likely your Prior. If you are not aware of what your worldview is when it comes to strangers, totally innocent people will be unwitting actors in a play you set up in your head.
Even if you start out as positive as Susan, if the stranger says something you disagree with, do you get curious why they say that – or do you suddenly think they are not really as kind, interesting, or valuable?
This is another of the central points in Bayesian thinking: If there is a significant Prior, the real outcome doesn’t change very much by single random events.
Let us take another example:
Think about your valedictorian high school student with a clean 4.0 GPA just before the final examen. 4.0 is the prior. Even if the student gets a D on a test that counts 30% towards the grade in that class, the impact on the overall GPA earned over four years is miniscule. (Literally from 4.000 to 3.996.)
But I bet that the student thinks they are a complete failure. Their focus is on the one 30% D, not on the 47.7 A+s.
This indicates that we are not very good in Bayesian thinking – neither when it comes to our successes, nor when it comes to evaluating the value of other people with whom we may disagree.
So what can we do about that?
We can emulate Susan and stay curious.
You may have had the MMR vaccine and hence never have had mumps, measles, or rubella. But you probably trust your parents if they tell you how unpleasant that was. They are from a different generation so you trust that their life has unfolded differently from yours.
If people don’t look or sound like you, is it likely that they also may have life experiences different from yours and hence they say something that may be true for them even if it is not true for you? Stay curious.
If your trusted friend over 30 years says something that is “way out”, your many years of friendship should be the prior. The blip that is way out should not make you jump to the conclusion that you have been mistaken about this person for 30 years; it should lead you to think that you misunderstood what your friend said. Stay curious.
And if the friend really means whatever was said, stay with them and figure out what their prior is for what they said, what is the context, what data informs their opinion. If they have been duped or fallen into a Facebook wormhole, isn’t it likely that they might listen to somebody with whom they already have a 30 year prior, who is truly curious about how they think, and who is willing to help them research data supporting or contradicting their position?
After all, you have 0% chance of influencing people with whom you are not on speaking terms.