2023 Q2 | Edition 2 | Article 1
A message from the Reverend Bayes
Want to know if your latest flame is the one? A 300 year old vicar and founder of machine learning has the answer for you.
In a churchyard in Islington in the heart of London, you’ll find the tombstone of a non-conformist priest, the Reverend Thomas Bayes. In 1761, when he was laid to rest, the burial ground was reserved for people who challenged the status quo. Bayes campaigned for religious autonomy but also challenged the dominant mathematical doctrine of the day, frequentist mathematics.
Two centuries later, another great thinker, Alan Turing, was working on what would become the first computing machine. His machine to crack the Enigma Code, a cypher used during the war, worked, but not fast enough. Turing, influenced by Bayesian mathematics, realised that sometimes a bit of human intervention is needed. The beginnings of machine learning had arrived.
Bayesian Theory and the Enigma Code
Thomas Bayes was a vicar, but he had a hobby he was passionate about – mathematics. In the world of mathematics, there was an important assumption all mathematicians followed - that all variables were random variables until we evidence to the contrary. In other words, every possibility has the same chance of being true until the maths proves that assumption to be wrong.
Bayes disagreed. He argued that there are some conditions which limit the possibility that everything can be true that are completely obvious to humans. If you want to know if a person is pregnant and that person has male sex organs, then you probably wouldn’t bother requesting a pregnancy test. The probability that the person is pregnant is zero. A machine might need to run some tests before reaching the same assumption, but collective human knowledge just knows it to be impossible. Bayes called this human understanding ‘prior knowledge’.
Turing was able to use prior knowledge to halve the run time of his machine, and that was fast enough to produce usable results. Turing realised that all messages sent from distant island outposts generally contained the one same, short and predictable phrase, ‘Nothing to report.’ They were outposts and the war largely passed them by. It was enough prior knowledge, to be able to say with some confidence how those letters were coded. With that head start, Turing and his team were able to quickly find the solution to the remaining letters by the end of the day. Bayesian thinking may have won Turing the war.
The true power of Bayesian thinking in daily life today
Poor decisions and choices are always readily available and begging for attention, but Bayesian thinking can help to rationalise the process.
Bayes left us with three helpful questions to make sense of a range of problems and understand how to modify our behaviours to get better results.
Let’s take the example of deciding if your latest flame is a keeper. Here are the 3 questions you need to ask:
Question 1
What is the base rate likelihood that this person is the right person for me?
In other words, if we’re looking for ‘the perfect partner’ just how common or how rare is it that people in a long term relationship feel that they are in a perfect relationship, or that there is nothing about their partner that they would want to change? If the evidence suggests that the base rate is somewhat low, that many people - say 90% - do feel frustrated, disappointed or downright furious with their partner quite a lot of the time, then our base rate is going to have to be fairly low, too. In fact, if we are looking for a relationship where no compromises are needed and yet both partners are still fully satisfied, we may be looking for a one in a thousand possibility, or perhaps even lower.
Maybe if we changed the parameters a little, from ‘perfect’ to ‘good enough’ we might get a slightly higher base rate.
Question 2
How relevant is the evidence we have to the question that we are trying to solve?
You might have a piece of evidence in front of you about your hunny that seems like a little gold mine of information. You are compatible! They love Taco Tuesdays, and you love Taco Tuesdays too! From a mathematical perspective, a shared liking for tacos is very weakly correlated with having the skills to form a strong, intimate relationships with others over the long term and that’s more your lofty goal.
You might, at that Taco Bar, find it infuriating that your partner wants to have a frank and open discussion about how to split the bill. It’s not a common scene in romantic films, after all. But, assuming that the conversation is handled without judgment or criticism, being able to talk openly about money is fairly universally associated with having one of the skills needed to build a lasting and valuable relationship together. Let’s be frank, if your long term plans include a shared mortgage on a home worth five times your salary, this ability to discuss a restaurant bill without discomfort is quality evidence.
Make sure that the evidence you are using is relevant to the puzzle you are trying to solve.
Question 3
How common is the evidence that you are looking at?
Let’s say that you like men and you have found someone who is a man. That accounts for just under 50% of the population and is quite common evidence. Perhaps we need to start narrowing it down. You probably have an age range in mind, and a sexual orientation that you require and that will bring down the odds against you rapidly.
If you start to add filters for minimum height, a particular income bracket etc. AND you also have to filter out the ones who don’t like you or who terrible relationship skills … Yikes! That number is really going to start to dwindle. You really are looking for a rarity.
Armed with knowledge of just how rare that person might be, you can take steps to increase your odds.
1 One option is to commit to rigorously tracking down that 0.000000001% of the population who meet all your criteria. That may take longer than you want.
2 You could broaden your criteria a little. Is height really good evidence that a person will be a good partner?
3 You could try and turn yourself into more of a catch, so that you too are the perfect body shape and in a similar income bracket and any number of other criteria that could propel you to perfection. Again, are body shape and income parity really a path to a happy relationship?
Finally, after sifting through the options, you might realise that making some better choices - nurturing patience and kindness, developing independence and leading a full life on your own - is probably the best option. These tend to be universally well-liked traits because they are evidence of strong relationship potential. You could study resilience and learn other skills that will stop you from sabotaging your own chances once you do meet someone with true potential.
By really satisfying criteria 1 and 2, you can start to make the numbers game in criteria 3 work for you.
For every action, there is an opposite reaction
Very few solutions in life come with only benefits. Event the term ‘win-win’ symbolises not that there are only benefits for both parties, but that we reach a stable place in the negotiations where we are both happy and dissatisfied in almost equal proportions.
