Induction, Chaos, and Elections

Inductive reasoning, put simply, is when one takes a set of observations, and develops a rule to predict what will happen in the future based on those observations. Let us start with a couple examples. One day you say to yourself, “every morning the sun comes up, and every night the sun goes down. I propose that next morning the sun will come up. I have seen so many instances of this occurring that I am sure it will happen tomorrow.” What happens tomorrow, and every morning after that? The sun comes up in the morning, and so on until your death.

Let us consider a different example. Imagine you are a turkey. One day you say to yourself, “every morning the nice farmer comes out to my pen and places delicious food in my trough. I propose that next morning the farmer will place delicious food in my trough. I have seen so many instances of this occurring that I am sure it will happen tomorrow.” What happens tomorrow? The farmer comes and places delicious food in your trough. This continues for another three months. One day the farmer does not arrive with food, but an axe. That morning the farmer beheads you, and soon you find yourself served for thanksgiving dinner.

What is the difference between these two cases? In one case, the sun, more observations leads oneself to greater certainty of an event; the sun will come up tomorrow morning

In the other, more observations leads oneself to a certainty that tomorrow will bring more delicious food in the trough. In reality, there is a beheading in your future that is nowhere to be found in the data/observations.

This is the problem with induction. In one instance more observations leads to greater certainty, in another instance, more observations leads one to a grossly wrong sense of security.

Next, I want to introduce the idea of a chaotic system (specifically a level 2 chaotic system). A Level 2 chaotic system is something that reacts to predictions. The classic example is that of Oedipus. An oracles prediction leads to oneself reaction to the revelation bringing about that very prediction. As an aside, I could go on a tangent discussing Deleuzian Capture Magic, but I will refrain.

A more salient example is that of United States equity markets. Imagine the following: Minerva Theory publishes an article highlighting how X company will have significant issues in the next year for reasons 1, 2, 3. The CEO of X company sees our article and takes measures to make sure 1, 2, and 3 are not an issue. Over the next year, Minerva’s prediction is proved to be bullshit, and our credibility is diminished. Our original thesis may have been sound; however, because we published our prediction, the end outcome was drastically different.

Let me draw from another example highlighted by Israeli historian Yuval Noah Harai. Imagine you are a data scientist for the United States CIA. Based on your regression analysis and ongoing monitoring in internet data, you predict that a terrorist group will hijack a group of planes and fly them into various buildings on September 11, 2001. Congress takes your prediction seriously and augments TSA, increases background checks, and improves security on planes. September 11^th comes, everyone holds their breath, and September 11^th passes. There is no plane hijacking, you are ridiculed as a charlatan, and lose your job for wasting so much time and money on a bad prediction.

Only because of your prediction and subsequent security improvements did 9/11 not occur. Had you not made the prediction, the hijacking would occur, and the World Trade Center destroyed. Instead you made the prediction, nothing happened, and you lose your job.

Let us apply this idea to polling and elections. Imagine you are a perspective voter and supporter of Party A. Pollsters are aware of the above concept and has incentive to misrepresent polling data. You are at home and see on the telly, “Candidate A is at 37% in the polls, while their opponent is at 63% support.” Your heart sinks and you feel dejected. Your preferred candidate is facing an insurmountable lead. There is no possible way your guy will win the election, regardless of your vote or not. With that in mind, you decide to not go to the polls, resulting in a landslide win for the opposing party. In reality, the election was much closer to 50/50. Had you, and many others, persuaded not to “fruitlessly” vote by the poll actually voted, the election would have been a tossup. Instead, the “prediction” resulted in its desired outcome, discouraging many from voting and materially impacting the outcome of the election.

Are polls good or pernicious?

Charles Wood – Head of Research, Minerva Theory

November 6, 2022