The secret weapon in the Information War

People who often believe fake news and conspiracy theories are guilty of a specific cognitive error. We could call this error “failure to apply Occam’s razor”.

William of Ockham, portrayed in a manuscript of his “Summa Logicae”, 1341 (Wikipedia)

For those who are not familiar with Occam’s razor: roughly speaking, it is the principle saying that when multiple theories explain an observation equally well, we should choose the theory that makes fewest assumptions (the “simplest” one, in other words). Occam’s razor has a deep, elegant probabilistic justification1, and it was first formalised by the English Franciscan philosopher William of Ockham in the 1300s.

Why do I say that conspiracy theorists fail at applying the principle? Because conspiracy theories are incredibly complex. They make assumptions about the number and level of people involved in hiding some truth, assumptions about their skills, aligned motivations and about how they managed to keep everything secret so far. All of this to explain observations (the apparent flatness of our planet, terrorist attacks, murders, pandemics) that can easily be accounted for with much more parsimonious theories2.

I find it really ironic that conspiracy theorists don’t understand Occam’s razor, and for one reason in particular. It turns out there is a group of people who are really good at following Occam’s advice: they are the data scientists working on the fake news algorithms that target the conspiracy theorists.

The secret sauce of machine learning

Machine learning models, such as those that find the right audience for fake news online, can be quite complex. But a crucial factor makes them work: they are never more complex than they need to be. When a model is too complex, data scientists say that it is overfitting, and that’s a huge problem. An overfitting model can be disastrously inaccurate. A big part of modelling in data science is about avoiding overfitting – that is, applying Occam’s razor.

A typical task for a Machine Learning model is finding a curve that separates two groups of points (here represented with different colours). The model on the left is unnecessarily complicated and it is unlikely to perform well on new data (overfitting)

This is true for all kind of models: those applied to credit risk, marketing, image recognition, text analysis and so on. And, of course, for models whose purpose is to identify the right audience for fake news.

In other words, a big part of the job of those who spread fake news online is applying the logical principle (Occam’s razor) that should make people skeptical of the content of fake news and conspiracy theories!

An old story with new consequences

This is an old story in modern clothes. In history, the source of power has always been the unequal allocation of knowledge, not just of resources.

But there is something new here: today differences in knowledge can make an unprecedented difference in the balance of power. In the era of information, the ability to cut through the overwhelming amount of data and news is crucial to make sense of the world and act accordingly as consumers, voters and citizens.

Those who can master the subtleties of logic (and one of its modern incarnations, machine learning) have an advantage over those who cannot. The former can literally change the reality bubbles of the latter. Recalling the philosophical principle of a Franciscan friar of the 14th century can help us navigate this battle and fight back.

[1] MacKay, David J. C. (2003). Information Theory, Inference, and Learning Algorithms (chapter 28)

[2] This is not say that all conspiracy theories are necessarily wrong, of course. But they should never be preferred to simpler explanations, unless they can explain observations better than the simpler explanations – and in my experience that happens very rarely.

One thought on “The secret weapon in the Information War

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 )

Connecting to %s

%d bloggers like this: