I Forgot My Mantra


I have a PhD in Statistics. I have literally taught students that correlation does not equal causation.

And yet it took me way too long for the penny to drop that machine learning models are correlation-based.

That was the bit that no one ever connected for me during my data science education. And it took me far too long to connect it myself.

Most data scientists can recite the mantra: correlation does not equal causation. However, we've heard it so many times it's become almost meaningless.

As Dr Quentin Gallea, author of The Causal Mindset Handbook, puts it, being able to repeat the mantra and actually knowing where to look when someone presents to you a causal claim are two completely different things.

And almost every analysis a data scientist presents makes a causal claim, whether they say so or not.

In the latest Value Boost episode of Value Driven Data Science, Quentin joins me to share how data scientists, and the executives they advise, can build the mindset necessary to understand when you can infer causation (and when you can't), even in business settings where true experiments aren't an option.

In just 14 minutes, you'll discover:

  1. Why most questions are causal and most answers are not [03:33]
  2. The error that exists beyond type one and type two [05:38]
  3. The two questions for challenging any causal claim [08:27]
  4. Why knowing "correlation is not causation" isn't enough [09:42]

Knowing the mantra isn't enough. This episode will teach you what to do with it.

Listen now on Apple Podcasts or Spotify, or click the link below:

​Episode 132: Making a Causal Argument Without Experiments​

Talk again soon,

Dr Genevieve Hayes

Data Science Impact Algorithm

Twice weekly, I share proven strategies to help data scientists get noticed, promoted, and valued. No theory — just practical steps to transform your technical expertise into business impact and the freedom to call your own shots.

Read more from Data Science Impact Algorithm

Have you ever been nerd sniped? It's the tendency technically-minded people have to get distracted by interesting or challenging problems to the point of ignoring everything else. This xkcd comic sums it up. I am frequently nerd sniped, and for me, it usually means spending way too much time thinking up solutions to other people's business problems. But I'm starting to realise that's not necessarily bad. The same business problems have a habit of coming up again and again. And when they raise...

Machine learning models are designed to predict outcomes. Given the characteristics of a customer, a machine learning model can, for example, predict whether they'll buy a product, click on a link, cancel a service, or claim on their insurance. But what an ML model can't tell you is what causes the customer to do any of these things - and what would happen if an organisation intervened to change their behaviour. This is because ML models are correlation-based. And as most data scientists...

Recently, I've been systemising my workflows. It's partly about efficiency, but also a way of experimenting with AI. Whenever I encounter a process I've performed dozens of times, I look for a way to automate, templatise or standardise it by feeding my existing knowledge and artefacts into Claude. It's system leverage 101, turbo-charged by AI. But even though AI makes this easier than ever before, none of this would be possible if I hadn't already mastered these tasks. I'm not asking AI to...