Under the bonnet of AI: where the real questions begin

Back when I built statistical and predictive models myself, there was one aspect I found perhaps even more fascinating than the final outcome: playing with the variables.

Taking a variable out. Adding a new one. Watching what happened to the model. Which correlations held up? Which ones disappeared? And why did an effect suddenly change as soon as you factored in something else?

That wasn't just a technical trick. It was a way to get a grip on the mechanisms behind the numbers.

Data never provided a complete answer to how the world works. But by experimenting, you could get a slightly better look at it. You would see that a correlation was sometimes less obvious than you thought. Or that something which initially seemed crucial barely meant a thing once you included another factor.

It was precisely that process that made it interesting. Not the model's highest score. Not the prettiest dashboard. Not the conclusion on the final slide. But rather the moment a new door opened: hang on a minute, what is actually going on here?

Answers have become more accessible

Much has changed with generative AI. You no longer need to build a model yourself, plough through tables, or understand how to technically set up an analysis just to ask a good question. You formulate a query in plain language and within seconds you receive an answer, an analysis, a summary, or a proposed approach.

That is a massive step forward. Far more people can now work with complex information, explore hypotheses, and test ideas. The distance between a question and a preliminary insight is shorter than ever.

But therein lies a risk, too. The easier it becomes to get an answer, the less natural it becomes to probe further.

The temptation of a good answer

A well-formulated answer quickly feels like an endpoint. Especially when it is persuasively written, sounds logical, and is neatly structured. But a convincing answer is not the same as true understanding.

The truly interesting questions often only begin after that:

• What assumptions does this answer rely on?
• What information has been included, and what is missing?
• What alternative explanations are possible?
• What would change if the context were slightly different?
• What uncertainty lies behind this conclusion?
• What question have we actually failed to ask?

In the past, you would sometimes literally see that uncertainty reflected in a model's output: changing coefficients, an effect that vanished, an unexpected interaction. It practically forced you to remain curious.

With AI, that intermediate layer is often less visible. The interface is slick, fast, and human-like. But as a result, some of the friction that helps us think critically also disappears.

Looking under the bonnet

“Looking under the bonnet” doesn't mean that everyone has to become a data scientist, statistician, or programmer. However, it does mean that we shouldn't treat the answer as an automatic endpoint.

For instance, try asking:

“What assumptions are you making to arrive at this advice?”

“What information could fundamentally change this answer?”

“Give me three reasons why this conclusion might be incorrect or incomplete.”

These are not technical questions. They are questions that help us think better. Generative AI can actually be incredibly valuable in this regard. Not just as an answering machine, but as a sparring partner, a hypothesis generator, and a tool for exposing blind spots.

The value of curiosity

For me, the power of statistics was never just about predicting things. It was about practising curiosity.

A model was not an oracle. It was a way of having a conversation with the data. You asked a question, got a clue in return, tweaked something, and discovered that entirely new questions arose.

That attitude has perhaps become even more important now that AI produces answers so effortlessly. So, by all means, use AI to work faster, develop ideas, and make complex information more accessible. But occasionally, make a conscious effort to peek under the bonnet.

Not because the answer always lies there. But because that is often where you find the question that truly matters.