One evening, I knocked on James’s door. After a distracted “Please…” I entered, only to find him in an unusual situation. James was sitting on the floor, completely absorbed in building a model of an old sailing ship. He looked so much like a child at play that I couldn’t help but laugh.

"Back to childhood, are we?" I teased.

James seemed a bit piqued.

"One might think you're not addicted to modelling..." he retorted.

"I'm not," I replied. "I deal with real things, I don't waste time on games and models."

"What do you call real things, my friend?" James asked with a twinkle in his eye.

"Same things you call real — this room, this desk, you, standing by the desk..." I trailed off, but he cut me off.

"You never see me," he interrupted sharply.

"What do you mean? I see you right now, and—"

"What you see," he continued softly, "is just a distribution of light intensity — a rather weird distribution — that your brain associates with James Bond based on your previous experiences."

"No, I don’t make any associations," I objected. "I'm just seeing you directly."

"No. You have a model of 'James Bond' in your head," Bond insisted. "This model may or may not have anything in common with the real object.""James, your fantasies are sometimes interesting, but..." I trailed off.

Bond, looking tired, like someone explaining the obvious to others, lost his patience. "And now, do you see me?" he asked, suddenly switching off the light.

"Of course, some details are harder to discern, but I can still clearly see you in general." I didn't give up, adding more and more arguments, but faced only icy silence.

Unexpectedly, I felt a hand on my back. It was James, who had cleverly moved to another corner of the room. When the light came back on, I found I’d been arguing in front of a hat hanging on a vase atop an old secrétaire.

"So, you think you're not modelling reality based on a few weird signals coming from the external world?" James continued.

"You fooled me with this old spy trick, but it doesn't prove anything. Yes, modelling is sometimes useful, but ..."

"Sometimes? More like always. You look like a scholar about to publish an article titled 'Model-based Quantification of the Baldness Process.' In fact, all scientific knowledge is model-based." After a brief pause, he added, "And not just scientific knowledge — any knowledge, skills, beliefs, and so-called common sense are nothing but models."

Model Might Miss Essential Things

Honestly, James pissed me off. Even when I got home, I couldn’t calm down. What nonsense he sometimes spouts! Does he really issue absurd, harmful, and even extremist ideas just to appear wise, original, and mysterious? And all with cheap tricks…

To calm down, I decided to do what I love — data analysis. Let’s say we measure light from a distant supernova explosion. Theorists predict that its time dependence should follow a Gaussian curve. All we need to know about this event are the parameters of the Gaussian — the width and the magnitude — and once we have those, we know everything about the supernova. I quickly wrote a code, and voilà, the fit was perfect! I was especially proud of my model because it outputted the variance of the parameters – magnitude a, explosion peak moment mu, and width sigma. This serves as a reliability measure; if the variances exceed a certain threshold, we reject the model.

But then I tested the robustness of the model by introducing additional intensity dips just before and after the explosion’s peak. I expected that the estimated reliability would degrade, prompting us to reject the model. Unfortunately, the results were mixed. Some parameters showed worse reliability, others improved, like the model pretended to handle the data better. What went wrong? I clearly saw the extra features, but the model didn’t. It simply filled the non-fitting gaps and remained content with the result.

So maybe James was partially right — models can sometimes miss essential things.

But generally, he’s wrong. I’m not a model. I’m a data scientist, an experienced professional. I’m, last but not least, a member of the Royal Scientific Society. I see things as they are — I don’t just fill gaps in reality. Or do I?

The Python codes can be found in the pdf version of this document: Full Text with Codes.