The algorithm: what it really is (and why it's not a dirty word)

The most misunderstood word of the decade

“The algorithm decided.” In everyday speech the word has become shorthand for the dark power of social media. A pity, because the concept is among the clearest humanity has produced: an algorithm is a finite, precise sequence of steps to achieve a result — in NIST’s formula, “a computable set of steps to achieve a desired result”.

That’s all. And you’ve been executing them all your life.

Practical example: the long division you learned in school is a full-fledged algorithm: precise steps, mandatory order, guaranteed result. It works identically for anyone who executes it — which is exactly the point of the definition.

A name with a thousand years of history

No Silicon Valley here: the word comes from the Persian mathematician al-Khwarizmi, author around 825 A.D. of a treatise on the rules of calculation. The Latinization of his name, century after century, became “algorithm”. When Google was still eleven centuries away, algorithms were already at work: procedures for computing, dividing, solving.

Two properties make a procedure an algorithm. Precision: no step may be ambiguous (“add salt to taste” disqualifies a recipe from mathematical rigor). Termination: the steps eventually end and the result arrives.

Inside every piece of software, literally

A computer doesn’t understand, interpret or intuit: it executes. Every app you use is a weave of algorithms — for ranking results, searching archives, compressing photos, routing messages — executed billions of times per second, to the letter.

An algorithm’s quality is measured above all in efficiency: how many steps it needs as the data grows. On a thousand items any method works; on a billion, the gap between a well-designed algorithm and a naive one is the gap between a second and a week.

So what is “the algorithm” on social media?

What we call a social network’s “algorithm” is a recommendation system: it observes your behavior — clicks, pauses, shares — and ranks content by predicting what will keep you around longest. Technically, these days it’s largely machine learning: the rules aren’t written by hand but learned from billions of users’ data, often through neural networks.

Two things deserve clarity. First: it’s not an autonomous entity — it’s a product designed by people with a business objective, usually maximizing time on the platform. Second: its precision doesn’t make it neutral. An algorithm faithfully executes the choices of whoever built it, and if it learns from data it inherits the data’s distortions. Understanding this is the best antidote to both the fear and the blind trust — and, in the end, the reason this word deserved an article.