Are Algorithms Making Decisions for Us Now?

October 12, 2026

Your phone knows that you are hungry before you do. Or at least it knows you paused for two seconds on a video of a sizzling burger, which is close enough for the advertising department. By lunchtime, a delivery app has suggestions, a map has a route, and your news feed has a fresh dose of outrage prepared to your personal specification.

So, are algorithms making decisions for us? The honest answer is less dramatic and more unsettling: often, they are arranging the menu from which we decide. And when the menu is arranged often enough, most of us stop asking what was left in the kitchen.

That is not a complaint about technology. Algorithms can spot fraud, translate languages, predict traffic jams, and help doctors notice patterns that a tired human eye may miss. A calculator has never staged a coup. The trouble starts when we confuse a system that calculates probabilities with a system that should decide what is fair, sensible, or humane.

The quiet hand on the steering wheel

An algorithm is, at heart, a set of instructions for reaching an outcome. That sounds harmless because it is harmless in many cases. If an app calculates the quickest route to the airport, it has a clear task and a useful answer. You can still ignore it, take the scenic road, and miss your flight with full personal dignity.

But algorithms have moved far beyond routes and restaurant suggestions. They rank job applicants, determine which posts we see, flag insurance claims, set prices, assess credit risk, and influence who receives extra scrutiny from public agencies. They are increasingly used in schools, hospitals, police departments, and welfare systems. At that point, the question is not whether a machine is intelligent. The question is who gave it the authority to sort people into categories.

A hiring tool may claim to identify the most promising candidate. But promising according to what? Past employees who stayed longest? People who used certain words in a resume? Applicants who attended familiar schools? An algorithm learns from historical data, and history is not a neutral witness. It brings its old preferences, blind spots, and social inequalities to the meeting, usually without even bringing coffee.

The same applies to recommendation systems. They do not have to tell you what to think. They only need to repeatedly decide what you see first, what looks popular, and what never reaches your screen. That is an extraordinarily efficient way to shape a public conversation while insisting that everyone freely chose it.

Are algorithms making decisions for us or narrowing them?

There is a useful distinction that gets lost in the noise. An algorithm can assist a decision, automate a decision, or quietly narrow the choices before a human decision takes place. These are not the same thing.

Assistance can be valuable. A radiologist supported by software that highlights a suspicious scan is still using professional judgment. A city planner using models to test flood risks is making better use of available information. In these cases, the algorithm is a tool. The human remains responsible for the result.

Automation is more difficult. If a bank automatically rejects a loan application, a family may never know which data point sank them. Perhaps it was income volatility. Perhaps an outdated address. Perhaps a proxy that correlates with something the bank would never dare mention aloud. The decision arrives with the calm authority of a locked door.

The most subtle form is narrowing. A streaming service does not force you to watch another crime series. It simply makes that option so easy, so familiar, and so well placed that choosing something else begins to feel like administrative work. Social media works similarly. It rewards attention, not wisdom. Anger, fear, tribal certainty, and adorable dogs all perform well. Nuance is less reliable, unless the dog is also delivering it.

Convenience is the bargain. We save time, avoid endless searching, and receive services tailored to our habits. The cost is that our habits become the raw material from which future choices are shaped. This is not mind control. It is something more ordinary: frictionless repetition.

Efficiency is not the same as judgment

Organizations love algorithms because they promise consistency. Humans are slow, inconsistent, expensive, and occasionally distracted by lunch. A model can process thousands of cases in seconds. For routine decisions, that can be a genuine improvement.

Yet consistency is not automatically fairness. A system can apply a flawed rule with impressive discipline. If an eligibility model wrongly labels a group as high risk, it can repeat that error at scale, day after day, without a moment of doubt. Humans can be biased, certainly. But a human can also notice that a rule produces nonsense in an individual case. A machine only notices nonsense if someone has taught it what nonsense looks like.

This matters especially in public services. If a government uses algorithms to identify possible fraud, allocate housing, or prioritize care, citizens need more than a cheerful statement about innovation. They need to know what data is used, how errors can be corrected, and who is accountable when the system gets it wrong.

The practical test is simple: if an algorithmic decision can seriously affect someone’s livelihood, liberty, housing, health, or reputation, there must be a meaningful path to human review. Not a chatbot that apologizes in six languages. A person with the authority to reconsider the case.

The problem is not only the code

It is tempting to blame the algorithm as though it were a mysterious creature living in a server farm. But algorithms reflect choices made by people and institutions. Someone decides the goal. Someone selects the data. Someone defines success. Someone chooses whether an error is an inconvenience or a catastrophe.

Consider a delivery platform that sends drivers toward areas where demand is predicted to rise. That may improve efficiency. It may also send more traffic into already crowded neighborhoods, reward drivers who work the least desirable hours, and create a city where every street corner is optimized for a business model nobody voted for. The code did not invent that priority. It merely enforced it at speed.

That is why transparency matters, though transparency alone is not magic. Publishing 400 pages of technical documentation is a fine way to make sure nobody reads it. People need plain-language explanations: what is this system deciding, what information does it use, what can go wrong, and what can I do if it harms me?

Companies also need to resist the phrase “the algorithm decided.” That sentence is often a corporate raincoat: useful when responsibility starts falling from the sky. Algorithms do not have moral agency. They do not feel regret, face voters, lose sleep, or answer a difficult email. People do.

Keeping humans in the loop, and awake

“Human in the loop” has become a fashionable reassurance. It sounds sensible, but it depends on what the human is actually allowed to do. If an employee must approve 800 automated recommendations before 5 p.m., that person is not exercising judgment. They are functioning as a decorative seatbelt.

Real oversight requires time, training, access to relevant information, and permission to disagree with the system. It also requires measurement. Are certain groups rejected more often? Are appeals successful at unusual rates? Does the model still work after economic conditions change? A system that was accurate last year may become confidently wrong this year.

As citizens and consumers, we have smaller choices too. We can occasionally search beyond the first recommendation, adjust privacy settings, compare sources, and ask why a particular result appears. None of this will bring down the algorithmic empire before breakfast. But it restores a useful habit: noticing when convenience has begun to make choices on our behalf.

The goal is not to return to paper maps, handwritten ledgers, and waiting on hold for forty minutes. The goal is to keep technology in its proper role. Let algorithms calculate, suggest, and warn. Let them help us see patterns too large for any one person. But when a decision concerns dignity, opportunity, or rights, the final responsibility should belong to someone who can explain it, defend it, and, when necessary, say: the system was wrong.