Chapter 9 — What the Algorithm Sees

What the Algorithm Sees

There was a time when finding a book meant wandering.

A person entered a library with only a vague idea of what they were looking for. They drifted through shelves, reading titles that caught the eye. Sometimes the book they came to borrow was not the one they carried home. Discovery happened because neighboring books occupied neighboring space.

Libraries had always been organized. But they still allowed for accidents.

Today, many searches begin with a blank rectangle. A question is typed. A fraction of a second later, thousands—or millions—of possible answers have already been placed in order. Most of us click one of the first few. Very few of us ask why those appeared before the others.

The remarkable achievement of the internet was never simply that it connected information. It connected far more information than anyone could possibly navigate alone. Without systems of ordering, the web would have become less like a library than like a warehouse where every book had been dropped onto the floor.

Ranking solved a genuine problem. Someone—or increasingly, something—had to decide what should appear first.¹

The earliest search engines relied heavily upon relatively simple signals. Some counted words. Others measured links. A page referenced by many other pages appeared more authoritative than one standing alone. The approach was imperfect, but it allowed people to find what they were seeking with astonishing speed.

The ranking was useful because it usually felt reasonable. The better it became, the less visible it grew.

This is one of sediment's quieter habits. A successful solution disappears into ordinary experience. We stop noticing the ordering itself and begin noticing only the ordered world.

Long before algorithms, human beings had always arranged attention. Editors decided which stories belonged on the front page. Museum curators determined where visitors would begin. Teachers chose which books students would read before others. Parents decided which stories children heard before bed.

No society has ever existed without ranking. Attention has always required selection.

What changed was scale. No editor chooses the billions of recommendations flowing through streaming services, social platforms, online marketplaces, navigation systems, and search engines each day. The ordering happens continuously, adapting to millions of people at once.

The selection becomes personal. And because it becomes personal, it often becomes invisible.

Two people standing side by side can type the same question and encounter different worlds. One receives restaurants. Another receives news. A third receives advertisements. The ranking quietly learns what has held each person's attention before and uses that history to shape what appears next.

The process feels almost natural. Of course the computer knows what I like. Of course these are the best results. Of course this appeared first.

Those small assumptions deserve a second look.

A ranking system never asks, "What is true?" It asks questions that are, by necessity, narrower. What is most relevant? What is most useful? What is most likely to satisfy this search? What is most likely to keep this person's attention?

These are practical questions. They are also design choices.

Different systems answer them differently because different organizations optimize for different outcomes.

The interesting transformation happens slowly. At first, the ranking helps us navigate. Eventually, it begins helping us decide.

Consider how often we equate visibility with importance. A restaurant appearing first in search results feels more trustworthy than one buried on the fourth page. A video viewed ten million times appears more culturally significant than one viewed ten thousand times. An article recommended repeatedly begins to feel more central than one encountered only by chance.

Position quietly becomes persuasion.²

None of this requires manipulation. It requires only repetition.

Human beings have always learned from patterns. If something appears consistently at the top of lists, we naturally begin assuming it belongs there. This is not a failure of intelligence. It is one of the shortcuts that makes everyday life manageable.

We cannot investigate every possibility from first principles. We depend upon ordering.

That dependence grants remarkable influence to those who build systems of ranking. Not because they control what we think. More often because they influence what we are likely to encounter before we begin thinking.

There is an old saying that maps are not the territory. Ranking systems remind us of a related truth.

Lists are not reality.³

They are interpretations of reality, created according to particular purposes, using particular signals, at particular moments in time.

Some of those interpretations are extraordinarily good. Search engines have made vast stores of human knowledge accessible in ways earlier generations could scarcely imagine. Recommendation systems regularly introduce people to music, books, films, and ideas they might never have discovered on their own.

The achievement is genuine. Sediment is once again performing useful work.

But usefulness can become invisibility. After enough years, we stop experiencing rankings as recommendations. We experience them as landscapes. The first result simply feels like where the path begins.

Perhaps every civilization has built such paths. Ancient roads directed merchants through mountain passes. Railroads concentrated towns around stations. Highways redirected commerce toward exits.

Today's pathways are often digital. They guide attention instead of footsteps. The principle, however, remains remarkably familiar. Build enough paths and people begin mistaking them for the terrain itself.

This is why ranking deserves our attention. Not because it eliminates choice. Choice remains. But because it quietly shapes which choices become easiest to notice.

The most enduring structures rarely command us. They orient us. They make some directions feel obvious and others difficult even to imagine.

Perhaps that is the deepest sediment algorithms are laying down. Not the answers they provide. The paths they make feel natural.

Footnotes

1. See Cathy O'Neil, Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy (New York: Crown, 2016); Safiya Umoja Noble, Algorithms of Oppression: How Search Engines Reinforce Racism (New York: New York University Press, 2018). ↩

2. See Tim Wu, The Attention Merchants: The Epic Scramble to Get Inside Our Heads (New York: Alfred A. Knopf, 2016). ↩

3. See Geoffrey C. Bowker and Susan Leigh Star, Sorting Things Out: Classification and Its Consequences (Cambridge, MA: MIT Press, 1999). ↩