Chapter 11 — The Ethics of the Reading Eye

Chapter 11 — The Ethics of the Reading Eye

We should talk about what it feels like to be read.

For ten chapters we have stood beside the detective, sharing his pleasure in seeing through people. It is a delicious position, the reader's chair; the eye that reads is always the powerful one. But every act of inference has a second person in it — the one being inferred about, the one whose secrets are being lifted without consent from the traces they didn't know they were leaving. And when the science of deduction stops being a party trick and becomes an instrument of states and corporations, that second person is you. All of us. Read continuously, by machines that never blink, for purposes we never agreed to. This chapter is about the cost of the gift, and it is the one chapter Doyle never could have written, because in his world the eye that reads is always benevolent and always right. In ours, it is neither.

The prison that watches itself

Two centuries ago the philosopher Jeremy Bentham designed a building. He called it the Panopticon — an "all-seeing place." It was a prison built as a ring of cells around a central tower, arranged so that a single guard in the tower could see into any cell at any moment, while the prisoners could never tell whether they were being watched. Bentham's genius, and his horror, was the realisation that the guard need not actually watch. The possibility of being watched, made constant and inescapable, does the work. The prisoners, unable to know when the eye is on them, behave at all times as if it is. They internalise the watcher. They become their own jailers.

The philosopher Michel Foucault seized on this design in 1975, in Discipline and Punish, and argued that the Panopticon had escaped the prison. Its logic — control through the mere possibility of being seen — had, he claimed, seeped into the whole of modern life: the school, the factory, the clinic, the office, all of them arranging people so they could be observed and so, eventually, they would observe themselves. Foucault was writing about buildings and bureaucracies, filing cabinets and timetables. He died in 1984, before the internet, and it is one of the small tragedies of intellectual history that he never saw the tool that would make his metaphor literal — because we have now built the Panopticon out of glass and radio, and we carry it in our pockets, and we paid for it ourselves.

The engine that eats your behaviour

The modern name for the tower was given by the scholar Shoshana Zuboff, who in 2019 called it surveillance capitalism. Her analysis is precise and worth stating plainly, because it is the economic engine underneath every inference machine in Chapter 8. The great technology platforms, she argues, discovered that the traces we leave — our searches, our clicks, our locations, our pauses, our purchases, the milliseconds we linger on one image over another — are a raw material. They called it, in their internal language, "behavioural surplus": more data than any service actually needs to function, harvested anyway. And this surplus is fed to the abduction machines, which infer from it the things we never disclosed — our moods, our health, our sexuality, our politics, our susceptibilities, the moment we are most likely to be persuaded — and those inferences are sold, at auction, to whoever wishes to move us.

This is the Science of Deduction turned into an industry, and notice what has happened to Holmes's method in the transformation. Holmes read one client at a time, for a fee they paid him, to solve a problem they brought him. The machine reads everyone, all the time, for a fee paid by third parties, to solve problems we did not ask to have solved — chiefly the problem of how to make us buy, click, vote, and stay. The eye that reads has been detached from our consent and pointed at us from behind a mirror. And it is very, very good at its job. It is the calibrated, Bayesian, industrial descendant of everything Part II celebrated — which is exactly why it is dangerous. The better inference gets, the less of yourself you can keep.

The right not to be guessed

This forces into the open a kind of privacy that older ideas cannot handle. We used to think privacy meant control over our secrets — the facts we deliberately conceal. But the machine does not need your secrets. It infers them. It does not read your diary; it reads the ten thousand trivial things you never bothered to hide and deduces what the diary would have said. A now-famous illustration, whatever the exact details of its retellings: a retailer's model, reading nothing but shifts in a young woman's purchasing, inferred she was pregnant and began marketing to her — before she had told her family. No secret was stolen. A pattern was read. The pregnancy was an abduction — the best explanation of a change in buying — and it was correct, and it was nobody's business, and no diary had to be opened for her privacy to be stripped.

Scholars have started to call what we need here predictive privacy, or the right not to be inferred about — a protection not for the facts you hide but for the conclusions others draw about you from the facts you can't help leaving. It is a strange, new, necessary idea, and it exists because the science this book celebrates got good enough to require it. There is a chilling second-order effect, too, the same one Bentham designed into his tower: when people know they are being read and predicted, they begin to edit themselves — to avoid the search, the book, the route, the association that might be inferred against them. The reading eye, simply by being known to exist, narrows the range of things a free person feels free to do. That is the Panopticon's oldest trick, running now at the scale of nations.

The same eye that sets the innocent free

And here is where an honest book has to resist its own momentum, because the eye that surveils is the same eye that exonerates, and I will not let you leave this chapter with a comfortable villain.

Consider two cases. In 2005, a serial killer who had taunted police for three decades under the name BTK was undone by a single act of inference. He sent police a floppy disk, believing it anonymous; investigators read its hidden metadata, which quietly recorded that the file had last been saved by someone named "Dennis" at a local Lutheran church. A few minutes of deduction from a digital trace ended a thirty-year reign of terror, and Dennis Rader was caught. In 2018, the Golden State Killer — a rapist and murderer who had eluded capture for over forty years — was identified when investigators uploaded crime-scene DNA to a public genealogy database, found distant relatives, built out the family tree, and closed in on one man. Both cases are Holmesian miracles made real: invisible traces, patiently read, delivering a monster to justice. Both are triumphs of exactly the science this book has spent nine chapters resurrecting. You cannot cheer the exoneration of the innocent by DNA and boo the capture of the Golden State Killer by DNA; it is the same technique, the same reading of the same traces, pointed at the same kind of truth.

That is the knot, and it does not untie. The power to read the world's traces is genuinely one of the great instruments of justice — it frees the wrongly convicted, catches the truly guilty, and does both with a rigour Holmes only dreamed of. And the very same power, in the very same act, is the great instrument of surveillance — it strips the citizen, feeds the market, and builds the tower. There is no version of this science that only exonerates and never surveils, because they are one science. The trace that convicts the murderer is the trace that profiles the shopper. Locard's principle — every contact leaves a trace — was always morally double-edged; it just took us a century, and a machine, to feel the second edge.

The ethic the detective never needed

So what does the resurrected science ask of us, ethically? Precisely the thing that Part II identified as its core and that Part I found Holmes to lack: calibrated humility, made into a public virtue. An inference about a person is a form of power over them, and power over people demands the things Holmes never had to supply — consent, accountability, the right of appeal, and above all the built-in knowledge that the inference might be wrong. Robert Williams sat in a cell because a machine made a confident guess and no one around it had internalised that a guess can fail. The whole moral difference between a science of inference and a tyranny of inference comes down to whether the reading eye remembers it is guessing — whether it carries, into every judgment about a human life, the humble whisper of "Norbury."

Doyle never had to teach his detective ethics, because Holmes's inferences never touched anyone who didn't deserve it, and were never wrong, and served no master but justice. That was the deepest fantasy of all — deeper than the infallibility, deeper than the rigged clues. The real science has no such luck. It is fallible, it is for hire, and it is pointed at everyone. Which means the science of deduction, made real, is not only an intellectual achievement. It is a permanent political responsibility — the ongoing, unfinished work of deciding who may read whom, to what end, with what recourse when the reading is wrong. Holmes never had to carry that weight. We do. It is the price of making the dream come true.

And so we arrive, humbled and clear-eyed, at the end — ready to say at last what the Science of Deduction actually is, now that we have taken it apart and built it back, and what it means to be its practitioner in a century Doyle could not have imagined.