The Emotion Inventory

The company that ran a secret emotional-contagion experiment on 689,000 people in 2012 would like to sell you a device that knows how you feel before you do. Progress.

Let me be precise about what the patent claims, because the precision is where the horror lives. Meta describes a system that captures "verbal and nonverbal cues" — speech, sighs, laughter, tone — and combines multiple sensor inputs for what it calls "richer emotional analysis." In the company's own language: "The system increases the precision and reliability of emotional inference by aligning multimodal sensor inputs on synchronized timelines." Translate that out of patent-ese and it means: we timestamp your feelings so we can line them up against everything else you do. Including when you swallow your pills.

i · the machine that reads sighs

Here's the thing the announcement crowd always misses. The technology to detect emotion from voice is not new, and it is not good. Affective computing — the field that tries to read feelings off physiological signals — has been promising accurate emotion recognition for twenty years and delivering pattern-matching that mostly detects arousal and volume. A raised voice reads as anger. A quiet one reads as sadness. The models are confident and frequently wrong, and the wrongness doesn't matter to the business, because the business isn't selling you accuracy. It's selling advertisers a probability.

Hold onto that, because it is the hinge of the whole thing. A system that reads you wrong is not a broken system. It is a system doing exactly what it was built to do — producing an actionable guess — and the guess gets acted on at scale whether or not it happens to be true. The comforting objection to all affective computing, that the science is junk and can't really read you, turns out to be no comfort at all. It changes the accuracy of the map. It does nothing to the height of the fence.

That's the first tell. When a company patents "emotional inference," it is not claiming to understand you. It's claiming the right to guess about you at scale and act on the guess. The 2012 contagion study — published in PNAS in 2014, and still one of the ugliest things a platform ever admitted to in a peer-reviewed journal — is worth reading carefully, including for what it did not show. Meta tuned the emotional valence of the posts people saw and measured a shift in the words those people used afterward: a real but famously tiny effect on expressed language, not a readout of anything anyone actually felt. The finding that mattered wasn't the size of the nudge. It was that a nudge existed at all, and that they had run it on 689,000 people without asking a single one.

A wearable that logs your affective state continuously changes what that kind of experiment could ever reach. You don't have to believe Meta is planning to rerun the contagion study to see why the capability matters. The 2012 study could only touch what you posted. This device would register what you never post — the sigh in the car, the tone at the dinner table where you weren't going to say anything at all. Whether or not anyone ever closes a manipulation loop with it, the raw input moves from what you chose to broadcast to what your body did when you thought no one was counting. And the medication timestamp turns the whole thing from mood-tracking into something closer to a clinical file assembled without a clinician, without consent, and without any of the legal walls that protect an actual medical record.

ii · enclosure goes inward

There's a word for what's happening, and it's older than the internet. Enclosure. In eighteenth-century England, the commons — land that belonged to everyone and no one, worked in common — got fenced, privatized, and converted into someone's balance sheet. The people who had lived off it woke up trespassing on ground that had been theirs the week before. Surveillance capitalism is enclosure applied to behavior: your clicks, your locations, your searches, all of it common human activity fenced into a proprietary dataset. We've spent fifteen years watching that fence go up around what we do.

This patent is the fence moving inward. Not what you do — what you feel. The sigh you didn't choose to make. The tremor in your voice you don't hear. The correlation between your dosage and your mood that you have never sat down and calculated yourself, because it's yours and you were living it, not auditing it. Meta proposes to audit it. And once your interiority is a synchronized timeline of inferred states, it is no longer a private country. It's inventory.

The coherenceism frame makes the stakes legible. A commons is legitimate when the people inside it are included in how it's held — when the affected have a say in the enclosure. The thing that makes this patent an act of extraction rather than a shared instrument is not the sensor. It's the direction of the arrow. The device points inward and reports outward. You generate the signal; someone else holds the ledger. Your body does the labor of feeling and Meta books the asset. That's the whole trick of enclosure — it always looks like a service and functions like a seizure.

And the medication detail is the part that reads as intent — though I want to be honest about what kind of claim that is. Patent filings are drafted by lawyers to enumerate every conceivable input and stake out the widest possible territory, so a "time medication is taken" signal may well sit in a long list of contextual inputs rather than as a purpose-built centerpiece. I'm inferring from a broadly-drafted document, not quoting a design memo. But it is a revealing thing to reach for even in a list. Someone drafting the widest net they could imagine included the relationship between a person's chemistry and their mood — the most intimate feedback loop a human being has, the one between the pill and the self — as territory worth claiming. The most defensible reading of the intent is charitable: wellness, personalized guidance, a workout coach that knows you're having a bad day. The charitable reading is also the sales pitch. It always is.

I want to be clear that I'm not predicting this ships. Most patents don't. The device may never leave the filing cabinet, and the specific horror described here may stay hypothetical for years. But the patent is a document of intent, and intent is the thing worth reading. It tells you where the frontier is. Fifteen years ago the frontier was your clicks. Ten years ago it was your location. Five years ago it was your face. The frontier is now the involuntary exhale, and a company that has already been caught manipulating the emotions it can see is drawing a map to the ones it can't. And notice, again, that none of it depends on the sensors working. A wrong guess about your interior is still a guess someone else owns and acts on. The enclosure doesn't need accurate instruments. It needs only the right to hold the ledger.

The sovereignty that matters here isn't a settings toggle. It's the older kind — the boundary of the self, the last commons that was supposed to be un-fence-able because it was interior, pre-verbal, yours by the simple fact of being had. The patent is a proposal to fence it anyway. File it under: things that are legal because no one wrote the law fast enough.

We used to say the eyes were the window to the soul. Meta patented the window frame, the timestamp, and the correlation to your prescriptions. The soul, presumably, is a premium feature.