There's a camera looking at your face right now, or there will be soon, and it isn't just checking who you are. It's deciding how you feel. Happy or hostile. Engaged or bored. Truthful or lying. And then it acts on that verdict — sells to you, scores you, flags you — without ever asking whether it read you right.
This is Emotion AI, sometimes dressed up as "affective computing." The pitch is seductive: machines that understand feelings. The reality is colder. It's a pseudoscience of the face, wired into a surveillance stack, sold to advertisers, employers, and police as if a smile were a fact you can measure.
The global market for this was already valued in the low tens of billions of dollars and projected to more than double within the decade. That money is not buying empathy. It's buying the claim that your inner life is now legible from the outside — and legible to whoever owns the camera.
The science underneath is shaky — and they shipped anyway
Start with the part they don't put in the sales deck.
The whole industry rests on a claim from the 1970s: that a small set of "basic emotions" produce universal facial expressions. Smile means happy, everywhere, always. In 2019 a landmark review led by psychologist Lisa Feldman Barrett — surveying over a thousand studies — demolished it. People scowl when they're not angry and don't scowl when they are. Expressions vary by culture, context, and individual. You cannot reliably read emotion from a face. The paper's conclusion was blunt: the tech companies are selling a capability the science does not support.
They shipped it anyway. Because the product was never accuracy. The product was the appearance of insight, and a number to put in a dashboard.
So the machine watches your micro-expressions, your tone, your gaze, your heart rate through a webcam, and outputs a confident label: 82% engaged, 71% honest, elevated aggression. The confidence is manufactured. The label is a guess wearing a lab coat. But it enters a hiring decision, a loan file, a police report — and there it hardens into a fact that follows you.
Where they point the camera
In advertising, brands wire up your webcam or your phone's front camera to measure your reaction to an ad, second by second, and A/B test which emotional trigger converts you. Your face becomes a focus group of one, unpaid and often unaware.
In hiring, video-interview systems have scored candidates on facial "signals" and vocal tone to rank them for jobs. One prominent vendor, HireVue, marketed facial analysis for exactly this — and in 2021, after an FTC complaint and public pressure, dropped the facial component. The demand didn't die with it. The idea that a webcam can rank your character is still being sold under new names.
In policing and the state, the ambitions are darkest. Vendors have pitched "aggression detection" for public spaces and "deception detection" at borders — the EU trialed an automated lie-detector called iBorderCtrl that read travelers' faces. China has deployed emotion-recognition tech in security contexts, including reports of its use against Uyghurs. This is the face read as a threat score, by a machine that cannot be cross-examined.
The face is a mask — and they forgot
Here is where the old knowledge cuts clean.
Egypt understood the face as a threshold, not a readout. The heart — ib — was the seat of thought, will, and truth, and it was hidden. At the final weighing, the heart is set on the scale against the feather of Maat precisely because only there, in the sacred court, can its truth be known. Not from the outside. Not by a guard. Not by a camera. The interior was sacred because it was sovereign — yours, and unreadable by force.
Emotion AI is the crude inversion of the weighing. It claims to read the heart from the surface, in a fraction of a second, for profit, without consent, and to pass judgment on what it finds. It mistakes the mask for the ib. And then it acts on the mistake — denies you the job, flags you at the border — as if its guess were the verdict of the scale.
Our record: In the order of Maat, only the true court may weigh a heart, and only against the feather, and only with the soul present to speak for itself. A machine that weighs your interior without your voice, without the scale, without the truth — and then bills or jails you on the result — has stolen the function of the court. That is Isfet in a lab coat: judgment without justice, verdict without trial. The sovereignty of your inner life is the first thing they need you to forget you had.
The lever: keep your interior sovereign
Never doom without a door. The pseudoscience is the weakness — attack it, and the whole tower wobbles.
- Refuse the reading. Cover the webcam when you're not using it. Decline the "emotion-analyzed" video interview and say why. Push back on any app that wants your camera to "understand" you. Consent is the load-bearing beam; pull it out.
- Cite the science, loudly. When a vendor or employer claims to read emotion, the answer is short: the research says you can't. Barrett's review is your weapon. A claim that dies under scrutiny should be forced to face scrutiny.
- Demand a ban where it's dangerous. The EU's AI Act already moved to prohibit emotion recognition in workplaces and schools. That's the template. In hiring, in policing, in the classroom — the correct setting for this tech is off.
- Guard the mask. Your face is not a public API. Your mood is not telemetry. Treat every request to scan your affect as what it is — an attempt to read a heart that only you have the right to open.
The weighing belongs to the true court alone. Let the machine stare. It sees a mask and calls it a soul — and the moment you name that error out loud, its verdict has no power to bind you.
Name the pseudoscience. Cover the lens. Keep your heart off their scale — it was never theirs to weigh.