Blade Runner Taught Us to Fear the Emotion Detector. In January, California Bans It at Work

Futuristic cyberpunk office with an emotion-detection scanner analyzing an employee under neon rain, inspired by Blade Runner.

Written by

in

On November 25, Prime Video will release all eight episodes of Blade Runner 2099 at once. The series jumps fifty years past Blade Runner 2049, into a Los Angeles where replicants, not humans, now run the show. Michelle Yeoh plays Olwen, a replicant who works as a Blade Runner, hunting other replicants, in the final days of her life. Hunter Schafer plays Cora, a human who has to pretend to be a replicant Blade Runner just to avoid capture.

Read that last part again. In this future, passing as something you are not is a survival skill. And the oldest question in the franchise, who gets to decide what is real and what is only performance, suddenly has a very human face.

It is also a question that California lawmakers just answered in real life.

The scene that started it all

The test everyone remembers is the Voight-Kampff, and it opens the original 1982 film. A Blade Runner named Holden sits across from a worker named Leon at the Tyrell Corporation, a machine trained on Leon’s eye, and fires off questions built to provoke a reaction. The theory is simple: replicants can imitate emotion, but they cannot truly feel it, so the body gives them away. The twist, as Lovins notes, is what trips Leon up: not a lack of empathy, but the moment Holden asks about his mother and he has no memories to offer. He answers by attacking Holden.

The theory is tested again later, when Deckard examines Rachael. Deckard says it normally takes about thirty questions to identify a replicant; Tyrell remarks, a bit smugly, that with Rachael it took more than a hundred. According to Lovins, geographers Marcus Doel and David Clarke (1997) observed what that scene implies: there is no number of questions after which the test can declare someone human. It can only confirm that someone is a replicant, or that they have not yet been proven to be one. If a replicant’s response is convincing enough, the test stops measuring feeling and starts measuring performance.

Then someone actually built it

In 2016, a team at Lancaster University led by design-fiction researcher Paul Coulton presented a paper at CHI, the major human-computer interaction conference. They built a device inspired by the Voight-Kampff machine, an emotion detector, imagined for online dating, that would measure skin response, heart rate and pupil dilation. The point was not to sell a gadget. It was to use fiction as a lens and ask whether this is a world we actually want.

Literary scholars have long read the franchise the same way. Christopher Lovins, writing in the Journal of Science Fiction in 2019, compares Blade Runner with Ghost in the Shell and argues that science fiction has grown more comfortable with blurring the line between human and machine instead of defending it.

The problem is the science

Here is the twist the movie never needed: the real-world version of the test may be built on shaky ground. In 2019, psychologist Lisa Feldman Barrett and four colleagues published a major review in Psychological Science in the Public Interest, a review the authors describe as the most comprehensive and systematic to date, covering facial movements and emotion. People do sometimes scowl when angry, more than chance would predict. But the authors found substantial variation across cultures and situations, and the same movements can express more than one emotion, so facial movements are not reliable, specific signals of any single emotion. A meta-analysis cited in the review found that the supposedly typical facial configurations for each emotion showed up, on average, in only about 19% of emotional episodes. And the same movement, such as a scowl, often communicates something other than an emotion.

The authors warned that this idea already guides commercial applications, security screening, children’s education and court decisions. In other words, the fictional Voight-Kampff and its real-life descendants arguably share the same flaw: they read a signal from the body and call it a feeling.

California says no

On September 30, Governor Gavin Newsom signed AB 1883. Starting January 1, 2027, the law bars California employers from using AI-powered workplace surveillance tools to recognize, infer or predict an employee’s emotional state, or to collect neural data. It carves out tools used for safety, and it is deliberately narrow. Newsom himself noted when signing that more work remains, since the law leaves some key terms undefined and may create confusion about which tools are covered.

California is not alone. The EU’s AI Act has prohibited emotion recognition in the workplace since February 2025, with exceptions for medical and safety reasons.

Who is the replicant now?

Back to Cora. In Blade Runner 2099, a human survives by acting like a machine that acts like a human. The premise suggests that when a test decides who gets hunted, the test becomes the most dangerous thing in the room, especially when it is wrong.

For decades, Blade Runner asked what it means to feel. Starting in January, one state will ask a more practical question: should your boss ever be allowed to guess?

Sources