Tech companies claim AI can recognise emotions, but the science doesn't stack up

Emotion recognition technology poses a range of legal and societal risks - especially when deployed in the workplace.

- Natalie Sheard

Can artificial intelligence (AI) tell whether you're happy, sad, angry or frustrated?

According to technology companies that offer AI-enabled emotion recognition software, the answer to this question is yes.

But this claim does not stack up against mounting scientific evidence.

What's more, emotion recognition technology poses a range of legal and societal risks—especially when deployed in the workplace.

For these reasons, the European Union's AI Act, which came into force in August, bans AI systems used to infer emotions of a person in the workplace except for "medical" or "safety" reasons.

In Australia, however, there is not yet specific regulation of these systems. As I argued in my submission to the Australian government in its most recent round of consultations about high-risk AI systems, this urgently needs to change.

A NEW AND GROWING WAVE

The global market for AI-based emotion recognition systems is growing. It was valued at US$34 billion (S$45.8 billion) in 2022 and is expected to reach US$62 billion by 2027.

These technologies work by making predictions about a person's emotional state from biometric data, such as their heart rate, skin moisture, voice tone, gestures or facial expressions.

In 2025, Australian tech start-up inTruth Technologies plans to launch a wrist-worn device that it claims can track a wearer's emotions in real time via their heart rate and other physiological metrics.

Truth Technologies founder Nicole Gibson has said this technology can be used by employers to monitor a team's "performance and energy" or their mental health to predict issues such as post-traumatic stress disorder.