What would happen to human interaction if AI could read our emotions?
Artificial Intelligence image: iStock
Philosopher Alexandra Prégent has explored how emotion recognition technology could change the way we interact with one another. In her dissertation, she argues that protecting our emotional privacy is essential for a functioning society.
AI systems are increasingly being developed to recognise emotions by analysing facial expressions, voices, body language and other physiological signals. Although the scientific validity of many of these systems remains contested, they are already being tested in healthcare and security settings.
Rather than focusing on the technology’s current shortcomings, PhD candidate Prégent asks a different question: what if these systems work exactly as intended? ‘Companies are showcasing all the benefits of these technologies’, she says. ‘As philosophers and scholars, it’s our job to wave a flag and show what could be at stake if they become reliable.’
When technology starts making decisions
To explore the implications of fully functioning emotion recognition technologies (ERTs), Prégent developed several future scenarios. One involves a surgeon arriving at a hospital to perform an operation. Before she enters the operating theatre, an ERT system flags her as emotionally unstable. She is taken aside by her employer and prevented from operating.
‘The technology forces her to expose something that belongs to her private life’
The surgeon insists she is perfectly capable of carrying out the procedure. The reason for the alert has nothing to do with her professional skills. She is going through a difficult divorce and is worried about one of her children. Although she deliberately left those concerns at home, the technology has interpreted her emotional state as a reason to question whether she can do her job.
‘The technology forces her to expose something that belongs to her private life’, Prégent says. ‘She has to explain what is happening at home and justify why she believes she can still perform the operation. But the system has already judged her emotional state and a decision has been made.’
For Prégent, the example illustrates a broader societal risk. People occupy many different social roles throughout their lives: they are surgeons, but also parents, partners, friends and neighbours. They usually decide for themselves which aspects of those roles they bring to a particular situation.
Changing the rules of social interaction
ERT, Prégent argues, could blur those boundaries. If people can no longer hide their inner emotions, the distinction between their private and professional lives is at risk of collapse. Instead of deciding for themselves what they reveal, people may find that technology exposes parts of their emotional lives and forces them to explain or defend them.
Prégent sees similar tensions beyond the workplace. She imagines a government representative delivering a public apology after a fatal incident. However, ERT indicates that they do not feel genuine sadness or regret. Instead of listening to the apology, the public focuses on the mismatch between the emotions being expressed and those the technology claims the speaker is feeling.
‘Public apologies are necessary, but they can’t always be accompanied by genuine emotions, if only for self-protection’, she says. ‘Take police officers, doctors or politicians who regularly deal with traumatic events. They can’t relive those emotions every time they address the public.’ That does not make the apology meaningless, she argues. A public apology is also a social ritual: it acknowledges what has happened, expresses responsibility and explains what will happen next.
According to Prégent, ERT could fundamentally change how we interpret one another. ‘Instead of relying on what people say and do, we may increasingly rely on what technology tells us they feel.’
Preparing for the future
The European AI Act already restricts some uses of emotion recognition technology, but this is largely because today’s systems are unreliable. Prégent’s research takes a longer-term perspective. If the technology becomes significantly more accurate, society will need stronger ethical arguments for deciding when it should – and should not – be used.
‘Protecting emotional privacy is not only about protecting information’
Prégent emphasises that ERT is not inherently negative. It could, for example, improve the early detection of neurological disorders or help some people recognise emotions. But potential benefits should not prevent society from considering the broader consequences. ‘Protecting emotional privacy is not only about protecting information’, she says. ‘It’s about protecting the social space in which people can decide who they are, how they present themselves and how they relate to others.’
Whether ERT will reach its full potential remains uncertain. But according to Prégent, now is the time to think about what kind of society we want to live in before the technology and companies make these decisions for us.
PhD defences by researchers from Leiden University can be watched live online. Alexandra Prégent will defend her dissertation entitled Emotional Privacy: A Philosophical Investigation of Emotion Recognition Technology on Wednesday 2 September, from 11:30 to 12:30 hrs.