Mind over Matter

“The central novelty across the three studies is a new generation of AI-inspired brain decoding models that, for the first time, allow us to precisely read out how people consciously experience emotions from brain activity,” said research leader Professor Benjamin Becker from the MIND & AI Lab at the Department of Psychology. “This includes fear as well as the intensity of emotional experiences (affective arousal) under highly dynamic conditions close to real life.”

The team combined advanced MRI-based brain scanning with AI-inspired decoding to overcome what he terms a fundamental limitation of traditional approaches. For decades, researchers in hundreds of studies have measured how single brain regions respond to isolated, static stimuli – such as briefly presented pictures of fearful faces. But this approach could not comprehensibly capture how the brain actually constructs thoughts and feelings.

“Our AI-inspired models change that,” said Professor Becker. “They simultaneously read out information from hundreds of brain regions and the communication between them, allowing us to measure what a person consciously experiences.”

In more detail, the technique used combines advanced MRI scanning with AI-inspired decoding and carefully designed psychological experiments that induce immersive and dynamic emotional experiences resembling everyday life. The models learn the relationship between momentary changes in the data – widely distributed brain activity patterns and the communication between brain regions – and link these to what the person actually reports feeling.

Combination is key

“This combined approach is key,” said Professor Becker. “There have been previous models for tracking fear and other emotions from brain activity, but we found that every single one of them failed when it came to fear experienced in more realistic, life-like contexts. Our models are the first to succeed in this.”

The new models provide objective, brain-based measures of conscious emotional experience, which opens up new strategies for diagnosis and for testing new treatments.

“These advances are important beyond technical progress or scientific debates,” said Professor Becker. “Emotional dysregulation is at the heart of the most common mental health disorders – that is, depression and anxiety – which represent the leading cause of disability and suffering worldwide.”

Among these disorders, social anxiety disorder is one of the most prevalent mental health conditions – roughly one in 13 people will suffer from it during their lifetime. It is characterised by intense fear in social situations that goes far beyond ordinary shyness, triggering overwhelming distress and leading people to avoid the social interactions that are essential for a fulfilling life.

One of the new strategies with a hypothesised efficacy in reducing anxiety is the neuropeptide oxytocin, and Professor Becker’s laboratory tested this hypothesis.

“There have been no fundamentally new pharmacological treatments for anxiety disorders in decades,” he said. “We wanted to advance treatment development and tested whether the neuropeptide oxytocin can reduce fear specifically in social contexts. It did – Oxytocin selectively reduced subjective fear in social situations but not in non-social ones, an effect we could precisely track at the brain level using our AI-inspired models.”

This was exactly the profile the team was hoping for: they would not want to simply switch off all fear, given that fear is vital for survival, but to specifically reduce it in social situations when it becomes disabling.

Social fear

“We are currently testing oxytocin in mental disorders, including autism, where social fear is a core challenge,” said Professor Becker. “More broadly, our AI-inspired brain models can accelerate treatment discovery and evaluation – for instance by objectively testing whether a new treatment truly changes conscious fear experience in the situations that matter most to patients.”

Asked why it is important to specifically measure the subjective experience of emotions such as fear, Professor Becker said: “Emotions include our subjective experience but are always accompanied by strong bodily responses – your heart races when you are afraid. But these bodily reactions are not really specific to any particular emotion. Think of two people on a rollercoaster: one is terrified, the other is thrilled – they share the same racing heart, yet their subjective emotional experience is completely different.

“For over a century, psychology has debated whether these automatic bodily reactions can truly be separated from the conscious feeling itself. Our technological advances now allow us to demonstrate that the brain basis of subjective feelings is indeed distinguishable from the more general bodily responses.”

“This also carries important implications for current debates about AI and consciousness: one of the most fundamental conscious experiences – emotional arousal – does not appear to require bodily input. This bears critical implications for debates on whether AI – lacking a body and physical presence – can develop consciousness. To take these implications forward, we have launched a new Strategic Research Theme ‘AI, Society & Social Dynamics’ at the Faculty of Social Sciences.”