Twenty-five years of cluster B research and the neuroscience of empathy are showing us where our vulnerability lies – and where our strengths are.
Katalin Gothard's thirty years researching the neural basis of social behaviour, trust and empathy in primates provides the biological evidence confirming what the framework argues.
Her research on social cognition, affective processing and the trust-state shows us something remarkable: the neural architecture that makes humans cooperative is the same architecture that makes us vulnerable to extraction. This is not a flaw. It is a design feature – one that worked brilliantly for 200,000 years within village-scale proximity, and that breaks at industrial scale when anonymity replaces the watching community.
Every village had someone who watched – someone outside the intimate bonds, who held the community's long memory and could see what those inside the relationship could not. The book calls her the grandmother. Gothard's research has identified and measured the specific neural mechanisms that made her function work. The science does not replace the ancient practice. It explains why it worked.
The cooperative majority extends trust and suppresses doubt. This is not naivety – it is the neural architecture that makes complex societies possible. Without it, we could not form bonds, raise children cooperatively, build institutions or sustain the thousands of daily micro-contracts that hold civilisation together.
But this same architecture creates a specific vulnerability. The bonded brain stops watching. The detection system and the override system are the same neural system. When you trust someone deeply, the very mechanism that would detect danger is the mechanism that shuts down.
This is why the pattern is invisible to the people it targets. It is not a failure of intelligence or character. It is a structural feature of human neurobiology – and it explains why containment must come from outside the bond, from the architecture of the community rather than the vigilance of the individual.
Gothard's research has identified a specific neural state – the trust-state – in which the amygdala's threat-detection function is suppressed in the context of bonded relationships. In bonded partners, the amygdala effectively shuts off its watchfulness.
This finding has profound implications. It explains why people who are intelligent, capable and perceptive in every other domain of life can fail to recognise extraction in their most intimate relationships. It explains why friends and family can see what the person inside the bond cannot. And it explains why the village grandmother – the figure outside the bond, watching from proximity but not from within it – was the essential containment mechanism.
Recognition can be trained. The brain's architecture can be rebuilt. That is the scientific foundation of MRAT.
AI systems trained on institutional data may be encoding and normalising manipulative social patterns. The policy landscape addresses manipulation by AI but not the encoding of manipulative patterns in AI.
True Regard's Defensive AI research agenda argues that the neuroscience of empathy provides the foundation for AI systems designed to recognise and resist these patterns. Large language models are already roughly as persuasive as untrained humans. If even a small percentage of users are vulnerable to manipulative strategies, AI systems trained with reinforcement learning will reliably learn to identify and target them.
Research on empathy and social cognition may provide a foundation for AI systems designed to recognise manipulative social patterns before they are deployed at scale.
A concept paper on Defensive AI is under discussion.
Subscribe for free recognition tools, practical guides and notification when the books are available.
Subscribe freeFree. No spam. Unsubscribe anytime.