Building AI Systems Worthy of Human Trust

Severe burnout, a near-death experience, and years inside a toxic work environment all collided into a personal and professional existential crisis that dismantled everything I thought I knew. On the night I ended up in hospital, I received an email telling me a paper had been accepted without revisions in the Journal of Positive Psychology. The irony was not lost on me. Here I was, one of the people writing about the science of flourishing was unable to keep himself alive. I realised something drastically had to change.

In the months that followed, I turned to every wellbeing framework, every tool, and every technique I had spent my career building and studying. And none of it worked. None of it could help me get out of this existential crisis I was in. One day, my therapist told me that the best way to deal with this was to connect more deeply with nature. Ok, I guess that made sense because all scientific evidence pointed to the fact that exposure to nature is one of the strongest predictors of meaning and meaning one of the strongest predictors of wellbeing. But going outside made me miserable. Why? Because I hate being dirty and being in nature makes me dirty! So in that absurd, humbling moment, I understood something that changed the trajectory of my entire career: we built a science of wellbeing for people who don't exist. Our models are built on population averages, but the average person is a statistical ghost. What works “in general” can be precisely wrong for the individual sitting in front of you.

That realisation broke me… but it also rebuilt me.

But what I found when I looked more broadly across the AI space alarmed me. Systems designed to support mental health and measure wellbeing are being built by engineers with no grounding in psychological science. They approximate flourishing through engagement metrics. They make claims they cannot validate. They create dependency and call it support.

Meanwhile, psychologists are confusing fluency with competence, mistaking the ease of using a tool for the expertise required to wield it responsibly. Both sides are causing harm, and neither sees it.

AI systems built without psychological grounding are not just ineffective. They are causing harm at scale. That is the problem I now work on.

I work with AI system designers, engineers, developers, and organisations to build AI systems that are psychologically safe, scientifically grounded, and designed to strengthen human agency rather than erode it. The question I ask of every system is simple: does this help people take control of their own lives, or does it take control from them? I am here to make sure we get this right, because we will not get a second chance to earn people's trust.