The Erasure of the Outlier

This week, I joined a room in Durham as a digital shadow.

Through the miracle of a pre-recorded video stream, I acted as a “witness” to the DIRDI AI conference. There is a specific, jagged irony in speaking about the “Privatisation of Being Human” via an MP4 file - my face reduced to pixels, my voice compressed into an algorithmic thread - arguing fiercely for the parts of us that refuse to be compressed.

I wasn’t there to provide the industry-sanctioned optimism that usually greases the wheels of these events; I was there to challenge the very vocabulary of the room. I was also, quite literally, not in the room. That’s the messy reality of life with lupus - it turns you into a signal in the noise, a reminder that the human body is the one 'outlier' no algorithm can ever truly smooth away.

Currently, our discourse is obsessed with “efficiency.” In the neoliberal university and the tech-sector alike, efficiency is treated as a moral good. But we must remember: efficiency is often just a faster way to do the wrong things.

What we are actually building is a Global Averaging Machine. This is a technology that mistakes statistical probability for scientific truth. It is a system that looks at the bell curve of human experience and decides that the “mean” is the only thing worth replicating.

As someone who has spent a career advocating for neurodivergent minds and equity in education, my concern isn’t that AI will fail. My concern is that it will succeed in making us all “average.”

Presently, we are building systems that function as high-speed sandpaper, designed to smooth the ‘human’ out of the process entirely. We are rubbing away the spiky profiles of genius until we are left with a sterile, predictable surface.

Being a Professor carries a certain institutional weight - a high rank that usually demands a 'visionary' answer. But when I'm asked about 'opportunity,' I refuse the standard script of more, faster, bigger. My answer is not 'more data!

The real opportunity is to use AI as a High-Contrast Mirror.

AI reflects back to us exactly where our systems have failed to account for human messiness. It shows us the structural gaps we’ve ignored for decades:

• The missing gender data in medical research.

• The neurotypical bias baked into our pedagogy.

• The invisible labour that sustains our institutions.

AI is taught by a global underclass of click-workers - a massive shadow-force of invisible human labour performing the digital drudgery required to make the machine appear 'intelligent.' This is the hidden friction we pretend doesn't exist.

The opportunity is to use the machine to unmask our own blind spots, not to automate the ones we already have.

However, the shadow side is what I call Cognitive Mono-cropping. Science has never moved forward because of the “most likely” outcome. It moves forward because of the outlier - the person who sees the pattern that shouldn’t be there.

If we allow AI to gatekeep peer reviews, grant funding, staff recruitment, or student admissions, we are effectively sterilising the scientific method. We are creating an Epistemic Enclosure where truth is defined as that which is most statistically probable.

The conversation at DIRDI turned to “National Security.” In a week where the headlines are dominated by the tragic movement of physical borders and the reality of kinetic war, it might seem counterintuitive to talk about data.

But the most vulnerable border we share right now isn’t geographic—it is the sovereignty of the human mind.

The real security threat is the Privatisation of Being Human. When the infrastructure of “truth” - what we see, what we believe, how we perceive “the enemy” - is owned by a handful of offshore corporations in California, the social contract dissolves. “National security” becomes a shadow when a citizen can no longer discern an algorithmic hallucination from a state mandate.

From an equity standpoint, this is where it becomes dangerous. If our security systems are built on “normative” data, then being neurodivergent or “atypical” - being someone like me - is no longer just a difference.

It is flagged as a security risk.

We are building a digital panopticon that treats “different” as “deviant.” This is the algorithmic policing of the human spirit. It is systemic exclusion disguised as “safety.”

Durham has always been a place for deep, often difficult inquiry. My challenge to my colleagues - and to you - is to refuse the efficiency of the easy answer.

In your labs, your classrooms, and your policy-making, look for the friction. Look for the “noise” in the data and the “difficult” person in the room. Because that texture - that refusal to fit the model - is exactly what it means to be human.

In my new role as Co-Director for the Being Human theme at the Leverhulme Centre for Algorithmic Life (launching April 1), this will be our mandate. We are not here to help the machine “understand” us better. We are here to protect the right to be un-optimised.

Science lives on the fringe. If we lose the outlier, we lose the breakthrough.

Don’t let the AI sterilise your science.

Protect your outliers. Protect the noise. Because whether it’s a scientific breakthrough or a body that refuses to comply with a 'normative' schedule, the friction is where the truth lives. Don't let the AI sterilise your science - or your humanity

As a post-script to the conference, I received an email from a mathematician named Jeff who co-directs an AI research hub at Oxford. He spoke about the need for his technical team to “lift their eyes up” and look at these broader issues. It is a small signal in the noise, but it suggests that even in the world of pure equations, there is a hunger for the messiness of being human. I’ll be heading to Oxford later to continue the provocation.