The Rigged Game: Why AI Recruitment Is Failing Neurodivergent Talent
In a striking clip from BBC3’s Computer Says No, Nat Hawley, founder of Divergent Thinking, makes a sharp and urgent point: the modern recruitment process is often sold as faster, smarter and fairer, but for many neurodivergent candidates, it is doing the opposite. Instead of widening access, AI-led hiring tools can quietly amplify exclusion.
The problem is not only that technology is being used in recruitment. It is that too much of it is being built around narrow assumptions about what talent looks like, how competence should appear, and how people are supposed to think, communicate and respond under pressure.
You can watch the snippet here: BBC3 “Computer Says No” Nat Hawley snippet
The promise of fairness can hide a new kind of bias
AI recruitment tools are often marketed with the language of objectivity. They are meant to reduce human bias, speed up decision-making and create more consistent processes. On the surface, that sounds like progress.
But Nat’s point in this clip is that consistency is not the same as fairness.
If an assessment tool is built around one narrow cognitive style, then it may consistently disadvantage anyone who thinks differently. In that case, the system is not becoming neutral. It is becoming automated bias at scale.
That matters because neurodivergent candidates are often already navigating processes that over-reward speed, conventional communication style, eye contact, working memory, verbal fluency or a certain type of linear reasoning. When those preferences get embedded into software, the exclusion becomes harder to spot and easier to defend.
Neurodivergence is variation, not error
A key part of Nat’s message is clarifying what neurodivergence actually is.
Whether someone is dyslexic, dyspraxic, autistic or ADHD, they are not simply failing to meet a universal standard. They are showing a different pattern of processing, responding and interpreting information compared with the neurotypical majority.
That sounds obvious, but it matters deeply in recruitment.
If hiring systems treat variation as noise, or treat cognitive difference as a performance problem before the person has even been given a fair chance, then neurodivergent applicants are not being assessed on ability. They are being filtered through someone else’s model of “normal”.
This is one reason Divergent Thinking focuses so strongly on challenging narrow workplace assumptions. A lot of exclusion happens not because organisations actively reject difference, but because they quietly design systems that cannot interpret it properly.
“Gamified” testing is not as neutral as it looks
One of the most revealing points in the snippet is Nat’s critique of so-called gamified recruitment tests.
These kinds of assessments are often presented as modern, engaging and data-driven. But the format itself can still privilege one kind of thinker over another. Nat’s example of a puzzle-based mental maths task is a simple way of showing the problem: once you build an extra layer of cognitive demand into a task, you are not only testing the outcome. You are testing whether someone processes the route to the answer in the specific way your system expects.
That is a serious issue.
A person might be excellent for a creative, strategic or technical role and still perform less well in a test that rewards one narrow style of linear puzzle-solving. If the employer mistakes that test performance for overall capability, then the system is not identifying talent well. It is screening it out badly.
This is where AI recruitment often becomes most dangerous: it creates the illusion that because a process is data-based, it must also be valid.
The hidden disadvantage is often invisible to the employer
A particularly powerful part of Nat’s argument is that the disadvantage is often built into the design of the test itself, not just into the candidate experience.
That means employers may never realise what they are missing.
A neurodivergent candidate might underperform in an assessment not because they lack skill, but because the assessment demands:
rapid processing in a narrow format
comfort with unnecessary ambiguity
a specific social presentation
fluent verbal performance under pressure
a cognitive route that has little to do with the job itself
If employers do not question the structure of the process, they may assume the outcome is fair simply because everyone took the same test. But fairness is not sameness. A process can be identical for everyone and still be deeply unequal in how it operates.
That is why inclusive hiring has to look beyond standardisation and ask a more important question: what is this test actually measuring?
Automated video interviews raise even deeper concerns
The snippet also highlights one of the most troubling developments in recruitment: automated video assessment.
This is where the risks become even more obvious.
If an AI system is trained to score candidates based on cues like eye contact, speech fluency, facial expressiveness or social delivery, then it is making assumptions that may have very little to do with job performance and a great deal to do with neurotypical norms.
Nat gives examples that make this especially clear:
a dyspraxic candidate may struggle with fluent speech under pressure
an autistic candidate may not use eye contact in the expected way
a technically brilliant applicant may be flagged as weak because they do not perform social confidence in a conventional format
That is not fair assessment. It is a category error.
A candidate can be excellent at coding, systems thinking, analysis, design or problem-solving and still perform differently in a social metric that has nothing to do with those strengths. If an AI system rules them out on that basis, then the organisation is not only excluding talent. It is doing so for reasons that are largely irrelevant to the job.
When “social competence” becomes a false gatekeeper
This is the deeper warning in Nat’s BBC3 appearance.
Too much recruitment still treats social performance as a proxy for employability. AI tools can make that problem worse because they encode those assumptions into a system that looks objective from the outside.
But many roles do not require the kind of polished, fast, socially normative behaviour these systems reward. And even when communication matters, that does not mean one narrow presentation style should be treated as the gold standard.
This is one reason Divergent Thinking’s workplace assessments are so useful in practice. They help employers move away from assumption-led judgements and toward a clearer understanding of what actually creates disadvantage, what the role genuinely requires, and what support or redesign would make the process fairer.
The cost is not only ethical — it is strategic
Nat’s critique is not only about inclusion in a moral sense. It is also about loss.
When a company allows poorly designed AI systems to exclude neurodivergent candidates, it is not just creating unfairness. It is losing people who may have been:
highly creative
deeply focused
technically brilliant
systems-oriented
loyal
original in how they solve problems
That loss is often invisible because the person never reaches the point where their strengths can be seen.
The company may believe it is improving efficiency. In reality, it may be narrowing its talent pool, reducing cognitive diversity and weakening its own capacity for innovation.
That is why this issue matters far beyond accessibility language. It is about whether organisations are unintentionally purging difference from their workforce while convincing themselves they are modernising.
A black box is still a box
One of the most dangerous things about AI recruitment is that it can make bad decisions harder to challenge.
When a human interviewer makes an obviously poor judgement, there is at least a visible interaction to examine. But when a candidate is rejected by an opaque system, the reasoning may disappear behind a black box. The result feels neutral because nobody can see the assumptions clearly enough to argue with them.
That is a major problem for neurodivergent applicants.
If a candidate is screened out because of eye contact, pacing, speech pattern, response time or test structure, they may never know that was the issue. The employer may not know either. The software simply outputs a decision, and the exclusion becomes tidier, faster and less accountable.
Nat’s warning is so important precisely because it exposes that false neutrality.
The answer is not blind faith in “better AI”
It is tempting to respond to these concerns by saying the solution is simply better technology.
But Nat’s message points somewhere deeper. The core issue is not only the sophistication of the tool. It is the assumptions sitting underneath it.
If employers continue to design hiring around narrow ideas of competence, then “better AI” may just become a more polished way of enforcing the same limitations. What is needed is a more human-centred understanding of talent — one that recognises that different brains process information differently, communicate differently and show ability differently.
That means asking:
what are we actually trying to measure?
are we assessing job-relevant skill, or social conformity?
where are we mistaking difference for weakness?
what kinds of candidates are our systems quietly filtering out?
These are the questions that make recruitment more accurate as well as more inclusive.
What employers should take from this
The most useful takeaway from Nat Hawley’s BBC3 appearance is not simply “be careful with AI”.
It is this: if your hiring process cannot recognise talent that presents differently, then it is not as fair or as smart as you think it is.
Employers should be reviewing:
what their assessments actually measure
whether their video tools privilege neurotypical social cues
whether “gamified” tests are relevant to the role
how candidates can request alternatives or adjustments
whether recruitment is selecting for job performance or recruitment performance
And more broadly, they should be asking whether their current process is helping them find the best people — or only the most conventionally legible ones.
Final thought
Computer Says No is such a revealing title because it captures the bluntness of what is happening.
For too many neurodivergent candidates, the problem is not that recruitment is hard in the ordinary sense. It is that the game has been designed around assumptions that were never built for them. And when those assumptions are handed over to AI, the exclusion becomes faster, tidier and easier to miss.
Nat Hawley, founder of Divergent Thinking, cuts through that neatly. His warning is simple: if employers let black-box systems decide who looks competent, they risk screening out exactly the people who could bring the insight, loyalty and innovation they claim to value.
The goal is not to reject technology. It is to stop mistaking automation for fairness.
To explore more of Nat Hawley’s work, visit Divergent Thinking, read more on the blog, or learn more about workplace needs assessments.
Comments