The Propinquity and the Probability

Why our proximity to the consequences of AI may matter more than we care to admit

I've long been interested in the relationship between what people believe and what they might stand to lose by changing those beliefs. Hardly a novel observation, and something psychologists have been examining for decades. Nevertheless, I suspect we're witnessing a rather interesting demonstration of the phenomenon in the public response to artificial intelligence.

What interests me is whether some of the increasingly polarised commentary surrounding AI has less to do with the technology itself than with our proximity to its consequences. Specifically, whether our assessments of its dangers and opportunities are being influenced by what might happen to our livelihoods, professional identities and, perhaps most interestingly, our positions of authority.

There are, of course, legitimate concerns about AI, including its reliability, governance, environmental impact and the concentration of considerable technological power within a relatively small number of companies. I share several of these concerns. What I am less convinced about is the extent to which we recognise our own interests when forming opinions about it.

I wonder how often we confuse a principled objection with one that happens to preserve something we already possess.

Stephen Fry and the ownership of language

I have considerable affection for Stephen Fry. I've always admired his intellect, his humour and his obvious delight in language. In his 2008 essay, Don't Mind Your Language…, he challenged the pedants who seemed rather more concerned with correcting other people's grammar than appreciating what they were trying to communicate.

I rather liked his position. Language is not some carefully preserved museum exhibit. It changes, adapts and occasionally does things that would send the custodians of grammatical propriety reaching for the smelling salts.

There was also something generous about Fry's argument. Language needn't belong exclusively to those who have mastered its formal conventions. People should be free to experiment with it, misuse it and occasionally improve upon it.

Which makes his more recent experiences with AI interesting.

In 2023, Fry discovered that his voice had apparently been reproduced using AI without his knowledge or permission. Recordings of his narration of the Harry Potter books had been used to create a synthetic voice capable of delivering words he had never spoken. Understandably, he was alarmed.

Now, it would be rather unfair to suggest that someone who had previously championed linguistic freedom was suddenly less enthusiastic when technology began reproducing his own linguistic identity. Defending people's freedom to use language is not the same thing as defending another person's freedom to impersonate you. Fry's concerns about consent and authenticity are perfectly reasonable. Indeed, his subsequent involvement in an AI voice installation at the Hay Festival in 2025 suggests he is willing to engage with the technology rather than simply reject it.

Nevertheless, his experience got me thinking about whether our relationship with technology changes when its consequences begin to affect us personally. It is one thing to admire something from a comfortable distance, and quite another to discover that it is capable of interfering with something we regard as our own.

The intellectual territory we occupy

Consider the different responses to AI from Noam Chomsky, Steven Pinker and Brian Cox.

Chomsky, whose contribution to linguistics and cognitive science is considerable, has been sharply critical of large language models. In a 2023 essay with Ian Roberts and Jeffrey Watumull, he argued that their statistical approach to language differs fundamentally from the explanatory reasoning associated with human cognition.

His objections are not trivial. They arise from a theoretical position developed over many decades, and they deserve serious consideration. Yet the emergence of systems capable of producing sophisticated language without apparently requiring all the mechanisms traditionally associated with human linguistic competence raises some awkward questions about our existing assumptions.

That doesn't make Chomsky wrong, and I have no evidence that his criticism is motivated by a desire to protect his intellectual legacy. But I do wonder at what point defending an established theoretical framework becomes difficult to distinguish from defending the intellectual territory built around it.

Pinker offers a somewhat different perspective. He acknowledges the extraordinary capabilities of AI whilst maintaining important distinctions between human and artificial intelligence. He also recognises its potential to augment our cognitive abilities, much as previous technologies have extended our capacity to calculate, remember and retrieve information.

I find this position rather more appealing, although I recognise that may simply reflect my own preferences.

Brian Cox, approaching matters from physics, has similarly acknowledged the possibilities and uncertainties associated with AI. What appeals to me here is the willingness to admit that we don't yet know where this technology might lead. It's an attitude I would hope to find amongst scientists, although I suspect we are not always as comfortable with uncertainty as our professional commitments might suggest.

We understand the architecture and mathematical operations of contemporary transformers, but our understanding of how their learned internal representations produce particular capabilities remains incomplete. We are also some distance from knowing how reliably those capabilities will generalise or what further developments might reveal.

As someone with a longstanding interest in consciousness, I find this particularly intriguing. We have yet to arrive at a universally accepted explanation of human consciousness, which makes me reluctant to accept categorical declarations about what artificial systems could never become.

I would rather acknowledge what we don't know than use that uncertainty to defend what I already happen to believe.

Propinquity, probability and personal consequence

There are some established psychological ideas that may help explain what we're seeing.

Kunda's (1990) work on motivated reasoning describes how our desires and existing commitments can influence the evidence we attend to and the conclusions we consider reasonable. Importantly, this needn't involve deliberate dishonesty. We can arrive at conclusions that suit us whilst sincerely believing we're being entirely objective.

A slightly uncomfortable possibility for academics, given that evaluating evidence independently of personal preference is supposed to be part of the job.

Kahneman and Tversky's (1979) prospect theory also offers something worth considering. People don't necessarily weigh potential gains and losses equally. The prospect of losing something we already possess may carry rather more psychological weight than the possibility of acquiring something of equivalent value.

I wonder whether these processes help explain some of the polarisation surrounding AI.

For someone who sees AI as an opportunity to acquire new capabilities, investigate ideas or bypass established barriers, technological progress may seem largely beneficial. Someone whose professional identity depends upon possessing expertise that others do not might see the same development rather differently.

The technology hasn't changed. What has changed is the position from which it is being observed.

This is where I find the relationship between propinquity and probability interesting. By propinquity, I mean the perceived proximity of AI to something an individual values, whether that is employment, authority, professional identity or intellectual standing. By probability, I mean how likely they believe the anticipated consequences are to occur.

These judgements are unlikely to be independent. A development that appears distant from our circumstances might attract curiosity or even enthusiasm. Once we believe it could affect our own position, we may become rather more circumspect.

There is some empirical evidence consistent with this possibility.

Grant, Green and Evans (2026), examining attitudes amongst British workers, found that greater occupational exposure to AI was associated with stronger support for government regulation. Those anticipating negative consequences for their employment prospects were also more supportive of regulation. Interestingly, expecting benefits from AI did not produce an equivalent reduction in regulatory support.

A separate study by Pye and colleagues (2026), involving 201 UK psychology professionals, found that perceived threats to professional capabilities and personal identity were associated with lower intentions to adopt generative AI, even after accounting for established technology acceptance factors.

Neither study demonstrates that criticism of AI is irrational or self-serving. Supporting regulation may be entirely sensible, and concerns about deskilling may reflect genuine risks. The latter study was also cross-sectional, so it cannot establish that perceived identity threat caused resistance.

Still, the findings suggest that people's attitudes towards technology are influenced by more than their assessment of what it can accomplish. What it might do to them also appears to matter.

When knowledge becomes less scarce

I think this is particularly relevant to higher education.

Having spent more than three decades working in and around academia, I've witnessed the expansion of systems through which knowledge, authority and professional legitimacy are organised and maintained.

Universities are ostensibly concerned with the creation and dissemination of knowledge. They are also institutions in which qualifications, disciplinary boundaries, publication records and professional positions help determine who is entitled to speak, who is believed and whose ideas receive serious consideration.

Much of this is entirely reasonable. Expertise matters, scientific methods matter, and there are very good reasons why an unsupported opinion shouldn't automatically carry the same weight as a carefully conducted programme of research.

But I am not convinced that every feature of our existing academic arrangements can be justified on those grounds.

AI presents an interesting problem because it is beginning to reduce some of the practical barriers that have traditionally separated those who possess particular forms of expertise from those who don't.

Someone with limited formal training can now explore theoretical arguments, interrogate statistical methods, develop computational models or structure an essay with assistance that might previously have required considerable institutional support.

None of this makes them an expert. Nor does having access to plausible answers guarantee understanding. In fact, AI may allow people to become confidently wrong with considerably greater efficiency, which is hardly something we're short of already.

But it may also enable someone with a worthwhile idea to investigate it, develop it and communicate it in a form that others can evaluate.

I suspect this is where some of the discomfort originates.

If language, technical fluency and familiarity with academic conventions become less effective barriers to participation, what happens to the authority historically associated with possessing those capabilities? And if AI can reproduce some of the outputs through which academic expertise is recognised, will universities need to reconsider which elements of that expertise are genuinely indispensable?

I've explored related questions in Beyond Manufacturing Consent, particularly the relationship between intellectual authority, institutional gatekeeping and access to ideas.

I'm not suggesting academics are universally hostile to AI. Many have been instrumental in developing it and evaluating its capabilities. Nor does every objection to AI-assisted work reflect professional insecurity.

But when I see declarations that AI-assisted writing is inherently inferior, or that an argument should be dismissed simply because its prose appears machine-generated, I wonder whether we're evaluating the substance of the work or defending conventions through which intellectual legitimacy has traditionally been assigned.

There is a difference, and I suspect it will become increasingly difficult to ignore.

The profitable business of optimism

Of course, this argument cannot be directed exclusively at those who oppose AI.

Those promoting the technology are no less susceptible to the influence of their own interests.

An entrepreneur whose company valuation depends upon expectations of continued technological progress has every reason to emphasise its transformative potential. An investor anticipating substantial returns may be inclined to regard concerns about employment, reliability or regulation as temporary obstacles.

Equally, those who have found AI personally empowering may underestimate its consequences for people whose circumstances differ from their own.

I include myself in that observation.

I use AI extensively to organise ideas, examine arguments, test assumptions and investigate areas in which my knowledge is limited. I regard it as a potentially extraordinary extension of human intellectual capability.

But my enthusiasm may reflect the opportunities it affords me personally. I'm interested in what becomes possible when specialist assistance is no longer restricted by institutional position, technical proficiency or financial resources.

Someone whose livelihood depends upon providing those services may understandably see things differently.

Neither perspective is automatically correct.

There is money to be made from promises of technological salvation, just as there may be professional authority to preserve by predicting catastrophe.

I wonder whether some of the apparently irreconcilable positions dominating public discussion reveal rather more about the incentives surrounding AI than its eventual capabilities.

The middle road is rather less comfortable

Historical responses to technological development offer some parallels, although none is exact.

The Luddites are frequently caricatured as irrational opponents of progress, when their concerns involved employment, wages, skills and economic power. Photography altered the relationship between artistic skill and accurate visual reproduction without bringing painting to an end. Calculators and computers changed educational practice whilst provoking continuing arguments about what students should understand independently.

What interests me is that resistance to technological change has frequently been grounded in consequences that were real, or at least plausible.

We shouldn't dismiss concerns about AI because they arise from threatened interests. Someone may have a personal reason to fear a development and still be correct about its dangers.

Nor should enthusiasm for technological progress automatically be mistaken for intellectual openness. The anticipated consequences may simply be rather favourable to the enthusiast.

I suspect things become particularly troublesome when the future is uncertain. Where evidence is incomplete and the consequences potentially profound, we have considerable freedom to select interpretations that accommodate our existing beliefs.

Perhaps this helps explain the confidence with which some predict either the end of human intellectual endeavour or a future in which AI resolves almost every difficulty we face.

Both positions strike me as rather ambitious, given how much remains unknown.

I've previously considered the value of occupying a middle position in Take the Middle Road: Existential Angst in the Age of AI. I remain drawn to it, although I suspect it requires rather more effort than simply choosing a side.

It isn't a matter of splitting the difference between two confident opinions. Both may be mistaken. Nor should uncertainty become an excuse for doing nothing where there is good evidence of harm.

It requires a willingness to revise one's assessment as the evidence changes, even when that revision has uncomfortable implications for one's own circumstances.

For academics in particular, that ought not to be especially controversial.

The problem of standing too close

I began by wondering whether our attitudes towards AI are influenced by how close we stand to its consequences.

I suspect they are, although the relationship is unlikely to be simple. Financial security, professional identity, intellectual commitment, personal experience and perceived opportunity may all contribute, and not necessarily in the same direction.

What interests me is that these influences may not always be apparent to the individuals concerned.

We're rather good at constructing principled explanations for positions that happen to accommodate our own interests. This doesn't mean those explanations are necessarily false. It simply means that our confidence in them might deserve a little more scrutiny.

Perhaps the most revealing aspect of the current debate is how frequently we see the same technology described as either a remarkable opportunity or an existential threat, depending upon who happens to be doing the describing.

I don't think AI will inevitably produce a utopia, nor do I accept that it must lead to intellectual or social catastrophe. Its consequences are likely to be unevenly distributed, beneficial to some and damaging to others. That alone gives us good reason to investigate them carefully rather than leave their distribution entirely to commercial or institutional interests.

But I wonder whether we might learn something by paying closer attention to the relationship between the opinions being expressed and the circumstances of those expressing them.

This is not an invitation to dismiss an argument because its author has something to gain or lose. That would be a poor substitute for examining the argument itself.

It is instead an invitation to apply the same scepticism to our own certainty that we so readily apply to others.

Perhaps the question is not simply what AI is capable of doing, but what we're capable of admitting about it when the consequences begin to reach our own doorstep.

References

Chomsky, N., Roberts, I., & Watumull, J. (2023, March 8). The false promise of ChatGPT. The New York Times.

Fry, S. (2008, November 4). Don't mind your language…. The Old Friary.

Grant, Z., Green, J., & Evans, G. (2026). New tech, new threat? Occupational exposure to artificial intelligence increases support for AI regulation. Journal of European Public Policy. https://doi.org/10.1080/13501763.2026.2710714

Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47(2), 263–291. https://doi.org/10.2307/1914185

Kunda, Z. (1990). The case for motivated reasoning. Psychological Bulletin, 108(3), 480–498. https://doi.org/10.1037/0033-2909.108.3.480

Pinker, S. (2023, February 14). Will ChatGPT supplant us as writers, thinkers?. The Harvard Gazette.

Pye, S., Hazell, C. M., Smith, K. J., & Bergin, A. (2026). Identity threat as a barrier to NHS psychology professionals' adoption of generative AI: A technology acceptance model study. Behavioral Sciences, 16(10), Article 1827. https://doi.org/10.3390/bs16101827