To have your cake and eat it: AI, The Dunning-Kruger Effect and Critical Thinking

 To have your cake and eat it: AI, The Dunning-Kruger Effect and Critical Thinking

There is a very nice, very French expression: ‘Vouloir le beurre et l’argent du beurre’, which in English is simply ‘To want one’s cake and eat it too’: both point to the impossibility (or the greed perhaps) of wanting two opposed things at once. So you have butter, you want to keep it and eat it, but at the same time you want to sell that butter to make money. Can you have both at the same time? Can the cat be both dead and alive?

The recent, barely believable hypocrisy displayed by the Tech Bros and their AI companies is an interesting starting point. They now call for ‘Regulation of AI’ because the pace of its development has been too rapid[1]. Well yes, I wonder who pushed that development, and who put AI out there without ever having thought the consequences through. It’s not new of course (DDT, tobacco, pollution) but it does still rankle.

What is ever more painful, however, is that those CEOs’ attitude really is a case of ‘having your cake and eating it’, and unfortunately it mirrors exactly what most people’s attitude today is as well. Basically, what I see around me is people using AI for minor tasks while advising others to be careful when doing so. Do as I say, not as I do. In essence, those people are asking for AI to be regulated while making use of it for things they could easily do themselves – or should be able to do.

It is very reminiscent of a recent study that investigated the Dunning Kruger Effect [DKE] on car drivers, which concluded that ‘The data consistently confirms the DKE across driving situations, highlighting its universality[2]’. The Dunning Kruger Effect is a well-established cognitive bias that says that ‘individuals with lower skills tend to overrate their abilities, while high achievers often underestimate themselves’. When it comes to driving, it means that the majority of drivers rank their skills very highly while ranking the skills of others very low, which leads them to, for example, drive over the speed limit (‘I’m good enough to do it even if others aren’t’), ignore road signals (‘these are for bad drivers, not for good ones like me’) and run through red light (‘my handling of the car is of superior quality so I can make it where others couldn’t’). Obviously, that is a major source of road accidents and casualties.

And that is what seems to be happening with AI too, and at different levels.

In July 2025, a report said that ‘ChatGPT receives 2,5 billion prompts every day[3]’ – BILLION! And that was a year ago, time enough for Gemini and whoever else to accumulate prompts themselves. In fact, you can’t even type something in Google without AI being automatically activated. The data centers necessary for all this – and more – are depleting lands of water, sending electronic chips’ prices through the roof and generally scarring the landscape and polluting the environment. And people don’t like that, and they protest, and they say how terrible it all is. Yet those prompts do come from somewhere, surely? I want my cake, I want to eat it, and, as the French sometimes add, I want to marry the baker’s daughter as well.

In education, we have seen hundreds, thousands of workshops on AI-use, seminars and conferences on how to use AI, papers published semi-humanizing AI by referring to it as ‘a partner’, a ‘buddy’. We have seen governmental reports urging the implementation of AI in schools, in the administration, in services. We have seen governmental agencies publishing reports on AI-use, and national educational bodies including AI in their list of necessary competencies for learners. And all of them, always, use that phrase: ‘Use AI, but do so critically’ – see for example, in the Netherlands, the SLO’s kerndoelen. ‘Use AI, you have no choice, but think critically’. A nice one, this, as it manages to encourage the use of critical thinking while committing a logical fallacy: no choice, really?

And yet what I do not see is what is being done about the ‘Thinking critically’ bit: where is that learned? Where is that taught? Critical Thinking is typically understood as being made of two related yet distinct concepts: the Critical Thinking Skills (e.g. making connections between ideas, drawing conclusion, checking sources, double-checking, inductive and deductive thinking), and the Critical Thinking Dispositions: the mindset necessary to make use of those skills – for example, the desire to know, to be objective, to take different views into account. Here again, the DKE strikes its blow: it’s always other people who cannot think, other people who are uncritical – it’s always someone else who must learn it all. It never applies to me. How many times have I had students (and colleagues) saying they were ‘very critical thinkers’ and yet prove unable to spot a logical fallacy from close-up, or worse, producing such fallacies themselves?

Yet the tentative results from a three-year long inquiry into critical thinking at school with my Master students would indicate that while critical thinking is highly valued in mottos and slogans and mission statements and discourse, it tends to be poorly implemented in everyday teaching. At my own institution, I have yet to be offered the chance to take part in a training session on critical thinking, or see a symposium organized around it. I was given the chance to learn how to use AI many times over, but never that of learning about critical thinking. Odd, I think. And really concerning too, when considering one little piece of data in the recent (2025) PISA report:

When considering the average changes observed across OECD countries, least-affected were items that demanded from students to locate information, which students could successfully solve even with only a superficial, yet targeted, engagement with the text. Conversely, the most-affected items relate to higher-level processes (evaluating and reflecting), which require careful reading, and often required students to remember and integrate multiple pieces of information, either from a single source or from multiple sources.

There’s a problem with reading skills, certainly, but mechanically finding some words in a text (turn-taking markers, logical connectors) is not the same as actively thinking about disparate bits of information and synthesizing them. The latter is critical thinking, both skills and dispositions, while the former consists of applying simple rules which do not necessarily lead to understanding. As an illustration, I asked my first-year Bachelor students, fresh from secondary school, whether they knew the word ‘Metaphor’: 90% did. ‘Where do you know that word from?’, I asked. ‘From school, in the language and literature classes’, they said.

But when I asked them to tell me what it meant, or what they thought, in their own words, that a metaphor did or how it worked, only one had a go (and it wasn’t very convincing): all the others admitted that while they knew the word, they didn’t know the concept behind it. Most could even give an example, more or less, and they could probably find one in a text. Yet what it does, how it enriches meaning, how it translates from the source domain to the target one and how that creates meaning – all this was completely lost.

We urgently need to address the problem of systematic, curriculum-wide, fully embedded teaching and practicing of critical thinking in all its dimensions. As the Dunning Kruger Effect is taking hold in everyday AI use, more and more people begin to think that what is absolutely needed when using AI – Critical Thinking – is only needed by others and not by them. We cannot afford to keep saying that we must use AI critically (eating your cake) while urging the use of it without a clear plan on how to support that use (having your cake). Developing Critical Thinking Skills and Dispositions takes time, repetition, full engagement from the teaching staff and therefore full support from the management. It needs to cut across the whole curriculum and reflect the variety of ways we can’t always trust ourselves to think the right way (Paul 2005).

Let’s not be stupid drivers of fancy cars who are such easy victims of the DKE, and cause so much damage to the rest of us. Sam Altman is not going to save you, nor is Elon Musk, so we must do it ourselves.

 

Navarro, J. et al. We all fall for it: Influence of driving experience, level of cognitive control engaged and actual exposure to the driving situations on the Dunning-Kruger effect. Transportation Research Part F: Traffic Psychology and Behaviour, Volume 109, 2025, Pages 1-18, https://doi.org/10.1016/j.trf.2024.11.021.

Paul, R. (20012). Critical thinking. Tomales : FCT.

SLO (2025). Actualisatie kerndoelen en examenprogramma’s
https://www.slo.nl/thema/meer/actualisatie-kerndoelen-examenprogramma/actualisatie-kerndoelen/definitieve-conceptkerndoelen-digitale/

https://www.theguardian.com/technology/2026/sep/14/ai-ceo-safety-slowdown

https://www.sciencedirect.com/science/article/pii/S1369847824003231

https://techcrunch.com/2025/07/21/chatgpt-users-send-2-5-billion-prompts-a-day/

 

I'm Vincent Hernot - born and bred in St Malo, Brittany, France, educated there and in Rennes, then in Ireland and in Normandy, pushed it up to a PhD in Irish things (Linguistics and Political Science), before moving to the Netherlands at the tail-end of my 20s. Found a job, then another, and finally stuck at it at the Hogeschool van Amsterdam (University of Applied Sciences) where I've mostly been teaching all things Linguistics and a whole lot of Literature (at Bachelor and Master level). The one constant since my early teens has been reading - reading, reading, reading :) (and playing music, so that's two constants)

Comments

Post a Comment

Popular posts from this blog

Five things I hate about you: AI and those small things that make us human

  Those apparently 'simple' tasks we delegate to Chatgpt will end up costing us more than we know In his 1994 noir novel ‘Gun, with occasional music’, the American writer Jonathan Lethem thought up a bleak version of the future in which the State makes the taking of a certain drug compulsory: that drug is called ‘Forgettol’, and it basically erases your memories, short or long, and makes it impossible to remember even basic facts about yourself, let alone others. That then leads to the invention of the ‘memory-box’: a device characters use to effectively externalize their brain. That memory-box stores all those memories your brain can’t handle; in fact, it stores all the information your brain can’t handle. The benefits for the state are obvious (no possibility of rebellion), but the effects on people and society are terrible: a fragmentation of one’s consciousness which leads to a fragmentation of the social fabric. Sounds familiar? When the printing press started to spre...
Read more

AI, creativity and citizenship

Image
A rush of thoughts is the only conceivable prosperity that can come to us (Emerson) In the numerous debates around AI that have been taking place lately, we have all heard ‘ tout et son contraire ’, as we say – we heard everything in all directions, for all arguments. Some say it will save the world, some that it will destroy it; some claim AI will make us better, or freer, or more creative, or more intelligent, while others will claim the exact opposite – time gained will be spent on scrawling, not learning, we will forget simple stuff, we will lose ourselves and what makes us who we are. Bertrand Russell wrote (I think in the 30s?) a little book on how we would all benefit from having 4-hour work- days: instead of slaving away behind a desk or at a machine for eight hours each day, let’s work less and create time for ourselves. Russell’s hope was that we’d use that time to educate ourselves so as to become better people. A claim many AI-supporters replicate today, in all shapes a...
Read more

The real problem with AI? Inertia

  I work in an area – Education – where the latest developments in AI technology have had an immediate impact – and yes, the same has happened across the board, I know. Since last year and Chatgpt’s arrival on the world scene, everyone – from pupils to students and teachers – has been using it. By ‘Everyone’ I obviously do not mean literally everyone, but it is clear that will soon be the case. Pupils use it to get their homework done, or pass tests; students ditto ; teachers to prepare lessons, develop material, design tests and have them marked – among many other things. Unsurprisingly, all those people claim that technology is great and helpful – well, they would, wouldn’t they? And to be fair, ‘great’ and ‘helpful’ are reasonable words to use in this case. So that’s not the problem. Equally unsurprisingly, those users will make sure to remind you of two things at all times: AI is not intelligent as we know and understand that word – it’s mindless, really – just a mach...
Read more