The Study That Proved Everyone Right

A LinkedIn thread this week revealed something more interesting than the research it was discussing.

A study landed in my feed this week. Xia, Aldharman & Chiu (2026), published in TechTrends, surveying 333 art and design students at a Chinese university to find out what drives effective self-regulated learning with AI. The headline finding: it’s not enthusiasm for AI that predicts good learning outcomes. It’s whether students perceive AI through the lens of ethics and social good. A positive attitude toward AI alone didn’t predict anything.

Dr Sam Illingworth shared it on LinkedIn. The post was measured, thoughtful, and well-framed. “The students who learn well with AI are the ones who can interrogate it and weigh what it is for. That is a skill of judgement, and it has to be taught.”

Within hours, the comments were flowing. And that’s where things got interesting. Not because of what the study found, but because of what the reaction revealed.

The Methodology Nobody Mentioned

The study has real limitations. Cross-sectional design, so no causation can be established. Every measure was self-reported, no one observed how students actually used the tool. All participants were art and design students at one university. The task was planning a summer holiday. The workshop lasted one week. There was no control group. The model measured only positive perceptions of AI, not concerns, harms, or dependency.

These aren’t obscure technical quibbles. They’re the kind of limitations that shape whether a finding generalises, and the authors themselves acknowledge several of them in the paper.

Yet in a thread of nearly 30 comments from academics, educators, and AI literacy advocates, almost nobody raised them.

What Happened Instead

The responses clustered into a pattern so clean it could have been designed as a demonstration.

The Validators used the study as scaffolding for positions they already held. “This is an important reminder that AI literacy is not about becoming better at using AI, it is about becoming better at thinking with AI.” Beautiful sentence. Has nothing to do with whether this study proves anything.

The Expanders took the finding and generalised it far beyond the evidence. “Those that choose to use AI support without critical thinking are condemning themselves to long-term neurocognitive degeneration.” A study of art students planning a holiday becomes evidence for cognitive collapse.

The Integrators wove it into their existing frameworks and products. One commenter pitched their AI platform. Another linked their own published research. The study became a vehicle for other agendas.

The Sharers amplified without interrogation. Tags, reposts, enthusiasm. The study was true because it felt right, and it was shared because it confirmed what the sharer already believed.

A handful of people pushed back. One called it “flimsy.” Another called it “absolutely contaminated research.” Another cited a retracted study with similar claims. But they were the minority, and their comments received less engagement than the affirmations.

The Psychology of the Thread

What happened in that thread is a textbook case of confirmation bias, but it’s more specific than that. There are at least three cognitive mechanisms at work.

Confirmation bias is the broadest one. We seek, interpret, and remember information that confirms our existing beliefs. If you already believe AI literacy requires critical thinking and ethical judgement, a study that says exactly that doesn’t get interrogated. It gets validated. The methodology becomes invisible because the conclusion is welcome.

The halo effect is doing work here too. The study comes from three reputable universities, is published in a peer-reviewed journal, and was shared by a credible academic with a clear AI literacy agenda. That institutional framing creates a credibility signal that overrides methodological scrutiny. We trust the source, so we don’t interrogate the claim.

And then there’s what I’d call opportunity framing. The study doesn’t just confirm a belief, it creates an opening. If ethical perception and critical thinking are what matter, then there’s a role for people who teach those things. Every educator in that thread who championed the findings was also, implicitly, championing their own relevance. That’s not cynical. It’s human. But it’s worth noticing.

The Uncomfortable Question

Here’s what bothers me about the thread. The study found that positive attitude toward AI didn’t predict effective learning. The implication is that enthusiasm without critical evaluation is insufficient. But the response to the study was overwhelmingly enthusiastic, and almost entirely uncritical.

The thread performed the exact behaviour the study cautioned against.

And that raises a question that goes beyond one LinkedIn post. If academics and educators, people whose professional identity is built on critical thinking and evidence-based reasoning, will share and amplify research without interrogating its methodology because it confirms what they already believe, what hope is there for the general public navigating AI claims?

This isn’t an attack on the people in the thread. Several made genuinely thoughtful points. But the pattern is telling. When a study tells us what we want to hear, we share it. When it tells us something inconvenient, we scrutinise it. That’s not a flaw in AI literacy. It’s a flaw in human cognition.

What This Actually Tells Us

The most interesting comment in the thread came from Steve Wright, who wrote: “This conversation always happens like there is no history. All of the things this paper says that students need to use LLMs well are the things that students need to be positive agents in their own lives. In this context, they are literally the things that students need to choose NOT to use LLMs.”

That’s a genuinely subversive point. The critical thinking and autonomy the study celebrates might be the very skills that lead you to reject the tool. But nobody engaged with it, because it doesn’t serve the narrative that AI literacy is the answer.

I’m not saying the study’s finding is wrong. It might well be true that how you think about AI matters more than whether you like it. But this study doesn’t prove that. It suggests it, in one context, with one population, using one method. And the speed with which it was shared as settled truth tells us more about our cognitive biases than it does about AI in education.

The next time a study lands in your feed that confirms exactly what you already believe, that’s the moment to be most sceptical. Not because it’s necessarily wrong. But because your brain has already decided it’s right, and stopped looking for reasons it might not be.

David Chadderton spent his twenties and thirties teaching people how to make life-or-death decisions at forty thousand feet. He now applies the same principles to consumer psychology, which, depending on the brief, can feel equally high-stakes. He’s the creator of the STAR Framework and the author of The STAR Framework: Rewriting the Rules of Consumer Engagement (NYC Big Book Award 2025), The STAR Operating System: Decode Mindset, Understand Motivation, Transform Human Behaviour, and Dear Algorithm, It’s Not Me, It’s You. By day, a Chief Marketing Officer. By night, a behavioural science obsessive who writes The Unoptimised Human because he can’t stop thinking about why people do what they do.