Recruiters were asked this year how they separate a candidate from a pile of resumes. The top answer was critical thinking, at 78%, ahead of work ethic and communication, from a national survey. Also this year, Cengage surveyed nearly 800 college instructors and found 73% reported declines in critical thinking, with 70% naming growing dependence on AI and technology as a contributing cause. More than half of respondents said that this year's graduates are less prepared than graduates from a decade ago. The skill everyone is hiring for is the exact skill everyone is losing, and the tool driving the demand for the skill is implicated in its loss.
This is a skill gap that hides, particularly when organizations aren't built to evaluate it. When critical thinking gets weaker while everyone gains increasing access to AI tools that produce competent output, nothing looks wrong. Documents still read. Decks get produced and approved. What goes missing is the internal moment where somebody would have stopped and asked whether this is the right question to be asking, whether the work answers it, and whether it serves the goal that started the process in the first place. This moment leaves no trace when it gets skipped, and the tools won't flag it either, since they're built to affirm, not diagnose. You end up with a room full of people losing a skill none of them can tell they're out of practice at.
The vocabulary is interesting. I argued that creativity got rebranded as judgement and taste because those sound more like business skills. Critical thinking is what sits beneath both, and unlike the newer terms it already has a research tradition and a measurement history behind it.
What's new is where it's being demanded. Critical thinking has always been valued as you move closer to the top of an organization, in the roles that set direction. It wasn't as important in the person producing the deliverable, because that role was focused on production and making more time for production. Increasingly, critical thinking is in-demand at a lower altitude. The skill itself hasn't changed.
Everything I've argued about formal critique training being good preparation for this moment could be read as an argument on the credential. That's not really what I'm arguing. "Go back and get an art degree" isn't advice. It's my own humorous vindication of my own personal training being persistently devalued. But the training wasn't the art degree itself, it was a format for developing a skill. Formal critique is an established and repeatable meeting structure that anyone can run in any field for any kind of work. It's been comfortably running outside of art school forever. None of that requires knowing anything about painting. It requires a space, a piece of work, and an agreement on how the conversation flows.
There's no "one weird trick" to the format. Somebody presents work and states what they were trying to do before anyone reacts. This gives the room something to evaluate against besides personal preference. Feedback is anchored to that stated intent rather than whether people like the work. Teams that run this well have people write reactions silently before discussion opens, so the first confident voice doesn't set the frame for the entire conversation. The most senior person speaks last for the same reason. Comments are expected to identify what isn't working and say something about why. This is the difference between a reaction and an instruction.
Every one of those actions is a piece of critical thinking, and many disciplines claim to teach it. It isn't unique to art school, but what I identified as unusual from my own experience was the rigor and concentration. The same exercise in front of the same people every week for four years, with the difficulty scaling as the team improved, and then looping back into making and starting all over again. It's not something you study in a unit on critical thinking, demonstrate on an exam, and set down.
A systematic review in 2024 of peer assessment in higher education found strong evidence that it improves learning outcomes under specific conditions: assessor training, multiple rounds, and structure around how the assessment happens. Separate research looking directly at critical thinking found that structured peer assessment improves it, alongside creative self-efficacy and learning outcomes. Those conditions are exactly what casual workplace feedback lacks. It sniffs of a formal critique without the proven parts that make a critique function.
It's also been demonstrated that a person giving feedback learns more than the person receiving it. A controlled experiment by Culver put students into groups that either exchanged feedback or only assessed other people's work without getting anything back, and the students who only assessed improved more in their own writing. Lundstrom and Baker found the same thing. Graner found that students who only assessed did as well as students who both assessed and received. A meta-analysis in Educational Psychology Review covers the pattern across a broader set of studies. The proposed mechanism is exactly what the crit format demands. You have to recognize criteria, judge quality against them, and make decisions.
The students who benefitted the most from the crit exercise were the ones who rated it as less valuable, because they didn't receive anything they could feel. Organizations unable to measure judgement measured output instead and rewarded the visible part of the effort. The valuable thing remains invisible to the people holding it.
Which is precisely why the formal crit holds up against what AI use erodes. The concern in the research isn't that people are using AI to produce worse work. It's cognitive offloading. The pattern where handing a task off to a system means the underlying capability stops being exercised. Michael Gerlich studied 666 people across a range of ages and education levels and found a significant negative relationship between heavy AI use and critical thinking scores, with cognitive offloading being the mediating factor. The learning lives in performing the evaluation, and offloading removes exactly that step. That's true whether you're relying on AI or a metrics dashboard to provide your answer. A well-run critique structurally requires that cognitive step. You can't reconstruct what someone was reaching for and defend it when everyone disagrees without doing the thinking yourself. There's no version of that you can hand to a machine that will agree with whatever you feed it.
The objection that always comes next is that critique is slow and demanding. It takes an hour a week that nobody has and we've shown that people don't notice the benefit it produces. That argument won for most of the history of professional work, because the constraint was always production capacity, and any hour spent talking about the work was an hour that wasn't spent making it. That constraint has changed, and continues to change. The hours that used to go into making things are simply available now in a lot of jobs. The default assumption is to convert that time into more output, but there's no rule that says you have to. Increased productivity means there is room in the day for increased quality. Quality is the thing that organizations have never been able to measure, so it keeps losing to whatever is countable. The time to do critique properly is the very thing the tools just handed over, killing the traditional objection.
It solves the other problem I pointed out in the second post too. The routine junior work that used to function as apprenticeship is disappearing. A junior person no longer spends their first three years in production drudgery, absorbing judgement slowly by proximity and hoping a senior leader has time to explain a decision. They can spend those hours in the room where decisions get argued out loud, doing the evaluation and building the skill that the research says teaches more than receiving feedback does. That's a faster path from junior to senior than the one it replaces, and it runs on time that already exists.
However, it's not a magic meeting. Remember, the research showed that peer assessment works under specific conditions. It also takes months before a group gets good at it, because the quality of a critique depends on group norms, not just how good one person in the room is. Plenty of art school crits were competition dressed as rigor, which taught people to get defensive, rather than to think. A team can also run critiques until a group converges on one house standard and easily mistake that agreement for quality. A shared method for building judgement is safe to standardize. A shared standard for what counts as good is not, which is part of what makes this hard to measure.
That last point is where the four pieces of this series converge.
Critical thinking isn't issued at birth. It's trained. Formal critique is a standard concentrated way of training it. The systems we built to recognize that skill assumed measured output was a reliable stand-in for the thinking behind it, and AI broke that link faster than any hiring or review process can adapt. Meanwhile the tools pull hard toward the middle by default, so differentiation stopped being a natural result of talented people working together and became something you have to want on purpose. Three separate problems. One shared cause. What's strange is that the same solution addresses all three.
You make something. You put it in front of people who will tell you where it fails your own stated goals. You learn to find the real note in the noise. You go back and change it. Then you do it for someone else's work, which teaches you more than your own turn did. That loop is old and it isn't proprietary. Art school just kept running it after most of professional life decided the cost was too high and the result didn't show up on a performance dashboard.
What's changed recently is the cost. Critique always lost the cost argument, because it took hours from production and production was the constraint. It isn't anymore. The tools took away the drudgery and handed back time. That time can go to the work that has always resisted measurement and has always been where quality comes from, and back to junior people who now have room to build judgement by exercising it instead of waiting years for someone to hand it over. Or it can go into producing more, faster, which is where most of it is going right now, because that's the part the dashboard can see.
These tools will keep getting better at producing confident, competent answers. Whether they're good was never really the question. The question is whether anyone in the room is still practiced enough to tell, and whether the organization would notice if they weren't. Everything we can currently measure says that capability is going down while demand for it goes up. That's a strange thing to watch happen in real-time and decide to do nothing about.