Curiosity

Curiosity is the one of the seven that almost everybody believes they already have. It is also the one the tools make easiest to stop using, because the discomfort that starts an inquiry is now relieved in about a second.

Curiosity is the disciplined drive to explore, learn and update beliefs in the face of new evidence. The discipline is the load-bearing word. Interest is a mood and arrives on its own. Curiosity is what somebody does about it when the answer is already available for free.

What the evidence supports#

Four findings, from four literatures. None of them is about AI. Say that first, before they are used to argue about it.

von Stumm, Hell and Chamorro-Premuzic synthesised the prior meta-analytic evidence in 2011 and found intellectual curiosity, which they call a hungry mind, to be an independent predictor of academic performance. Curiosity and effort together explained as much variance in academic outcomes as intelligence did.

Harrison, Sluss and Ashforth followed 123 newcomers across twelve call-centre organisations in 2011. Those higher in specific curiosity sought more information from colleagues, and that information-seeking was associated with handling customer problems more creatively. The route from disposition to outcome runs through something a person actually did.

Swan and Carmelli followed 1,118 men with a mean age of 70 and reported in 1996 that those more curious at baseline were more likely to be alive five years later, after controlling for medical risk factors.

Gino surveyed more than 3,000 employees in 2018. Ninety-two per cent agreed that curious people bring new ideas to their teams. Twenty-four per cent reported feeling curious in their jobs regularly. The gap between those two numbers is the finding, and it points at the conditions rather than at the people.

What the evidence does not support#

None of the four studied AI, and none of them tested whether curiosity can be trained into somebody. Harrison and colleagues say so themselves.

So the claim here is narrower than the subject invites. Curious people do better on several measures. What a person can put on a calendar is the behaviour rather than the disposition, and the behaviour is the part the evidence connects to the outcome: Harrison's newcomers did better because they asked.

What AI changes, and how much of that is measured#

Two directions, and they do not have the same evidential standing.

The first is that a model answers faster than a colleague and is available at any hour, which removes the friction that used to sit between not knowing and finding out. That is an argument rather than a measurement, offered as one.

The second has measurement behind it, and none of it is about curiosity by name. The pattern is cognitive offloading: a capability held by a person moves to a system, and the person's own version of it thins because it is not exercised. In a randomised trial, experienced developers were 19 per cent slower with AI while believing they were 20 per cent faster, which is a failure to notice rather than a failure to work. Across 106 experiments, human and AI pairs did worse on average than the better of the two alone. AI-assisted writers produced work rated more creative individually and markedly more similar to each other.

Read those together and the risk is not that people stop being interested. It is that the first plausible answer arrives before anybody has decided whether it was the right question, and nothing in the exchange flags that.

What absence looks like#

Not silence. A person who has stopped asking still produces work, on time, that reads well. The signal is narrower. They cannot say where a number came from. They have nothing they were wrong about this quarter. The questions they do ask are all downstream of an answer they have already accepted.

At the level of a team the same absence shows up as a room in which the model's output is the agenda. Gino's 24 per cent is the measured version of it, and the conditions that produce it are older than the tools.

Permission, not reward#

Organisations that say they want curious people usually reward something else. Results are rewarded, and results contain curiosity somewhere inside them, which is not the same as rewarding it. Nobody is ever paid for the question that turned out to lead nowhere, and a great many of the questions that lead somewhere are indistinguishable from that one at the moment they are asked.

What curiosity needs is not an incentive. In Rahim Hirji's words: space and permission to explore, not permission to fail. A sliver of time. The distinction is doing real work. Permission to fail is the phrase organisations reach for, and it asks people to believe a promise about what happens after something goes wrong, which is a promise most of them have seen broken. Permission to explore asks for something smaller and more credible: a bounded amount of time in which looking into something is the job rather than a deviation from it. A sliver is enough. What kills curiosity is not the fear of failure so much as the absence of any hour in which exploring is legitimate.

Some organisations do not need curious people, and the honest ones should say so. Work that is genuinely procedural, where the answer is recoverable and the variation is the risk, does not benefit from somebody reopening settled questions. Saying that plainly is better than claiming to want curiosity and rewarding compliance, which teaches people that the stated value is decoration.

The organisations that do need it have two things to get right. The first is the permission above. The second is knowing when to stop: the five levels of questioning have a cul-de-sac at the end, and part of the skill is recognising that you have gone too far down one. Curiosity without a stopping rule is not a virtue, it is a way of never finishing anything.

Three procedures older than the problem#

Everything above argues that curiosity matters and that AI makes it easier to stop exercising. That leaves the question of what a person actually does on a Tuesday afternoon with a model open in front of them. Three procedures answer it, and none of them was invented for this. All three were built for other problems decades before generative AI, which is a point in their favour: they were not designed backwards from a conclusion about AI.

The five whys. Taiichi Ohno set this out in his account of the Toyota Production System, crediting the underlying approach to Sakichi Toyoda. Take a problem, ask why, take the answer, ask why of that, and keep going to five. Rahim Hirji's worked example, from the SuperSkills teaching deck, runs: I am always late for class. Why? I leave late. Why? I wake up late. Why? I go to bed late. Why? I scroll at night. Why? There is no cut-off time for screens.

The root cause was the phone at night. The mornings were never the problem. Stopping at the first answer would have produced an intervention aimed at waking up earlier, which is the wrong intervention delivered with confidence. That gap between the first plausible answer and the load-bearing one is the whole reason the procedure specifies a number.

The AI version of the same failure arrives faster. A model asked "how do I stop being late for class?" returns a competent answer about alarms and morning routines, because that is what the question was about. It has no way to know the question was aimed at the wrong end of the day.

SCAMPER. Bob Eberle assembled this mnemonic in 1971 from the idea-spurring checklist Alex Osborn published in 1953: substitute, combine, adapt, modify, put to another use, eliminate, rearrange. Seven ways to reopen something you had decided was finished. Its use against AI is narrow and specific. A model returns work that looks finished, formatted and closed, and the formatting is doing persuasive work the content has not earned. Running seven prompts over a finished-looking draft is a way of declining to accept the appearance as the fact.

Attention, search, knowledge. Notice what other people have walked past. Go and look on purpose rather than waiting to be told. Then use what you found, so the next loop starts from further along. This one is Rahim Hirji's mnemonic from the same deck, September 2026, and no claim of first use is made for it: the three words are ordinary English and the sequence is a teaching device rather than a finding.

What none of this establishes#

The evidence on this page supports the claim that curious people do better on several measures. It does not support the claim that running these procedures makes anyone more curious. Harrison and colleagues are explicit that their result cannot show curiosity being trained into people, and neither Ohno nor Eberle ever tested whether their procedures changed a disposition, because neither was trying to. Both are instructions with sixty and fifty years of adoption behind them and no controlled test.

Narrow the claim accordingly. These procedures make a person behave, for a few minutes, the way a curious person behaves anyway. Whether the behaviour reaches back and changes the disposition is unmeasured, and the research declines to assert it. What can be said is that the behaviour is the part the evidence connects to outcomes: Harrison's newcomers did better because they asked, and asking is the thing a procedure can put on a calendar.

There is also a reason to expect the procedures to matter more now than when they were written, offered here as an argument and not as a measurement. Each was designed for a world where the cost of the first answer was high. Ohno's engineers had to walk to the machine. Osborn's teams had to sit in a room. When getting a plausible first answer took effort, the effort itself supplied some of the discipline that the five whys formalises. That cost has gone to nearly zero, and the discipline it used to carry has gone with it.

Rahim Hirji's line from the same deck holds the whole section together: AI will answer any question you ask, and it will never tell you that you asked the wrong one. That part has not been automated.

Assessing curiosity#

Curiosity resists self-report more than most of the seven, because the disposition it describes is one almost everybody believes they have. What can be observed instead, organised by the ten subskills beneath it:

• Asking better questions, self-initiated discovery. Does a log of questions exist, and did any of them become a test? The Curiosity Loop produces the record: a question, a small test by Friday, and a sentence of the form "we learnt X, so we will Y". A person running it has the record. A person who is interested has recollections.
• Independent research and exploration, information synthesis. When they bring you something, can they say where it came from and what they checked? The source rule is the practice that makes the answer available.
• Metacognition, open-mindedness. Can they name something they were wrong about in the last quarter, with the date? The ability to produce an example on request is the closest available proxy, and its absence is informative.
• Learning agility, active learning strategies, lifelong learning. What did they learn most recently that nobody asked them to learn? The question separates curiosity from compliance with a development plan.
• Growth mindset. How do they respond to being corrected in front of other people. Observable, uncomfortable, and more reliable than any instrument.

What this assessment cannot do#

It is not scored and has not been validated. A person can produce a log without the disposition behind it, and a curious person in an organisation that punishes questions will produce no record at all, which makes the absence of evidence ambiguous rather than damning. The wider limits are at how you assess capability rather than output.

Where this goes next#

• What makes a good question, the practice underneath the first step of the loop, with the Question Formulation Technique attributed and three tests of Rahim Hirji's own.
• The Curiosity Loop, see, seek and shift, and the 48-hour rhythm that is the smallest version that works.

Key research and primary sources

• von Stumm, S., Hell, B. and Chamorro-Premuzic, T. (2011). The Hungry Mind: Intellectual Curiosity Is the Third Pillar of Academic Performance. Perspectives on Psychological Science, 6(6).
• Harrison, S. H., Sluss, D. M. and Ashforth, B. E. (2011). Curiosity adapted the cat: The role of trait curiosity in newcomer adaptation. Journal of Applied Psychology, 96(1).
• Swan, G. E. and Carmelli, D. (1996). Curiosity and mortality in aging adults. Psychology and Aging, 11(3).
• Gino, F. (2018). The Business Case for Curiosity. Harvard Business Review.
• Ohno, T. (1988). Toyota Production System: Beyond Large-Scale Production. Productivity Press. The source of the five whys, and a practitioner method rather than a tested one.
• Eberle, R. F. (1971). Scamper: Games for Imagination Development. D.O.K. Publishers. Built on Alex Osborn's 1953 checklist. The 1971 edition was not opened; the resolvable object is the 2023 combined edition.

Every figure on this page has been checked against the source that reports it. Where a number could not be confirmed, it was removed rather than left standing, and what was removed is recorded in the build register.

About this research#

Written by Rahim Hirji, author of SuperSkills (Kogan Page, 2026), keynote speaker on AI and human capability, and founder of The SuperSkills Intelligence Company.

How this research works · Reviewed quarterly · Found an error? Tell me and it is corrected on the page.

Listen. Heard on AI for Business Leaders, The Fourth Dimension: Building Human Capability into the AI Strategy with Rahim Hirji. Heard on Education Futures, SuperSkills: The 7 human skills AI can't replace (with Rahim Hirji). Heard on The Growth Hacking Culture Podcast, AI Deskilling: Why Your Team Is Getting Worse at Thinking. Heard on Inside Learning, Super Skills: The 7 Human Skills for the Age of AI with Rahim Hirji. Heard on Inside Learning, Super Skills: The 7 Human Skills for the Age of AI with Rahim Hirji.

• Curiosity
• Change
  readiness
• Principled
  innovation
• Global
  adaptability
• Empathic
  communication
• Big
  picture thinking
• Augmented
  mindset