The Open Question

The case for curiosity as the active ingredient in cognitive defence

Most counter-disinformation work rests on an assumption that has rarely been examined with any rigour; that people hold false beliefs because they lack accurate information. Correct the record, raise the literacy, label the content, fund the fact-checkers and the problem contracts. More than a decade of public investment has followed that logic across education, platform policy and strategic communications.

The assumption fails in an instructive way. It casts the target as an under-informed processor waiting for better inputs. What hostile influence exploits is a motivational state, specifically the desire to stop processing. An adversary does not need you to believe something false. They need you to stop asking.

That distinction changes what a defence should look like.

What is under attack

Kruglanski and Webster (1996) described the mechanism three decades ago in terms that have aged remarkably well. Under time pressure, ambient noise, fatigue or threat, people seize on the first available explanation and then freeze on it, becoming resistant to information that would disturb the settlement. Need for cognitive closure is not a fixed trait. It rises with the conditions.

Read that finding operationally and cognitive warfare, meaning the deliberate degradation of an adversary population's capacity to interpret events rather than the promotion of any particular falsehood, becomes a closure-manufacturing enterprise. Flood the information environment with mutually contradictory claims. Accelerate the tempo so that no claim can be checked before the next arrives. Raise the emotional stakes so that suspended judgement feels like disloyalty. The population then does most of the work itself, because closure under those conditions is adaptive; it restores function and relieves the intolerable sense of not knowing where one stands.

Hogg's uncertainty-identity theory (2007) explains where the seizing tends to land. Self-uncertainty, the uncomfortable state of being unsure who one is and what one's place is, motivates identification with groups that are clearly bounded, internally consistent and unambiguous about their own membership. The groups best equipped to resolve that discomfort are the ones offering the least ambiguity, which in practice means the most rigid. Uncertainty does not produce extremism directly. It produces demand and extremism is well positioned to supply it.

Kruglanski et al. (2014) completed the picture for the violent end of the spectrum. Radicalisation follows a loss or threatened loss of personal significance, an ideological narrative identifying a route to restore it and a network that validates the route and the person taking it. Ideology enters late in that sequence. It is the vehicle rather than the origin.

UK referral data now says something similar in an entirely different register, with the caveats that its own producers attach to it. In the year to September 2025, Prevent received 10,293 referrals, which the Home Office reports as the highest recorded in a single year since the data series began in April 2015 and a 39 per cent rise on the previous year. Concerns recorded under the two no-ideology categories accounted for 56 per cent of the 10,282 referrals where a type of concern was specified; 3,583 under 'no ideology identified' and 2,187 under 'no ideology - other susceptibility to radicalisation identified'. Those aged 11 to 17 made up 49 per cent of the 10,279 referrals where age was known, with the 11 to 15 band the largest single group at 3,655 (Home Office, 2026).

Three qualifications matter before that is used for anything. The figures are official statistics in development, revised between releases as recording practice improves. The record itself is being claimed across a changed series; Prevent has moved from an annual April to March cycle to a biannual one and this is the first release running to September, so the comparison is not quite like for like. And the annual rise is substantially a referral-behaviour effect. The January to March 2025 quarter alone produced 3,283 referrals, 32 per cent of the year's total and 82 per cent above the same quarter in 2024. Lord Anderson, then Independent Prevent Commissioner, suggested in Lessons for Prevent that the increase likely followed publicity surrounding the Southport trial; the release reports that view without adopting it as its own explanation. Health referrals rose 58 per cent and local authority referrals 57 per cent over the year. Something changed in who refers and how readily, which is not the same as a change in the referred population.

The obvious rejoinder is that the no-ideology caseload is an artefact of over-referral rather than evidence about radicalisation and the same dataset supplies the ammunition; those categories produced 412 adoptions from 5,770 referrals, an adoption rate near 7 per cent and 23 per cent of the 1,807 adopted cases where a type of concern was recorded, while extreme right-wing concerns produced 756 adoptions and 42 per cent of that same total. Read one way, the screening process is working, separating safeguarding cases from radicalisation cases.

Read another way, the rejoinder concedes the argument. Some 5,770 referrals entered a counter-terrorism pipeline and were filtered back out of it, because what they presented carried no ideological content for the system to act on. The people behind them are not false positives in any meaningful sense. They are the population the significance-quest and uncertainty-identity literatures describe, arriving at the door of the only mechanism that will see them and being returned to a general safeguarding environment not designed around epistemic vulnerability. A defence pitched at the level of belief arrives after the decisive event has already happened. A referral system pitched at the level of ideology sees people before that event and has nothing to offer them.

Sensemaking and the moment it collapses

Weick's account of the Mann Gulch fire (1993) remains the sharpest description of what that decisive event looks like. Thirteen firefighters, twelve smokejumpers and a fire guard, died in a Montana gulch in 1949 after a routine fire behaved in a way their existing frame could not accommodate. Weick's argument is that the fatal failure was not tactical. The crew lost the shared frame that allowed them to interpret what they were seeing and to recognise each other as a functioning unit. When sense collapsed, structure collapsed with it and instructions that would have saved lives became unintelligible.

The case is an interpretive reconstruction built from Norman Maclean's narrative account (1992) rather than a body of evidence. It carries the argument as illustration and should not be asked to carry it as proof.

Weick et al. (2005) set out the properties that make sensemaking exploitable. It is retrospective, working backwards from cues to a story that accounts for them. It is social, drawing on what others in the environment treat as obvious. And it runs on plausibility rather than accuracy. People do not adopt the most defensible account available. They adopt the most plausible one to hand, where plausibility is largely a function of identity fit and social supply.

This is the vulnerability that influence operations are built around, whether or not their designers have read Weick. Degrade the conditions under which sensemaking stays open. Wait for the frame to buckle. Supply a plausible account at the moment of collapse, delivered by a source that appears to be part of the in-group. Recruitment performs the same sequence on one person at a time and with more patience.

Sensemaking is the process under attack. The question for defenders is what keeps that process open when conditions are pushing it shut.

The evidence for curiosity

The most useful finding in this area comes from a study that was not designed to address influence operations at all. Kahan et al. (2017) developed a behavioural measure of science curiosity, capturing the disposition to seek out scientific information for its own sake rather than knowledge of scientific facts. Science literacy, they found, widened polarisation on contested issues; better-informed partisans were further apart, not closer, because greater capability was being used to defend prior commitments more effectively. Science curiosity behaved differently. Curious individuals across the political spectrum were more willing to expose themselves to evidence that cut against their existing position and the polarisation gradient flattened rather than steepened.

That result deserves more weight than it has been given in the security community. It identifies the counter-agent as a motivational disposition rather than a stock of knowledge. Capability without the appetite to use it against oneself makes matters worse.

Pennycook and Rand (2019) point in a compatible direction from another angle. Susceptibility to partisan falsehood in their data was better explained by under-engagement of analytic reasoning than by motivated reasoning in its strong form. People fall for material they would reject if they paused over it.

The intervention literature is more equivocal than its headlines. Prompting people to consider accuracy improved the quality of what they shared in Pennycook et al. (2021), though a preregistered direct replication of the closely related 2020 study found a markedly smaller effect and questioned how much work a minimal attentional nudge can do (Roozenbeek, Freeman and van der Linden, 2021). Guess et al. (2020) found that a digital media literacy package improved discernment between mainstream and false headlines and the pattern across their three samples is instructive; the US effect attenuated by more than half within about three weeks while remaining measurable, the India online effect could no longer be distinguished from zero at the second wave and the face-to-face Indian sample showed no effect at any point. Roozenbeek et al. (2022) demonstrated that short inoculation videos improved recognition of manipulation techniques at platform scale, though the field estimate of d = 0.09, taken a median of 18.4 hours after viewing, narrowly missed the authors' own preregistered smallest effect size of interest.

Protection also decays at rates that differ by format and the decay literature is partly an artefact of how decay is measured. Maertens et al. (2021) found gamified inoculation stable over three months only where participants were reassessed at intervals and decaying to non-significance within two months without them. Maertens et al. (2025) report that text-based inoculation held for roughly a month unboosted while gamified and video-based effects lost significance within a fortnight and they identify post-testing itself as one of the mechanisms that sustains them. Decay is a design problem rather than a refutation. What the design has to sustain, though, is not the skill. It is whatever keeps reactivating it.

Correction also fails less comprehensively than the counter-disinformation debate assumes. Wood and Porter (2019), across five experiments and more than ten thousand participants, found no evidence of the backfire effect on polarised issues and considerable evidence that people update towards accurate information when it is supplied. The problem with corrections is narrower and harder. They move belief. They do not touch the motivational state that made the false account attractive, nor the conditions that will make the next one attractive.

Measured against the scale of public investment, these are real effects with short half-lives, delivered through a skills architecture. A skill that is never reached for is a dormant capability and what determines whether it is reached for is motivation.

The harder question; how to encourage curiosity

Here the literature imposes a caution that is easy to miss. Curiosity is not one thing and one of its forms is precisely what hostile ecosystems feed on.

Loewenstein (1994) framed curiosity as a response to a perceived gap between what one knows and what one wants to know, a gap that produces a deprivation-like state which information closes. Litman (2005) drew the practical distinction; interest-type curiosity, oriented towards the pleasure of discovery and deprivation-type curiosity, oriented towards relieving the discomfort of not knowing. Conspiracy belief is driven in large part by epistemic, existential and social motives, by the need to understand, to feel secure and to maintain a valued image of one's group (Douglas, Sutton and Cichocka, 2017). A pipeline built to serve those motives opens gaps deliberately, advertises itself as the only source willing to close them and rewards the seeker with certainty rather than with a better question. Telling a susceptible population to be more curious, without qualification, risks accelerating the very process one is trying to interrupt.

What defence requires is interest-type curiosity coupled with a tolerance for gaps that stay open. That combination is buildable and it is mostly built by institutions rather than by campaigns.

Make uncertainty survivable. Where saying "I do not know yet" carries a status cost, premature closure becomes the rational choice. This holds in classrooms, newsrooms, operations centres and boardrooms alike. Leaders who narrate their own unresolved questions out loud are doing cognitive defence work, whether they describe it that way or not. Organisations that treat an ambiguous report as a contribution rather than a nuisance are doing the same. The test is simple; what happens, socially, to the person who raises something they cannot yet explain.

Reward the question rather than the answer. Education systems optimised for retrieval produce people who are good at closing gaps quickly, which is the trained reflex an influence operation is counting on. Assessment that credits the quality of a question, the identification of what would change one's mind or the ability to hold two accounts in play while evidence accumulates, trains the opposite reflex. This is a curriculum argument more than a counter-terrorism one, which is partly why it has been so hard to get funded.

Pair curiosity with intellectual humility. Leary et al. (2017) found that recognising the fallibility of one's own beliefs is associated with more careful attention to the quality of evidence and with less derogation of those holding opposing views. Curiosity without that companion disposition is appetite and appetite is steerable. Some of the most committed extremists are voracious readers.

Give the question a social home. Recruitment succeeds in large part because it takes the recruit's questions seriously when nobody else will and supplies a community in which asking is a form of belonging. Braddock and Horgan (2016) make the case that counter-narratives work through story and identity rather than through correction and the same logic applies upstream. A correction delivered by an out-group source competes badly against an invitation delivered by an apparent in-group. Youth provision, mentoring, sport, faith settings and online spaces that will sit with an unresolved question all function as cognitive defence infrastructure and are almost never funded as such.

Train the response at the point of contact. The moment that matters is rarely the moment someone encounters a false claim. It is the moment they bring it to a teacher, a line manager, a safeguarding lead or a parent and discover what asking costs. A correction closes the conversation and moves the enquiry somewhere it will be welcomed. An answer that treats the question as reasonable, takes the underlying concern seriously and leaves the matter open keeps the person in a setting where evidence can still reach them. Frontline practitioners are already making this call daily, mostly without training and often with institutional incentives pushing them towards correction because correction is what can be recorded as an action taken.

Measure it. Dispositional measures of curiosity and epistemic motivation exist and are usable. If curiosity is the active ingredient, programmes intended to build resilience should be reporting movement in it, rather than reporting reach, impressions and message recall.

These principles apply inwardly as much as outwardly. Security functions are themselves sensemaking organisations operating under time pressure and the seizing and freezing dynamic does not exempt analysts. An intelligence culture that penalises revised assessments, or that treats a withdrawn judgement as a professional failure, is manufacturing closure inside its own walls while campaigning against it outside them. Weick's Mann Gulch crew were trained professionals. What they lost was the shared frame and they lost it under exactly the conditions that a modern operations centre reproduces routinely.

Limits worth stating

The evidential base for the central claim is thinner than the base for the interventions it criticises and the asymmetry should be conceded rather than managed. Kahan et al. (2017) is a more substantial paper than a single correlational finding and a weaker foundation than it first appears. Its information-selection experiment set a surprising counter-attitudinal story against an unsurprising congenial one, so in the comparison it reports, novelty and slant move together. Kahan's own hypothesis is that the science-curious will prefer the surprising story regardless of its relationship to their predispositions, which leaves the study unable to distinguish an appetite for disconfirmation from an appetite for surprise. The design could separate them, since the surprising story carries the climate-sceptical finding in one condition and the non-sceptical finding in the other, so for roughly half the subjects the surprising story was the congenial one. The reported analysis gives probabilities only for the uncongenial-surprising choice and the contrast is not presented. The paper does anticipate a neighbouring objection, that curious subjects seek out uncongenial evidence in order to argue against it and answers that subjects highest in science comprehension, who have most reason to counterargue, showed no such preference either way. Counterarguing was never measured, so that answer runs by inference from a null. The estimate is also imprecise. The headline contrast is a gap of 20 percentage points with an interval of plus or minus 19, which clears zero and barely. Curiosity was measured throughout and never manipulated, so the causal direction is assumed. And nothing in that literature establishes that science curiosity, or any near relative, can be raised at population scale by deliberate intervention. Prebunking and media literacy are held here to a decay standard that curiosity has never been asked to meet, because nobody has yet run the trial that would ask it. On the present evidence, curiosity is the most promising untested candidate for the active ingredient. Treating it as established would repeat the error the article is complaining about.

Nor does it immunise anyone. Curiosity operates on the probability that a frame stays open for a little longer under pressure, which is a real effect and a modest one. It does no work at all against coercion, against material grievance or against the sort of targeted grooming that recruits through relationship rather than through argument. It cannot be surged; an epistemic disposition is built over years and no communications budget released six weeks before an election will produce it.

There is also a governance problem the security sector has not confronted. Interventions that shape how a population thinks, rather than what it is told, sit closer to the state's relationship with its citizens than most counter-disinformation activity does. The obvious answer, locating the work in education, civil society and the professions rather than in the security apparatus, only partly holds; a state curriculum shaping epistemic disposition is still the state shaping how its citizens think, with a longer lead time and considerably less visibility than a communications campaign. What distinguishes it is that the disposition being built is one that can be turned on the builder. That is a real distinction and a thin one and it deserves to be argued rather than assumed.

The reframe

The standing question asked of a population under cognitive attack is what it believes and the standing answer is measured in poll movement and narrative penetration. Both are lagging indicators. The leading indicator is whether people are still asking anything at all and whether the institutions around them make asking safe, social and rewarding.

An adversary can manufacture closure faster than any state can manufacture truth. What a state can do, over a longer horizon than any campaign cycle accommodates, is raise the cost of closure and lower the cost of the open question.

References

Braddock, K. and Horgan, J. (2016) 'Towards a guide for constructing and disseminating counternarratives to reduce support for terrorism', Studies in Conflict & Terrorism, 39(5), pp. 381–404. doi;10.1080/1057610X.2015.1116277

Douglas, K.M., Sutton, R.M. and Cichocka, A. (2017) 'The psychology of conspiracy theories', Current Directions in Psychological Science, 26(6), pp. 538–542. doi;10.1177/0963721417718261

Guess, A.M., Lerner, M., Lyons, B., Montgomery, J.M., Nyhan, B., Reifler, J. and Sircar, N. (2020) 'A digital media literacy intervention increases discernment between mainstream and false news in the United States and India', Proceedings of the National Academy of Sciences, 117(27), pp. 15536–15545. doi;10.1073/pnas.1920498117

Hogg, M.A. (2007) 'Uncertainty–identity theory', in Zanna, M.P. (ed.) Advances in Experimental Social Psychology, vol. 39. San Diego, CA; Academic Press, pp. 69–126.

Home Office (2026) Individuals referred to and supported through the Prevent Programme, October 2024 to September 2025. Official statistics in development. London; Home Office. Available at; https;//www.gov.uk/government/statistics/individuals-referred-to-prevent-to-september-2025 (Accessed; 11 September 2026).

Kahan, D.M., Landrum, A., Carpenter, K., Helft, L. and Hall Jamieson, K. (2017) 'Science curiosity and political information processing', Political Psychology, 38(S1), pp. 179–199. doi;10.1111/pops.12396

Kruglanski, A.W. and Webster, D.M. (1996) 'Motivated closing of the mind; "seizing" and "freezing"', Psychological Review, 103(2), pp. 263–283. doi;10.1037/0033-295X.103.2.263

Kruglanski, A.W., Gelfand, M.J., Bélanger, J.J., Sheveland, A., Hetiarachchi, M. and Gunaratna, R. (2014) 'The psychology of radicalization and deradicalization; how significance quest impacts violent extremism', Political Psychology, 35(S1), pp. 69–93. doi;10.1111/pops.12163

Leary, M.R., Diebels, K.J., Davisson, E.K., Jongman-Sereno, K.P., Isherwood, J.C., Raimi, K.T., Deffler, S.A. and Hoyle, R.H. (2017) 'Cognitive and interpersonal features of intellectual humility', Personality and Social Psychology Bulletin, 43(6), pp. 793–813. doi;10.1177/0146167217697695

Litman, J.A. (2005) 'Curiosity and the pleasures of learning; wanting and liking new information', Cognition and Emotion, 19(6), pp. 793–814. doi;10.1080/02699930541000101

Loewenstein, G. (1994) 'The psychology of curiosity; a review and reinterpretation', Psychological Bulletin, 116(1), pp. 75–98. doi;10.1037/0033-2909.116.1.75

Maclean, N. (1992) Young Men and Fire. Chicago, IL; University of Chicago Press.

Maertens, R., Roozenbeek, J., Basol, M. and van der Linden, S. (2021) 'Long-term effectiveness of inoculation against misinformation; three longitudinal experiments', Journal of Experimental Psychology; Applied, 27(1), pp. 1–16. doi;10.1037/xap0000315

Maertens, R., Roozenbeek, J., Simons, J.S., Lewandowsky, S., Maturo, V., Goldberg, B., Xu, R. and van der Linden, S. (2025) 'Psychological booster shots targeting memory increase long-term resistance against misinformation', Nature Communications, 16, article 2062. doi;10.1038/s41467-025-57205-x

Pennycook, G. and Rand, D.G. (2019) 'Lazy, not biased; susceptibility to partisan fake news is better explained by lack of reasoning than by motivated reasoning', Cognition, 188, pp. 39–50. doi;10.1016/j.cognition.2018.06.011

Pennycook, G., Epstein, Z., Mosleh, M., Arechar, A.A., Eckles, D. and Rand, D.G. (2021) 'Shifting attention to accuracy can reduce misinformation online', Nature, 592, pp. 590–595. doi;10.1038/s41586-021-03344-2

Roozenbeek, J., Freeman, A.L.J. and van der Linden, S. (2021) 'How accurate are accuracy-nudge interventions? A preregistered direct replication of Pennycook et al. (2020)', Psychological Science, 32(7), pp. 1169–1178. doi;10.1177/09567976211024535

Roozenbeek, J., van der Linden, S., Goldberg, B., Rathje, S. and Lewandowsky, S. (2022) 'Psychological inoculation improves resilience against misinformation on social media', Science Advances, 8(34), eabo6254. doi;10.1126/sciadv.abo6254

Weick, K.E. (1993) 'The collapse of sensemaking in organizations; the Mann Gulch disaster', Administrative Science Quarterly, 38(4), pp. 628–652. doi;10.2307/2393339

Weick, K.E., Sutcliffe, K.M. and Obstfeld, D. (2005) 'Organizing and the process of sensemaking', Organization Science, 16(4), pp. 409–421. doi;10.1287/orsc.1050.0133

Wood, T. and Porter, E. (2019) 'The elusive backfire effect; mass attitudes' steadfast factual adherence', Political Behavior, 41(1), pp. 135–163. doi;10.1007/s11109-018-9443-y