I. Beginning with an abused word
The word hype is a self-referential trap. When we say that something “is hype,” we are making an implicit epistemological claim: that we can distinguish an object’s real value from the false value attributed to it. The arrogance of that claim far exceeds most people’s awareness of making it, because it presupposes an Archimedean point—some cognitive position uncontaminated by hype, from which we can survey the whole and pass a cool judgement of value.
But does such a point exist? When a venture capitalist says AI is over-hyped, on what cognitive basis does the judgement rest? When a technological optimist says AI’s potential is under-hyped, where is he standing? Two people work from the same information, reach opposite conclusions, and each is convinced he is the clear-eyed one who has seen through the hype.
This is the deepest strangeness of the phenomenon: the identification of hype may itself be another form of hype.
The aim here is to raise “hype” from an everyday term of abuse into a serious analytical concept. I will argue that hype is a structural, ineliminable, and under certain conditions functional feature of human collective cognition—and that understanding its mechanisms matters far more than simply being “anti-hype” or claiming to see through it.
II. What is hype, exactly?
2.1 A working definition
We need a definition precise enough to work with. I would put it this way:
Hype is an overproduction of collective expectations, in which the narrative construction of an object’s future value departs systematically from the currently verifiable boundary of its capability.
Several elements deserve unpacking.
First, “collective expectations.” The subject of hype is always plural. One person’s excessive enthusiasm is a private bias; only when that enthusiasm becomes a shared emotional and cognitive state through mechanisms of social transmission—media, social networks, conferences, investor roadshows—does it become hype. Durkheim’s collective effervescence applies here with some discomfort: at its peak, hype is phenomenologically almost isomorphic with the collective ecstasy of religious ritual.
Second, “overproduction.” The metaphor is economic. Just as Marx located the core of capitalist overproduction crises in a structural imbalance between production and effective demand, the overproduction in hype is a spillover of expectations relative to current realizability. The qualifier current is crucial: many hyped technologies did eventually deliver their wild early promises, only on a badly compressed timeline. The online shopping, video calls, and instant access to information imagined during the dot-com bubble all exist today—delayed by ten to fifteen years.
Third, “narrative construction.” Hype never travels as a table of data. It travels as a story. Robert Shiller’s Narrative Economics (2019) analyses this well: economic fluctuations are often driven by contagious narratives whose transmission dynamics resemble epidemiological models rather than the uniform diffusion of information assumed by rational expectations theory. A narrative that “AI will replace all white-collar work within five years” propagates far faster than a factual statement about where current LLMs sit on a particular benchmark.
2.2 Distinguishing hype from its neighbours
To sharpen the concept, it helps to separate hype from three things it is often confused with.
Hype ≠ optimism. Optimism is a disposition toward positive expectation about the future; it may be grounded or ungrounded. Hype specifically denotes collective expectation that is amplified and accelerated by social transmission, appreciating in value as it spreads. An engineer’s optimism grounded in deep technical understanding and the excitement of someone who has never read a paper but is surrounded on Twitter by an “AGI is imminent” narrative are epistemologically quite different in texture, even if their outward behaviour—buying AI-related stocks, say—is identical.
Hype ≠ bubble. A bubble is an economic concept: the systematic departure of asset prices from fundamental value. Hype can generate bubbles, but its extension is far wider. Cultural hype, technological hype, and political hype can all exist without any financial market involvement. The “phenomenal anticipation” around a film’s release is hype but involves no bubble. Conversely, participants in certain financial bubbles—the CDO market of 2007, for instance—were often aware that it was a bubble but unwilling to leave “before the music stopped,” which is closer to a coordination failure in game theory than to hype in the classical sense.
Hype ≠ propaganda. Propaganda has a definite agent and a definite intent. What is uncanny about hype is that it is typically a subjectless process. Of course many hypes contain conscious boosters—founders, media, key opinion leaders—but the dynamics as a whole cannot be reduced to the intent of any single actor. As the inverse of the invisible hand, hype is an invisible mouth: everyone is only saying what they take to be true or interesting, and these local utterances aggregate into a self-reinforcing torrent of narrative whose direction and force are under no individual’s control.
2.3 The ontological status of hype
Which raises a more fundamental philosophical question: is hype real?
In an important sense, yes. Hype may rest on a misjudgement of the future, but as a social-psychological phenomenon it is entirely real, and it produces entirely real consequences. The Thomas theorem applies: if men define situations as real, they are real in their consequences.
This is most conspicuous in finance. George Soros’s theory of reflexivity provides the most precise framework. In Soros’s model there is two-way causation between participants’ cognitive function and the participating function through which they act on the situation. When enough people believe a technology will change the world, capital flows in, talent clusters, infrastructure gets built—and all of that raises the probability that the technology really will change the world. Hype is not merely a (possibly mistaken) prediction about the future; it is a force that alters the future.
Which yields hype’s deepest paradox: a “false” collective belief can, by mobilising enough resources, make itself “true.”
III. The dynamics of hype: a machine that feeds itself
3.1 The Gartner curve and its limits
Any discussion of hype dynamics has to pass through Gartner’s Hype Cycle, proposed in 1995, which divides a technology’s life into five phases: Innovation Trigger, Peak of Inflated Expectations, Trough of Disillusionment, Slope of Enlightenment, Plateau of Productivity.
The model’s virtue is its intuitive plausibility: almost everyone can recognise the shape from experience. Blockchain, VR, the metaverse, 3D printing, nanotechnology—every technological concept seems to find its place somewhere on the curve.
As an analytical instrument, however, it has several fundamental defects.
First, it is descriptive and lacks explanatory power. It tells you the shape of the curve without telling you why the curve has that shape. Why do expectations first overshoot and then undershoot? What are the underlying micro-mechanisms? On this the model is silent.
Second, it assumes a single-peaked, deterministic trajectory. Reality is more complicated. Many technologies go through several hype cycles—AI alone has had at least three, in the 1960s, the 1980s, and the 2010s—each with a different shape, amplitude, and frequency. Others were forgotten entirely before reaching any plateau.
Third, and most important, it smuggles in a teleological assumption: as though every technology were destined to pass through all five phases and arrive at a rational plateau. This narrative quietly defines hype as an irrational stage that must be overcome, an unfortunate detour on the way to a rational terminus. But what if hype is an intrinsic component of how the system runs?
3.2 Micro-mechanisms: information cascades and social proof
To supply the explanatory power Gartner lacks, we need to descend to the micro-mechanisms of hype formation.
The 1992 paper on informational cascades by Sushil Bikhchandani, David Hirshleifer, and Ivo Welch provides the first piece. Their model shows that in a sequential decision environment, even if every individual is Bayesian-rational, it takes only a few early decision-makers happening to choose in the same direction for later decision-makers to rationally ignore their own private signals and follow their predecessors. Once such a cascade forms, it produces extreme informational fragility: the behaviour of an entire population may rest on a very small quantity of initial information.
Map this onto hype. When a handful of influential early voices—top-tier venture funds, well-known academics, technology media—express strong optimism about a technology, subsequent observers face an asymmetry. They cannot tell whether those early voices rest on deep technical assessment or on limited information plus cognitive bias. But under the logic of social proof, “so many smart people believe in it” is itself a powerful signal.
Robert Cialdini treats social proof as one of the most powerful decision heuristics available to us. In domains of high uncertainty—and the future value of a new technology is a paradigm case—reliance on social proof rises sharply. The result is a positive feedback loop:
the higher the uncertainty → the greater the reliance on others’ judgement → the more easily consensus forms → the more easily the consensus that forms departs from the true value.
3.3 Viral narrative transmission
Cascades explain why consensus forms easily, but not the characteristic overheating of hype. For that we need the dynamics of narrative transmission.
Shiller borrows the epidemiological SIR model to analyse the spread of economic narratives. Transmission depends on three parameters: the contagion rate (the probability an “infected” individual passes the narrative on), the recovery rate (the probability an individual stops transmitting it), and the initial infected base.
The key insight is that contagion rates differ. Narratives with the following features spread fastest.
- High emotional arousal. “AI could put three hundred million people out of work within a decade” provokes far more fear and excitement than “AI has reached human-level accuracy on a specific task.”
- Identity provision. “Are you someone who understands the future, or someone the future will discard?” Binary narratives of this kind bind a person’s technological position to their self-conception, which raises stickiness enormously.
- Simplicity and retellability. “The metaverse is the next internet” travels across a dinner table far more easily than any technical white paper. And each retelling simplifies, dramatises, and sharpens it—what folklorists studying serial reproduction have long described.
- Time pressure. “The window is short.” “First-mover advantage is decisive.” “If you don’t move now it will be too late.” Scarcity narratives exploit loss aversion.
When a narrative has all four at once, its contagion rate becomes very high while its recovery rate is pressed very low—because exiting the narrative means losing an identity and incurring the fear of missing out. This is the overheating mechanism: positive feedback in narrative transmission inflates collective expectations far faster than information actually accumulates.
3.4 Feedback loops and self-fulfilment
There is a further layer. As Soros’s reflexivity implies, hype does not merely reflect reality; it alters it.
Take AI. Once the “AI will change everything” narrative reaches sufficient intensity, the following begin to happen:
- capital pours in, so more research gets funded and more companies get financed;
- talent flows toward AI, accelerating the actual rate of technical progress;
- governments come to see AI as a focus of strategic competition and issue supportive policy;
- universities expand enrolment in AI-related programmes;
- public acceptance of AI products rises, lowering the cost of market entry.
These real changes then supply the hype narrative with fresh “evidence”: look, the capital is flowing in; look, the best people are switching fields; look, even governments are paying attention. A self-fulfilling prophecy takes shape—though the degree and direction of that fulfilment may diverge widely from the original narrative.
Which is hype’s most unsettling epistemological predicament: while you are inside the rising phase, all the evidence points to the reasonableness of the hype, because the hype is manufacturing the evidence. You cannot judge whether hype is excessive by observing current evidence, because you cannot separate “capital flowed in because the technology is genuinely good” from “capital flowed in because the narrative said the technology is good—and the inflow then made the technology look better.”
3.5 What triggers collapse
If the positive feedback loop is so powerful, why does it ever collapse?
Because the stability of a positive feedback system depends on a continuing inflow of new “infections.” Once the susceptible population is exhausted—everyone persuadable by the narrative has been persuaded—the increment of transmission begins to fall. Meanwhile the negative feedback has never gone away; it has only been suppressed:
- Accumulated reality gaps. Early promises come due. Products fall short. The ROI does not materialise. The use cases do not appear. Every gap is a small seed of counter-narrative.
- Defectors. Some early participants begin to voice doubt in public, and their defection carries asymmetric weight: a former believer saying “I was wrong” is more contagious than a lifelong sceptic saying “I told you so.”
- Competition for attention. Collective attention is finite. A new hype topic drains energy from the old one.
When the increment of positive feedback falls below that of negative feedback, the system reverses. And because the same positive-feedback machinery now runs backwards—“the smartest people are leaving,” “investors are pulling out,” “the press has turned”—the descent is usually faster than the ascent. That is why the trough of disillusionment on the Gartner curve tends to be steeper than the peak.
IV. The political economy of hype: who produces it, who consumes it?
4.1 The production side
The emphasis above on hype as a subjectless process does not mean there is no structural distribution of interests inside it. Hype has a clear political economy.
Venture capital is the most important institutionalised producer of modern hype. The VC business model depends intrinsically on producing it. A power-law return structure means a fund needs a few investments to return enormously in order to cover many failures. To return enormously, a portfolio company must raise later rounds at higher valuations and eventually exit through IPO or acquisition. And higher valuations depend heavily on the market’s expectations about the future growth of that company’s category.
In other words, VCs have a structural incentive to produce and maintain hype. This requires no individual VC to be insincere: one can sincerely believe the category will change the world while objectively functioning as a transmission node for hype. There is no contradiction between institutional logic and personal sincerity.
Technology media is the second great producer, and its incentives deserve the same dissection. Media business models depend on capturing attention, and in the attention economy the mild narrative—“this technology may develop somewhat over the next decade”—never beats the violent one—“this technology is about to upend everything.” The result is a systematic selection bias toward amplifying the most extreme voices, because extreme voices generate the most clicks, shares, and engagement.
Note that this bias is symmetric: media amplify the rising phase (“AI will change everything”) and the falling phase (“the AI bubble is about to burst”) alike. Their interest is not in sustaining any particular direction of narrative but in sustaining its extremity. Moderate middle positions do not generate traffic.
Academia plays an ambiguous role. On one hand it positions itself as a rational producer of knowledge, immune to hype. On the other, grant funding requires researchers to argue for the transformative potential and broader impact of their work, which institutionally rewards systematic exaggeration of significance. Daniel Sarewitz (2016), in his critique of the science funding system, put it sharply: the competitive structure of modern research funding is in effect a machine for converting scientific knowledge into promissory narratives.
4.2 The consumption side
Consumers of hype have their own complex motivational structure.
For the general public, hype supplies a cheap cognitive frame. In a world of information overload, genuinely understanding what a new technology means requires enormous cognitive investment. A hype narrative offers a shortcut: “you don’t need to understand the details of the transformer architecture, you just need to know that AI is about to change everything.” This cognitive offloading explains why hype spreads especially violently in highly specialised fields: the higher the barrier to expertise, the greater the dependence on simplified narrative.
For corporate decision-makers, hype supplies legitimacy. Under high uncertainty, “everyone is doing X” is itself the best reason to do X. This is not entirely irrational: in a world with network effects and ecosystem lock-in, “right direction, late start” can be more fatal than “perfect judgement, missed window.” Which is why FOMO is hype’s most faithful co-pilot.
For individual practitioners, hype supplies an identity narrative and a sense of career direction. “I work in AI” carries far more social identity value at the peak than “I am a database engineer,” even where the latter’s actual skills may be scarcer. This identity dimension makes exit especially hard: admitting that “the direction I have been chasing may not be as revolutionary as I thought” requires not merely a revision of opinion but a reconstruction of self-narrative.
4.3 Redistributive effects
A political economy of hype cannot avoid the basic question: who gains and who loses?
Broadly, hype produces a systematic redistribution of wealth and attention:
- From late participants to early ones. Most nakedly in finance: later investors enter at higher valuations and bear the largest losses if the hype collapses.
- From outsiders to insiders. Those who can produce hype narratives—investors, founders, industry analysts—typically get more information and better exit options than those who consume them.
- From the neglected alternatives to the hyped direction. This is the most hidden and possibly the most important cost: when resources pour into the hyped direction, fields that were never hyped but may be equally or more important go hungry. It is an invisible opportunity cost.
V. The deep historical structure of hype
5.1 Hype is not a modern phenomenon
If hype is the overproduction of collective expectations, its history long predates Silicon Valley and venture capital.
Dutch tulipmania (1637) is usually treated as the first well-documented speculative bubble. But Anne Goldgar’s revisionist history Tulipmania (2007) argues that the traditional account was itself enormously hyped by posterity: far fewer people actually speculated than later literature implies, and the price explosion was concentrated in a handful of rare varieties. A story about hype was itself hyped. This meta-level irony is nearly standard equipment for the phenomenon.
The South Sea Bubble (1720) offers a cleaner dissection. The company’s core narrative was that it held a monopoly on trade with South America—then regarded as a new world of inexhaustible riches—which would generate unlimited profit. That narrative had every high-contagion feature listed above: extreme emotional arousal (overnight wealth), identity provision (becoming a visionary of the new age), simplicity (“South America has endless gold and silver”), and time pressure (“the stock rises daily; buy today or pay more tomorrow”).
Newton’s famous loss in the South Sea Bubble—reportedly around £20,000, equivalent to several million today—is often cited to show that even the cleverest are fooled by hype. But the deeper lesson may be this: Newton’s genius equipped him to analyse deterministic systems, and hype is a reflexive social system. The tools that work on the former fail precisely here.
5.2 Why modernity intensifies hype
If hype is a structural feature of human cognition, why does contemporary hype feel more violent and more frequent than at any point in history?
The exponential acceleration of transmission. From the mail coach to the telegraph to radio to television to the internet to social media, the latency of information has been compressed to near real time, which sharply accelerates each iteration of the positive feedback loop. Tulipmania took years to peak; a crypto cycle can complete in months or even weeks.
Deepening financialization. The modern financial system has created ever more instruments for trading on expectations. Options, futures, leveraged ETFs, crypto tokens—each is a mechanism for converting expectation into a position, and the profit or loss on that position feeds back to reinforce or dissolve the expectation. The more financialized the system, the stronger the reflexive loop.
The extremity of specialisation and the cognitive division of labour. In a highly specialised society, no individual can independently evaluate knowledge claims across all fields. Dependence on “expert opinion” and “social consensus” is therefore higher than ever—which is exactly the breeding ground for cascades and for hype.
The democratisation (and industrialisation) of narrative production. Social media gave everyone the ability to produce and transmit narrative, and simultaneously created a large professional class whose occupation is narrative production—opinion leaders, content creators, industry analysts. The supply side expanded enormously while the quality-control mechanisms—editorial review in traditional media, peer review in academia—were sharply weakened.
VI. The epistemology of hype: can we get beyond it?
6.1 The counter-hype trap
The natural response to hype is counter-hype: stand opposite it and stay coolly sceptical. But counter-hype has epistemological traps of its own.
Counter-hype may be hype in reverse. When someone’s identity rests on being “the one who saw through the hype,” he acquires a systematic incentive to deny every optimistic signal. “AI is just statistics.” “Crypto is just a Ponzi scheme.” “EVs are just toys.” These are extreme simplifications of a complex reality, exactly like “AI will change everything,” “crypto will upend finance,” “EVs will kill the combustion engine.” Scepticism enjoys no epistemological privilege.
Nassim Taleb’s intellectual legacy makes an interesting case study here. His critique of the narrative fallacy and of overconfidence is genuinely penetrating, yet his writing style and public persona rest on a hype about being anti-hype: a self-narrative about being more clear-eyed than everyone else, whose transmission dynamics are nearly isomorphic with what it criticises.
Counter-hype makes errors of timescale. As noted, many hyped technologies did deliver on their early promises, only over a far longer horizon than the hype assumed. Roy Amara’s law captures it exactly: we tend to overestimate the effect of a technology in the short run and underestimate it in the long run. In this sense, “AI is over-hyped” may be right in the short run and wrong in the long run, while “AI will change everything” may be right in the long run and dangerous in the short. Correct analysis requires precise calibration of timescale—which happens to be one of the weakest points in human cognition.
6.2 A more mature epistemological posture
If neither embracing hype blindly nor opposing it wholesale is a good strategy, what is?
I favour a position that might be called participatory skepticism, with five elements.
First, acknowledge your own embeddedness. You are not outside the hype. Your sources, your social circle, your professional interests, your identity—all place you at a particular position in the field. Acknowledging that embeddedness is a more honest and more useful starting point than pretending to a view from nowhere.
Second, separate directional judgements from temporal ones. “Does this technological direction have value?” and “on what timescale will that value be realised?” are independent questions, and hype almost always conflates them. A technology can be simultaneously right in the long run and badly over-hyped in the short. Holding both requires cognitive flexibility.
Third, watch changes in material infrastructure more closely than changes in narrative. Narrative can flip overnight; material infrastructure—laboratories, factories, talent pipelines, regulatory frameworks, user habits—changes slowly and stickily. When you are trying to find signal in the noise, the most reliable indicators are usually these slow variables. Is real R&D spending rising or falling? Are key people flowing in or out? Is user retention rising or falling?
Fourth, think in probability distributions and resist the pull of binary narrative. “Either AI changes everything or it is just a bubble” is a false dichotomy. The truer picture is a distribution: different degrees of impact, in different domains, on different timescales, each with its own likelihood. Philip Tetlock’s work in Superforecasting (2015) shows that the best forecasters are precisely those most given to fine-grained probability estimates, least drawn to extreme positions, and most willing to update.
Fifth, keep metacognitive watch on your own degree of certainty. When you feel very sure that something is “definitely hype” or “definitely going to happen,” that feeling is itself a signal worth auditing. The intensity of certainty has no reliable correlation with the accuracy of judgement. Dunning–Kruger applies not only to assessments of competence but to assessments of hype: the more confident someone is about judging hype, the less they usually understand its complexity.
VII. The existential dimension: why do people need hype?
7.1 Hype as a secular religion
Raise the level of analysis one more step—out of sociology and economics into anthropology and ontology—and an uncomfortable but possibly important hypothesis emerges: hype satisfies something like a religious need.
Peter Thiel returns repeatedly in his public talks to a theme: the fundamental problem of the modern world is the exhaustion of imagination about the future. He argues that mid-twentieth-century Western society possessed a strong definite optimism—a belief in a concrete, plannable good future (the moon programme, the interstate highway system, the peaceful use of nuclear energy). That belief dissolved after the 1970s and was replaced by an indefinite optimism: the future will be better, but nobody knows how.
If part of Thiel’s diagnosis is right, hype can be understood as a compensatory response to the scarcity of definite optimism. In a “postmodern” world where the grand narrative has died (Lyotard, 1979), every round of technological hype is a brief resurrection of one. “AI will change everything.” “Blockchain will rebuild trust.” “The metaverse will surpass reality.” These narratives supply a sense of direction, of meaning, of a future one can get hold of—precisely what secular modernity finds it increasingly hard to provide.
Seen this way, the recurrence of hype cycles—including the fact that a new hype topic always arrives after each disillusionment—is not merely a product of market dynamics. It reflects a deep human hunger for a collective sense of purpose. That hunger is structural and ineliminable, and so, therefore, is hype.
7.2 Hype and temporality
Heidegger’s analysis of human temporality offers another angle. For Heidegger, human existence is essentially directed toward the future (Sein-zum-Tode, being-toward-death). We are always already projecting (Entwurf); our present actions, moods, and judgements are permeated by anticipation of what is to come.
Hype can be read as a collective expression of that ontological future-directedness. Human beings do not merely happen to be prone to excessive expectation about the future; in a sense, expectant projection toward the future is a basic structure of human existence. Hype is that structure developing, like a photograph, under particular social and historical conditions.
This does not mean we should embrace hype or abandon criticism. It means that criticism of hype cannot stop at “people are too easily fooled.” Behind the being-fooled is a fundamental need for meaning, direction, and hope, and that need does not disappear because you have pointed out the irrationality. Any project to eliminate hype that cannot supply an alternative mechanism for producing meaning is bound to fail.
VIII. Coda: living with hype
Return to where we began: the identification of hype may itself be another form of hype.
Having written this far, I have to admit a self-reflexive predicament. Is this essay not also a hype? A narrative about deeply understanding hype, offering in academic wrapping a feeling of cognitive superiority, so the reader may feel clearer-eyed than all those people swept along by it?
Yes. To some degree, yes.
But admitting it does not invalidate the analysis. It only reminds us that there is no Archimedean point with respect to hype. We are permanently embedded in the system we are trying to analyse. This is not a defect of cognition; it is the condition of cognition.
A mature intellectual posture might look like this.
You know how hype works. You know about information cascades, social proof, narrative transmission, reflexive loops. You know your own judgement is contaminated by your social position and your identity. You know that being anti-hype can be as wrong as embracing it. You know that all of this knowing may not help you judge any better.
And then you still have to act under uncertainty.
You still have to make decisions about direction, timing, and the commitment of resources, with incomplete information, with signal hard to separate from noise, and with a future that is not in principle predictable. All you can do is stay alert to the limits of your own cognition, distrust the feeling of certainty, and keep a respect for complexity.
To adapt the famous prayer from Reinhold Niebuhr:
Grant me the courage to join the hypes worth joining, the wisdom to avoid the ones that are collapsing, and the insight to know the difference.
Insight, of course, may also be over-hyped.