I am, by temperament, allergic to first numbers. When someone tells me an apartment is worth €200,000, my immediate reaction is to assume it is worth €150,000, because anyone offering me a number on first contact has a reason for offering that specific number, and the reason is rarely my benefit. I do not take prices at face value. I do not take expert opinions at face value. I do not even take meta-analyses at face value, which is why I read three of them before forming a view about anything that matters. This is not a virtue. It is a trait that has, on at least one documented occasion, made a real estate agent give up and suggest I might be more comfortable purchasing a house from a different person.
So when I tell you that anchoring is the most important cognitive bias in human history, you should know that I am the kind of person who came to that conclusion through aggressive contrarianism. I have spent my life trying not to anchor. I have read more skeptically than most people, demanded primary sources earlier than most people, and refused to accept the first explanation offered for almost anything. And here is what I have learned, after fifteen years of this approach: I am still anchored. Not on the first number. On the second-guess. On the conviction that the first number is wrong, which is its own anchor, established the moment I decided that all first numbers are suspect. The skeptic is not free of anchoring. The skeptic is anchored on skepticism itself, which is, at minimum, a more interesting anchor than the alternative, but anchored nonetheless.
What I have started to see, when I look around, is that almost everyone is anchored on something — and the something is rarely the thing they think they are reasoning from. My colleagues, when arguing about technical decisions, are usually arguing from the framework of whoever taught them their first programming language fifteen years ago. People in my industry, when arguing about AI, are usually arguing from the position they staked out publicly during the last hype cycle, even when the technology has changed enough that the position no longer maps onto the territory. People in politics, almost without exception, are arguing from the political identity they formed in their twenties, and the arguments themselves are post-hoc rationalizations of an emotional commitment that predates their access to evidence. I am no different. I am just anchored on different things, in different ways, with the small unearned advantage of knowing — sometimes — that this is what is happening.
Once I started seeing this pattern, I could not unsee it. Once a system establishes a reference point — a price, a worldview, an ideology, a power structure — it cannot easily move off it. Not because moving is impossible, but because the system has reorganized itself around the reference. The institutions, the incentives, the people promoted under the old logic, the stories told about why things are this way: all of them anchor the present to a past that may no longer be relevant. We are not designed to ask whether the anchor is correct. We are designed to ask whether each new piece of information is consistent with the anchor we already have.
This is, structurally, why nations fail to reform themselves. Why dictatorships do not voluntarily liberalize. Why the rich tend to get richer and the poor tend to stay where they were placed several generations earlier. Why sociopaths rise disproportionately into leadership and stay there once they arrive. Why religions splinter into a thousand sects each convinced the others have misread the same text. The mechanism is the same. The scale changes. The pattern is so consistent across human institutions that I find it hard to believe it’s a coincidence, and easy to believe that no one in particular is responsible — which is, paradoxically, the most disturbing part.
Through the lens of cognitive psychology
The phenomenon I have been describing has a formal name and a substantial scientific literature. Amos Tversky and Daniel Kahneman first documented it in their landmark 1974 paper in Science, “Judgment under Uncertainty: Heuristics and Biases.” They asked subjects to spin a wheel of fortune that landed on a random number — say, 65 — and then to estimate what percentage of UN member states were African nations. The median answer was 45%. When the wheel landed on 10, the median answer was 25%. The wheel was visibly random. Everyone knew it was random. The number affected the answer anyway, by a factor of nearly two.
What this means, when you sit with it, is that human reasoning is not a process of starting from evidence and arriving at conclusions. It is a process of starting from whatever happens to be in your head first and adjusting outward, usually insufficiently, in the direction the evidence pushes. The first number wins, statistically speaking, even when the first number is admitted to be meaningless. We do not have a mental “reset” button. We have an “amend” button, and the amendment is always partial. The effect has been replicated in hundreds of experiments across decades, in domains ranging from legal sentencing to medical diagnosis to salary negotiation, and it is one of the most robust findings in cognitive psychology — robust enough that it survived the replication crisis that has demolished other classic findings.
What I find interesting, and slightly humbling, is that even the skeptical strategy I described in the opening — refusing to accept the first number, assuming it is wrong, doing my own research — does not actually escape the anchoring effect. It just shifts the anchor. The skeptical position is itself an anchor. The decision that the first number is suspect is anchored on the framework I developed at some earlier point about how negotiation works, who can be trusted, what kinds of people offer numbers and why. Those frameworks are themselves anchored on earlier experiences, earlier teachers, earlier disappointments. There is no anchor-free position. There is only the question of which anchor you have chosen and whether you can articulate it.
Solomon Asch’s conformity experiments at Swarthmore in the 1950s — published in Scientific American in 1955 and replicated extensively — extended this finding into the social domain. Asch showed that participants would deny the evidence of their own eyes (incorrectly identifying which of three lines matched a target line) when surrounded by confederates giving the wrong answer. Approximately one-third of responses conformed to the obviously incorrect group judgment. The group anchor overrode direct sensory evidence. Subsequent meta-analyses by Bond and Smith (1996) in Psychological Bulletin, covering over 130 studies across 17 countries, confirmed the basic finding while showing that conformity rates vary by culture and historical period. The mechanism is universal. Its strength depends on context.
The implication is that we are not, individually or collectively, the rational agents we imagine ourselves to be. We are anchor-following animals with an unusually elaborate vocabulary for justifying which anchors we follow. The vocabulary varies. The following does not.
Through the lens of political economy
Daron Acemoglu, Simon Johnson, and James Robinson were awarded the 2024 Nobel Prize in Economics for their work on how institutions shape long-run prosperity. Their central thesis, developed across decades of papers and most accessibly presented in Why Nations Fail (2012), distinguishes between two institutional types. Inclusive institutions — those that protect property rights, enforce contracts, allow broad political participation, and constrain elites — tend to generate sustained economic growth. Extractive institutions — those that concentrate power and wealth in a narrow elite while suppressing broader participation — tend to generate stagnation, instability, or growth that ends abruptly when the elite’s interests change.
The work has been hugely influential and is now a central pillar of development economics. It has also been substantially criticized — Arvind Subramanian, Jeffrey Sachs, and Morten Jerven, among others, have raised questions about the empirical strategy, the simplifications required to fit hundreds of historical cases into a binary framework, and the partial neglect of geography and culture as independent factors. The critics are not fringe; some are senior economists who broadly agree with the thesis but doubt some of the specific causal claims. This is healthy. It is also instructive: even the work that wins the Nobel Prize for explaining inequality between nations is contested by serious researchers in the same field. Anyone offering you a simpler story than the Nobel committee’s is selling something.
What is less contested is the underlying mechanism: institutions, once established, are extraordinarily difficult to change. Robert Michels, a German sociologist, formalized this in 1911 as the Iron Law of Oligarchy, in his book Political Parties. Michels studied actual democratic socialist parties — organizations explicitly founded on principles of equality and broad participation — and found that they reliably evolved toward oligarchic structures regardless of their stated ideology. The pattern, he argued, was not a failure of will but a structural inevitability: any organization complex enough to act effectively requires specialization, specialization creates expertise asymmetries, expertise asymmetries create power asymmetries, and power, once concentrated, organizes its own preservation. Subsequent research in organizational sociology has largely confirmed Michels’ empirical observation while debating his stronger claim that oligarchy is literally inevitable.
Thomas Piketty, in Capital in the Twenty-First Century (2014), provided what is now the most empirically dense documentation of the long-term dynamics of inequality. Working with the World Inequality Database — an open dataset built from tax records, national accounts, and surveys across multiple countries over centuries — Piketty showed that when the rate of return on capital (r) exceeds the rate of economic growth (g), wealth concentrates over time. The condition r > g has held for most of recorded economic history, with brief exceptions during the post-war period (1945–1980), when high growth, progressive taxation, and explicit redistributive policy temporarily flattened the curve.
Piketty’s work has been criticized — Acemoglu and Robinson themselves have argued his framework underweights institutions, and economists like Matthew Rognlie have proposed alternative explanations centered on housing wealth — but the underlying empirical observation is robust: in the absence of deliberate, sustained intervention, capital concentrates. Wealth begets wealth. Power begets power. The system, left alone, anchors on past advantage and amplifies it.
This is the political economy version of cognitive anchoring. The first allocation of wealth determines the next. The next determines the next. And the institutions that might, in theory, redistribute, are themselves staffed by people whose careers depend on the distribution they would have to redistribute. The system is anchored in itself. It can be moved, but only by sustained external force — usually crisis, sometimes revolution, rarely policy.
Through the lens of organizational psychology
Here is one of the more uncomfortable findings in modern psychology, and I want to be careful with it because it is also one of the most sensationalized.
Psychopathy, in clinical terms, is a constellation of traits including grandiosity, lack of empathy, manipulativeness, shallow affect, impulsivity, and absence of remorse. It is not a synonym for “evil person.” It is a measurable personality profile, assessed using validated instruments like the Psychopathy Checklist-Revised (PCL-R) developed by Robert Hare. The general population prevalence is approximately 1%, with substantial methodological variation in estimates.
Paul Babiak, Craig Neumann, and Robert Hare published a 2010 study in Behavioral Sciences & the Law assessing 203 corporate professionals identified by their organizations as high-potential management candidates. They found a prevalence of psychopathic traits roughly four times the general population — approximately 4%. Subsequent studies have produced similar estimates, generally falling between 3% and 5% in senior corporate samples. A 2021 systematic review and meta-analysis in Frontiers in Psychology synthesized prevalence estimates across populations and noted, with appropriate caution, that workplace samples in certain industries do show elevated rates compared to community samples — though the authors warned that some of these workplace findings rest on small studies and should be treated as suggestive rather than definitive.
The structural argument — developed by Babiak and Hare in Snakes in Suits (2006) and elaborated by Clive Boddy in subsequent work on what he termed “corporate psychopathy” — is that modern organizational selection systems are not neutral with respect to personality. They preferentially reward traits that overlap with subclinical psychopathy: confidence presented as competence, willingness to make decisions without emotional weight, ability to manipulate impressions, comfort with deception, and ruthlessness in resource competition. The same traits that would mark someone as dangerous in a personal relationship become assets in a corporate hierarchy where the metric is upward mobility.
This is not a claim that all leaders are psychopaths. Most are not. It is a claim that the selection pressure in modern organizations favors a particular trait profile, and that profile is overrepresented at the top relative to its base rate in the general population. The mechanism is anchoring at the institutional level: the first generation of leaders selected under these criteria becomes the anchor for what “leadership looks like,” and subsequent selections, by people anchored on the original example, reproduce the trait profile while telling themselves they are selecting for “decisiveness” or “executive presence” or whatever language has been laundered for the purpose.
There are, predictably, critics of this entire framework. Some argue the prevalence figures are inflated by methodological artifacts. Others argue that the personality trait clusters being measured are not psychopathy specifically but a more general “Dark Triad” (psychopathy, narcissism, Machiavellianism). The honest summary is: the effect is real, the magnitude is contested, and the mechanism — institutional selection that rewards certain traits — is consistent with a much larger literature in organizational behavior that doesn’t depend on any specific clinical diagnosis to make the point.
Through the lens of history
Hannah Arendt, in The Origins of Totalitarianism (1951), and Robert Paxton, in The Anatomy of Fascism (2004), produced what are now considered foundational analyses of how totalitarian movements actually emerge. Their accounts are not, strictly speaking, science — they are works of historical and philosophical analysis — but the regularities they describe have been confirmed by subsequent quantitative work in political science, including Steven Levitsky and Daniel Ziblatt’s How Democracies Die (2018) and the long-running V-Dem Institute project at the University of Gothenburg, which has documented democratic backsliding across over 200 countries since the 1700s.
Paxton’s central observation: fascism does not arrive as fascism. It arrives as a coalition between traditional conservative elites and a populist movement, in conditions of perceived crisis (economic, cultural, demographic), with rhetoric emphasizing national renewal, enemies within, and the failure of liberal institutions to solve problems. The traditional elites believe they can use the populist movement and discard it. They cannot. The movement, once anchored in power, organizes the institutions to preserve itself, and the elites who thought they were the riders discover they are the horse.
Levitsky and Ziblatt, drawing on quantitative analyses of democratic transitions, identify four warning signs that have preceded democratic erosion in dozens of cases: rejection of democratic rules, denial of opponents’ legitimacy, tolerance or encouragement of violence, and readiness to curtail civil liberties of opponents. Their work, while not without critics, has been influential precisely because it identifies pattern rather than predicting any specific case.
The V-Dem data, which is the most rigorous longitudinal dataset on democratic quality, shows a measurable global decline in liberal democracy since approximately 2010, with more countries currently moving toward autocracy than toward democracy for the first time in decades. The dataset is publicly available, methodologically transparent, and updated annually. Whether the trend continues, reverses, or stabilizes, we will know in time. What we already know is that the institutions we treat as permanent are, by historical standards, recent and unstable.
The deeper pattern, across all of this literature: nations and political systems do not fail because of single dramatic events. They fail through gradual erosion of institutional anchors, replacement of one reference point with another, until the system that was once a democracy is, while wearing the same clothes, something else. The anchor moved. The vocabulary stayed. By the time anyone notices, the mechanism that would have allowed correction has itself been captured by what would need to be corrected.
Through the lens of game theory
Thomas Schelling, who shared the 2005 Nobel in Economics for his work on conflict and cooperation, formalized in The Strategy of Conflict (1960) the dynamics of what he called focal points and coordination problems. Schelling’s key insight: in many strategic interactions, there is no objectively best solution, only a solution that everyone expects everyone else to converge on. The focal point — the anchor — becomes the equilibrium not because it is optimal but because it is salient.
This applies to wars in a way that I find darkly clarifying. The security dilemma, formalized by political scientist Robert Jervis in his 1978 paper “Cooperation Under the Security Dilemma” in World Politics, describes a situation where each party’s defensive measures appear threatening to others, triggering escalating defensive responses that produce the very outcome each party was trying to prevent. Wars often start not because anyone wants war but because each party becomes anchored on a worst-case interpretation of the other’s behavior. The interpretation, once anchored, organizes all subsequent evidence. The worst case becomes the case.
The same mechanism explains why cooperation is so structurally difficult in international relations: the absence of a higher authority means each actor must solve coordination problems through anchoring on focal points, and the focal points are usually inherited from previous conflicts. Hostility, once established, anchors on itself. So does cooperation, when it is established — which is why the rare periods of sustained international order tend to require either hegemonic stability (one dominant power enforcing the anchor) or institutional structures that create new focal points (the post-war Bretton Woods system being the most studied example).
The problem, of course, is that anchors can shift. The hegemon weakens. The institutions decay. The focal point that everyone has been coordinating on no longer holds. And then, in the absence of a clear new anchor, the security dilemma takes over, each party retreats to defensive posturing, and the world drifts toward conflict not because anyone wants it but because no one knows where the new equilibrium is. Wars, in this framework, are coordination failures performed at scale by people convinced they are acting rationally.
Which they are. Locally. The anchoring effect is rationality, applied to the wrong reference point.
Connections
Anchoring is the structural cousin of Overfitting: both are forms of giving past data too much weight relative to present evidence. It is the social form of Fragility: institutions optimized so completely around their existing reference points that any disturbance threatens systemic collapse rather than local adjustment. It is the political form of Entropy: systems drifting toward concentrated equilibria not because anyone wills it but because that’s the lowest-energy state given the existing constraints. And it is the historical form of Noise: the same patterns appearing across millennia in different costumes, with each generation convinced its costumes are uniquely tailored.
If there is a single mechanism underneath every concept this collection has explored, it might be this one: humans, individually and collectively, do not reason from first principles. They reason from anchors. The anchors are usually inherited. The reasoning is usually about defending the anchor while telling ourselves we are exploring freely. We are extraordinarily creative about our anchors. We are not, often, free of them.
What I Don’t Know
I haven’t engaged with cultural anthropology — the work of scholars like David Graeber and David Wengrow, particularly The Dawn of Everything (2021), which argues that human social structures have historically been far more varied than the inevitability framings of Michels and others suggest. Their thesis, that hierarchy and equality have flexibly coexisted across history, is contested but worth taking seriously as a counter to the iron-law literature. I haven’t looked at network theory and the formal study of how power concentrates in scale-free networks — Barabási and others have shown that the rich-get-richer dynamic is mathematically inevitable in many network structures, which would be a more rigorous formalization of what I’m calling anchoring. I haven’t engaged with the recent literature on what political scientists call “competitive authoritarianism” — regimes that maintain electoral institutions while hollowing out the substance — which would refine the totalitarianism analysis considerably. I haven’t addressed the religious dimension at all, despite mentioning it in the opening — Max Weber’s analysis of how charismatic authority routinizes into bureaucratic authority is structurally relevant, and I left it on the table because the article was already long. And I haven’t confronted the most uncomfortable question: whether writing about anchoring while being myself anchored — to certain authors, certain frameworks, certain political assumptions — is a meaningful exercise or just a more sophisticated form of the thing I’m describing.
Where I Stand
I notice anchoring constantly, in myself most of all. I notice when I give a number more weight because it was the first one I heard. I notice when I trust an argument more because it’s made by someone whose previous arguments I’ve trusted, even when this particular argument should be evaluated on its own. I notice when I dismiss claims faster from sources I’ve decided are unreliable, regardless of whether the specific claim is true. The noticing does not stop the anchoring. It just adds a small thin layer of awareness around the anchoring, which, on good days, lets me adjust slightly. On bad days, the awareness produces the smug feeling of having understood the bias, which is itself a form of anchoring on the meta-level. Self-awareness is not exemption.
What I have come to believe — and I hold this provisionally, knowing I may be anchored by my own recent reading — is that the most useful response to all of this is not to try to escape anchoring, which is biologically impossible, but to deliberately diversify the anchors. Read across disciplines. Read people you disagree with, especially the smart ones. Spend time in environments and conversations that would not naturally come to you, because every comfortable environment is anchoring you toward a particular equilibrium. This is, I now realize, exactly what Lenses is for me — not a publication strategy, but an anchoring management protocol disguised as a writing project.
On the larger questions — the dictatorships, the inequality, the wars, the sociopaths in leadership — I have stopped expecting moral outrage to fix any of them. Outrage is itself anchored. It triggers in response to specific cues and ignores structurally identical cases that don’t carry the cues. What appears to actually move institutions, on the available evidence, is sustained pressure applied through institutions themselves: courts, journalism, organized civil society, electoral systems where they still function. None of this is glamorous. None of it produces the satisfaction of having identified a villain. It does, in measurable cases, produce change — slowly, partially, often reversibly. The historical examples of large-scale institutional improvement (the abolition of slavery, the expansion of suffrage, the construction of the welfare state, the reduction in major-power war since 1945) all share this profile: incremental, multi-generational, anchored to evolving institutions rather than charismatic individuals.
This is unsatisfying. It is also, as far as I can tell, true. And accepting that the mechanism is unsatisfying is, paradoxically, the only honest way to engage with it. The alternative — believing in dramatic personal transformations or revolutionary ruptures or charismatic leaders who will solve the structural problems — has been tested repeatedly across history and produced, with remarkable consistency, new versions of the same anchored systems wearing different costumes.
I do not know what the right level of cynicism is. I know that complete cynicism is functionally identical to support for the existing anchor, because it removes any energy from the slow institutional work that might shift it. I know that complete idealism is anchoring on a non-existent reference point, which produces frustration and burnout but rarely change. I am trying to live somewhere in the middle: clear-eyed about how systems actually work, while continuing to act as if change is possible, because the act of continuing — imperfectly, without certainty, without the satisfaction of being right — is itself the only thing that has historically produced the changes I would want to see.
The world is mostly anchored. Most people, including me, are mostly anchored. The work is moving the anchor, slightly, in directions that future generations will judge — and they will judge using anchors we helped install. Which is either humbling or terrifying depending on the day.
I’m going for humbling. The terrifying version is true regardless. There’s no point in adopting the costlier interpretation when both are available.
Written: May 2026 Version: 1.0 This is how I understand this concept today. It will change.
Leave a Reply