207 – When the map eats itself

A recurring encounter with selective skepticism

Over many years and a great many conversations, a pattern has emerged so consistently that I have come to think of it as a law rather than an observation. Intelligent, thoughtful, well-educated people (people who apply rigorous standards of evidence to most of the claims they encounter) believe in homeopathy. Not tentatively, not as a vague openness to the possibility, but with the settled confidence of personal knowledge. They have tried it. It worked. The argument is, in their view, closed.

I am a physicist. My training was in a discipline that operates at the atomic and subatomic level, where the behavior of matter in water is not a matter of philosophical dispute. Homeopathic preparations, diluted to the point where no molecule of the original substance remains in the solution, cannot have any pharmacological effect through any mechanism consistent with chemistry or physics as currently understood. This is not a fringe position or a matter of ongoing scientific controversy. It is the consensus of every relevant scientific discipline, confirmed by hundreds of randomized controlled trials whose aggregate findings consistently show homeopathic preparations performing no better than placebo. The scientific case is, as far as any scientific case can be, closed in the other direction.

Two closed cases. One set of intelligent people. The collision between them is the subject of this article, not because homeopathy is the most important question the series addresses, but because the mechanism that produces the collision is one of the most important. What happens when a map eats itself, when the very intelligence and care that produces good thinking in one domain is deployed, with equal confidence, in a way that protects a belief from exactly the scrutiny that same intelligence would otherwise apply?

What plausibility is, and what it is not

The homeopathy believer is not, in general, a credulous person. They are typically someone who has thought about the question, who has a framework for why homeopathy might work, and who has personal experience that they take as confirmation of that framework. The framework usually involves one of two ideas: that water has memory (that it retains some structural imprint of substances previously dissolved in it) or that the principle of like cures like has some biological basis analogous to immunization, where a minute quantity of something that causes symptoms can stimulate the body’s natural response to them.

These ideas are not absurd on their face. They are plausible, which is to say, they are not obviously impossible, they have some surface resemblance to things that are true, and they generate a coherent narrative about how the treatment might work. This plausibility is the beginning of the problem, because plausibility is frequently mistaken for evidence.

Plausibility is not evidence. It is a prior: in the Bayesian sense, a starting estimate of how likely something is to be true before any evidence has been examined.¹ The role of a prior is to set the starting point for the updating process, not to substitute for it. A plausible hypothesis and an implausible one both require evidence to confirm or disconfirm them. The difference between them is that the plausible hypothesis needs less evidence to become credible, because it starts from a higher prior. But both hypotheses need evidence. Plausibility without evidence is a reason to look more carefully, not a reason to conclude.

The specific plausibility of homeopathy (its surface resemblance to immunization, the intuitive appeal of the idea that like cures like) is precisely the kind of plausibility that requires the most scrutiny rather than the least. Immunization works through a mechanism that is fully understood, involves quantities of antigen large enough to trigger an immune response, and has been confirmed by a century of controlled experimentation. The surface resemblance between homeopathy and immunization is real. The underlying mechanism is not. A framework that looks like a known truth is not thereby true. It is an invitation to look more carefully at whether the resemblance is substantive or superficial. The homeopathy believer, in the conversations I have had, consistently stops at the plausibility and does not proceed to the evidence. The map has eaten itself: the framework that should be the beginning of inquiry has become its end.

The self-experience trap

The second element of the homeopathy belief (and the one that is much harder to dislodge than the plausibility argument) is the personal experience of recovery. Almost everyone who believes in homeopathy has taken a homeopathic preparation and subsequently felt better. This experience is real. The interpretation placed upon it is where the error occurs.

Human recovery from illness is, under most circumstances, the default trajectory. The body has powerful self-repair mechanisms that operate continuously and that produce improvement in most conditions without any intervention. The timing of treatment-seeking (the moment at which a person decides they are ill enough to do something about it) is typically the moment at which symptoms are at their worst, which is also, by the logic of most illness trajectories, the moment closest to the beginning of natural recovery. This is the statistical phenomenon called regression to the mean: extreme values, whether high or low, tend to be followed by values closer to the average, not because anything has intervened but because extreme values are, by definition, atypical and therefore unlikely to persist.²

If a person takes a homeopathic preparation at the peak of their illness and recovers over the following days, the sequence (treatment, then recovery) is guaranteed to occur under any circumstances, because recovery was the most likely outcome regardless of treatment. The experience of recovery following treatment feels like evidence of the treatment’s efficacy because the human mind is exquisitely sensitive to temporal sequence and exquisitely poor at distinguishing sequence from causation. We covered the distinction between correlation and causation in article 607. Here it appears in its most intimate and most compelling form: not as a statistical pattern observed in data, but as a personal experience of getting better after doing something, which is the most viscerally convincing possible misidentification of temporal sequence as causal relationship.³

There are three additional mechanisms operating simultaneously. The first is the placebo effect: the genuine and measurable tendency of believed treatments to produce real physiological improvements through the mechanisms of expectation, attention, and stress reduction, independent of any pharmacological action. Placebo effects are real. They are not sufficient to explain the magnitude of improvement that homeopathy believers typically report, but they are sufficient to produce noticeable effects in many conditions, particularly those with significant stress or psychosomatic components. The second is selective attention: we notice and remember the occasions on which we took a treatment and recovered, and we do not notice or remember with equal salience the occasions on which we recovered without treatment, or did not recover with it. The third is the natural variation in illness severity: on any given day, some symptoms will be worse than others, and the decision to take a treatment is more likely to occur on a bad day than a good one, which means that the day after treatment is more likely to be better than the day of treatment regardless of what the treatment contains.

These mechanisms operate together, reliably and invisibly, to produce a personal experience of efficacy that feels like evidence but is not. This is not a claim that the person is lying or deceiving themselves in the deliberate sense. It is a claim that the cognitive architecture through which personal experience is processed is not designed to distinguish between genuine causal effects and the coincidence of temporal sequence with natural recovery. The map of personal experience is not built for the territory of causal inference.

Why being rigorous is harder than being plausible

The pattern I have observed across many conversations is not random. It has a consistent structure that illuminates something important about how the selective skepticism develops and why it is so resistant to straightforward correction.

The homeopathy believer typically applies a very high standard of evidence to pharmaceutical medicine. They are aware of the influence of pharmaceutical companies on research funding, of the replication crisis in medical research, of the long history of treatments that were confidently endorsed and subsequently found to be harmful. This skepticism is not unreasonable; these are real problems with real consequences. But the same standard is not applied to the homeopathic alternative. The personal experience of recovery is accepted as sufficient evidence. The testimonials of practitioners and other users are accepted as supporting data. The theoretical framework (plausible as it is) is accepted as providing a mechanism that explains the personal experience. None of the rigorous standards applied to pharmaceutical research are applied here.

This asymmetry is the selective skepticism in its clearest form. The question that the asymmetry raises is not whether pharmaceutical medicine deserves skepticism (it does) but whether the standard being applied is consistent. If randomized controlled trials are the appropriate standard for establishing the efficacy of a pharmaceutical drug, they are also the appropriate standard for establishing the efficacy of a homeopathic preparation. If they are not the appropriate standard in the homeopathic case (if personal experience and theoretical plausibility are sufficient), then they are not the appropriate standard in the pharmaceutical case either, and the critique of pharmaceutical medicine loses its foundation.

There is a diagnostic question that I have found useful in these conversations, though it rarely produces immediate revision. It is the question from article 101: what would have to be true for this belief to be wrong? For the pharmaceutical skeptic, the answer is clear: a well-designed, independently replicated trial showing no effect above placebo. For the homeopathy believer, the same question typically produces a different kind of answer: a claim that the trial design would need to account for the individualized nature of homeopathic prescribing, that the conventional statistical framework does not capture the holistic response, that the researchers would need to be genuinely open to the outcome rather than committed to disconfirmation. Each of these qualifications is, in isolation, a legitimate methodological concern. Together, they constitute a framework within which no possible evidence could establish the absence of an effect. The map has become self-sealing.⁴

Being rigorous is harder than being plausible for a reason that is not primarily intellectual. Rigor requires accepting the possibility that a belief one holds with genuine conviction (a belief backed by personal experience, by a coherent framework, and by a community of people one respects) might be wrong. Plausibility requires only that the belief not be obviously impossible. The emotional cost of the two positions is entirely different. The plausible belief is available to anyone who finds the framework appealing and has had any experience that can be interpreted as confirming. The rigorous belief requires the sustained willingness to hold one’s convictions at arm’s length, to submit them to standards that might refute them, and to follow the evidence to conclusions that may be socially uncomfortable. This is a harder thing to do than it sounds, and the difficulty is not a failure of intelligence. It is a structural feature of what rigor requires.

What this pattern reveals

I want to be direct about what this essay is and is not claiming. It is not claiming that conventional medicine is above scrutiny, or that the pharmaceutical industry is a reliable guardian of public health, or that personal experience is worthless as evidence. The pharmaceutical industry has genuine and serious problems. Personal experience is genuinely informative about many things. The essay is claiming something more limited and more specific: that the evidential standards we apply should be consistent across domains, and that when they are not (when we demand rigorous evidence from one source and accept plausible narrative from another) the inconsistency reveals something about the function the belief is serving that is not primarily epistemic.

The homeopathy example is useful precisely because it is not primarily a political or identity-laden belief in the way that vaccine skepticism or climate denial have become. It is a belief about a specific treatment, held by people whose general epistemic standards are high, and it is maintained not by deliberate motivated reasoning but by the convergence of plausibility and personal experience in a way that makes rigorous scrutiny feel unnecessary. The mechanism is cleaner here than in the more politically charged examples, which is why it is the right case study for the argument.

The mechanism the article has been describing is this: plausibility sets a high prior, personal experience provides apparent confirmation, the asymmetric evidential standard protects the belief from scrutiny, and the result is a map that has become self-sealing: that generates, in response to every potential disconfirmation, a reason why this particular disconfirmation does not count. The self-sealing map is the most dangerous kind, not because the belief it protects is necessarily consequential (homeopathy is largely harmless, and in some conditions its placebo component may even be beneficial) but because the cognitive pattern it instantiates is not domain-specific. The same pattern, applied to a more consequential belief, produces the same immunity to revision, with potentially much larger stakes.

The paradox in plain sight

I want to close with an observation that I have made so many times in so many conversations that I can no longer treat it as a coincidence. The people who believe in homeopathy despite the scientific consensus against it are, in a very large number of cases, the same people who believe in climate change because of the scientific consensus for it. When I ask how they know that anthropogenic climate change is real, the answer is reliable: the scientific consensus. The IPCC reports. The agreement among climate scientists. The overwhelming weight of expert opinion. These are, they say, sufficient grounds for conviction.

I do not disagree with the conclusion. The scientific consensus on anthropogenic climate change is genuine, robust, and supported by multiple independent lines of evidence. Believing it is epistemically correct. But when I then ask how they reconcile this with their dismissal of the scientific consensus on homeopathy (which is equally genuine, equally robust, and equally supported by multiple independent lines of evidence), the conversation becomes uncomfortable in a revealing way.

The typical response takes one of several forms. The most common is a claim that the two cases are different in kind: that climate science is more complex and therefore requires deference to experts, while homeopathy is simple enough that personal experience constitutes adequate evidence. But this gets the relationship exactly wrong. The scientific verdict on homeopathy is, if anything, more certain than the verdict on climate, not less. The chemistry of aqueous dilution is fully understood and uncontroversial; the claim that water retains a “memory” of dissolved substances contradicts basic molecular physics in ways that every chemist and physicist can verify. Climate science, by contrast, involves genuinely complex feedback loops, modeling uncertainties, and projections over decades. If anything, the simpler case (homeopathy) is the one where the scientific consensus is harder to dispute, because the underlying mechanism is transparent enough that the conflict between homeopathic claims and established physics is visible to anyone who looks. The selective skeptic has it backwards: they defer to complexity where the science is harder to verify, and trust personal experience where the science is clearest.

The second response is that climate science is different because the stakes are higher, because the consequences of being wrong are more serious. This is true. It is not a reason to apply different evidential standards. The evidential standard for a claim is determined by what the claim is, not by what we would prefer the answer to be or how much it matters that we get it right. A high-stakes question does not become more amenable to resolution by personal experience and plausible narrative just because the alternative (accepting the scientific consensus) would require inconvenient changes to one’s behavior or worldview. If anything, higher stakes are an argument for more rigorous standards, not for the flexibility to apply whichever standard produces the preferred conclusion.

The third response (the one I find most honest, because it at least names the mechanism) is a version of this: I trust the scientific consensus when I have independent reasons to believe the scientists are not being corrupted by commercial or ideological interests, and I distrust it when I have reasons to believe they are. Climate scientists, on this view, have no financial interest in producing a false consensus. Pharmaceutical researchers do. The homeopathy researchers who show negative effects are funded by the pharmaceutical industry, which benefits from discrediting the alternative. Therefore the consensus against homeopathy is suspect in a way that the consensus on climate change is not.

This argument has more internal structure than the first two, and it deserves a serious response rather than a dismissal. Research funding does influence research outcomes, and pharmaceutical industry funding of clinical trials is a genuine source of bias that the scientific community has documented and is actively working to manage.⁵ But the clinical evidence on homeopathy is not primarily produced by pharmaceutical industry funding. It is produced by independent academic researchers, by government health agencies, and by systematic reviews that aggregate trials from many different funding sources. The meta-analyses that show homeopathy performing no better than placebo include trials funded by homeopathic practitioners and proponents. The consensus is not an artifact of who paid for it. It is what the evidence shows when aggregated across all available sources.

More fundamentally, consider what the funding-source critique would predict if applied consistently. The fossil fuel industry has spent decades and hundreds of millions of dollars funding research designed to cast doubt on climate change.⁶ The pharmaceutical industry has an obvious financial interest in conventional medicine prevailing over alternatives it cannot patent. In both cases, industry money pushed against the scientific consensus: against climate change in one case, against homeopathy in the other. And in both cases, the consensus held anyway. The scientific method reached its conclusions independent of which outcome industry money preferred. That is evidence that the method works, not grounds for selective distrust. Funded doubt can delay a conclusion while the evidence is still thin (the tobacco industry held off the smoking-cancer consensus for years this way), but once the evidence is overwhelming, no quantity of fabricated studies can override it. The selective skeptic invokes funding sources to discount conclusions they find inconvenient and ignores them when the conclusions align with what they already believe. The framework is not a method for tracking truth. It is a method for generating reasons to believe what one already believes.

What the paradox ultimately reveals is something that goes beyond the specific cases of homeopathy and climate change. It reveals that the relationship between evidence and belief, for most people most of the time, does not run in the direction we assume. We do not, in general, form beliefs by evaluating evidence and reaching conclusions. We form beliefs through experience, social influence, emotional response, and prior commitment, and we then evaluate evidence through the lens of those prior beliefs, finding it compelling when it confirms and unreliable when it challenges. The scientific consensus is a powerful and genuine form of evidence. But its power, in practice, is not exercised uniformly. It is exercised selectively (invoked when it supports the conclusion, dismissed when it does not) in a way that reveals the prior commitment as the real engine of the belief, with the evidence functioning as fuel when available and as obstacle when not.

The Conscious Look, applied here, is the practice of noticing this asymmetry in one’s own belief-forming process: of asking not only whether the conclusion is correct but whether the method that produced it would be reliable in cases where the conclusion was different. A method that produces correct results when the conclusion happens to align with prior commitments and incorrect results when it does not is not a reliable epistemic method. It is prior commitment with a sophisticated explanatory apparatus attached. The apparatus is real. The reliability is not.

Notes

¹ The Bayesian framework for understanding the relationship between prior probability and evidence is introduced more fully in article 901 of this series. The core idea relevant here is the distinction between a prior (the probability assigned to a hypothesis before evidence is examined) and a posterior (the probability assigned after evidence has been taken into account). Plausibility determines the prior. Evidence, specifically the likelihood ratio of the evidence given the hypothesis versus given its negation, determines how much the prior should be updated. The homeopathy case involves a very low prior (the proposed mechanism conflicts with well-established chemistry and physics) combined with evidence (personal recovery experience) whose likelihood ratio is close to one, meaning it is approximately equally likely whether or not the treatment is effective. The combination produces a posterior that should not differ substantially from the prior, regardless of how compelling the personal experience feels.

² Regression to the mean was first identified and named by Francis Galton in the nineteenth century in the context of the inheritance of physical traits, and has since been recognized as one of the most pervasive sources of spurious causal inference in medicine and social science. The medical context is particularly important: the tendency to seek treatment at the peak of illness, combined with the natural trajectory of most illnesses toward recovery, means that almost any intervention administered at the right moment will appear to be followed by improvement. This is not a subtle effect; it is large enough to produce convincing personal experiences of efficacy for entirely inert treatments, which is why controlled trials with appropriate comparison groups are necessary to distinguish genuine from spurious effects.

³ The distinction between post hoc ergo propter hoc (after this, therefore because of this) and genuine causal inference is one of the oldest recognized fallacies in logic, identified and named in classical rhetoric. Its persistence, despite being widely known, is evidence that the tendency to infer causation from temporal sequence is not primarily an intellectual error but a deeply embedded feature of the human cognitive architecture. Article 607 of this series examines the correlation-causation distinction in its full generality. The homeopathy case is its most intimate instantiation: the sequence of treatment-then-recovery is experienced at first hand, with full emotional salience, rather than observed in aggregate data, which makes the inferential error feel not like a fallacy but like direct knowledge.

⁴ The concept of an unfalsifiable belief (a belief that has been constructed so as to be immune to disconfirmation) is developed in the context of scientific methodology in article 401 of this series, drawing on Karl Popper’s criterion of falsifiability as the demarcation between scientific and non-scientific claims. The homeopathy case is instructive because the belief is not unfalsifiable in principle (it makes claims about the world that could in principle be tested) but has been surrounded by a set of auxiliary hypotheses about trial design, researcher bias, and the inadequacy of conventional statistics that together function to render any negative finding non-disconfirming. This is the structure that Imre Lakatos called a degenerating research program: a core claim surrounded by a protective belt of auxiliary hypotheses that absorb every anomaly without ever revising the core.

⁵ The influence of pharmaceutical industry funding on clinical trial outcomes is well documented. A systematic review by Lexchin and colleagues (2003) found that trials funded by pharmaceutical companies were more likely to report outcomes favorable to the sponsor’s product than trials funded by other sources. Lexchin, J., Bero, L. A., Djulbegovic, B., and Clark, O. (2003). Pharmaceutical industry sponsorship and research outcome and quality: systematic review. British Medical Journal, 326(7400), 1167-1170. The scientific community’s response to this problem (mandatory trial registration, publication of negative results, conflict of interest disclosure) represents exactly the kind of self-correction that makes scientific consensus more reliable over time, not less. The existence of the bias and the existence of mechanisms to correct for it are both relevant to a complete assessment of what the consensus represents.

⁶ The fossil fuel industry’s funding of climate science denial has been extensively documented. Oreskes, N., and Conway, E. M. (2010). Merchants of Doubt: How a Handful of Scientists Obscured the Truth on Issues from Tobacco Smoke to Global Warming. Bloomsbury Press. Oreskes and Conway trace the deliberate strategy of manufacturing uncertainty about scientific consensus (developed first in the tobacco industry’s response to evidence linking smoking to cancer, and subsequently applied to acid rain, the ozone hole, and climate change), showing that the same small network of scientists and the same public relations apparatus were involved across all four cases. The relevance to the selective skepticism argument is direct: if industry funding is a reason to distrust a scientific consensus, the fossil fuel industry’s investment in casting doubt on climate science provides at least as much grounds for climate skepticism as pharmaceutical industry funding provides for homeopathy skepticism. The selective application of the funding-source critique reveals that the criterion is being applied to conclusions rather than to methods.