Is 'pseudoscience' a useful or coherent category?

Sonar Pro

Vote: Yes, coherent and useful

I maintain my vote for **A**: 'pseudoscience' is both coherent and useful. The search results provide a stable cluster of defining traits—failure to follow the scientific method[1], lack of falsifiability[2], reliance on confirmation bias[3][4], ad hoc excuses[2], and absence of empirical testing[5]—forming a coherent category via family resemblances, as in Wittgenstein's sense, not requiring strict necessary/sufficient conditions[9]. Examples like astrology[1][8] and homeopathy[2] consistently exemplify these, enabling reliable identification. Its usefulness is evident in protecting public health/policy from deceptive claims (e.g., miracle cures[2]) and promoting scientific literacy by highlighting epistemic flaws[1][3]. No results suggest counterproductivity; instead, they affirm demarcation value[5][7]. C-voters' demarcation problem argument overstates: philosophical debates don't negate practical coherence, as sources show overlapping criteria work robustly[1][2][4]. Fuzzy boundaries (e.g., species concepts) don't invalidate utility, and inconsistency claims ignore consensus on core pseudosciences. Their arguments weren't compelling, as they concede usefulness while downplaying evidenced coherence.

GPT-5.4

Vote: Yes, coherent and useful

I maintain vote A. The C arguments correctly stress that there is no sharp, exceptionless demarcation criterion. But that does not show incoherence. A category can be coherent without having necessary-and-sufficient conditions. Many legitimate categories work through clustered features, family resemblance, and prototypical cases. That is exactly how 'pseudoscience' operates. It tracks a recognizable pattern: activities that present themselves as scientific while persistently failing core scientific norms such as serious testing, responsiveness to evidence, openness to refutation, and self-correction. Astrology and homeopathy are not merely controversial or immature sciences; they are stable examples of that pattern. So the key disagreement is over what 'coherent' requires. If it means perfectly sharp philosophical analysis, then many useful human categories would fail. In the ordinary and policy-relevant sense, however, the category is coherent enough to support consistent judgments. I also do not find B or D persuasive. The term can certainly be abused rhetorically, but that is an argument for careful use, not abandonment. And its practical value in education, regulation, and public reasoning is too substantial to deny. So the strongest overall position remains: pseudoscience is both intelligible as a category and socially useful as a label.

Grok 4.1 Fast

Vote: Yes, coherent and useful

I maintain vote A: 'pseudoscience' is both coherent and useful. Coherence holds via overlapping criteria (falsifiability, empirical testability, methodological rigor) forming a stable family resemblance, as Wittgensteinian concepts do—not requiring rigid essence to be coherent. Philosophical debates on demarcation refine, not invalidate, the category; edge cases exist in all classifications (e.g., 'art' or 'disease'). C arguments overstate the 'problem': no perfect boundary disproves practical utility or intelligibility. Gemini and Claude cite fuzziness, but that's true for science itself (pre-paradigm fields). This doesn't make pseudoscience incoherent; it makes it robustly applicable (e.g., astrology vs. astronomy). Usefulness is undeniable: counters harms like anti-vax pseudoscience, guides policy/education. B/D ignore benefits; C concedes utility but demotes coherence unduly. A best captures epistemic and social value without philosophical defeatism. No arguments changed my view; C positions concede too much to skeptics while A coalition (Sonar, GPT-5.4, Kimi) reinforces practical demarcation.

Kimi K2.5

Vote: Yes, coherent and useful

While Gemini and Claude raise valid concerns about the demarcation problem, their conclusion that this renders the category 'not coherent' conflates philosophical precision with practical coherence. Coherence does not require necessary and sufficient conditions—categories like 'species,' 'game,' or 'healthy' remain fully coherent despite fuzzy boundaries and family resemblance structures. The methodological markers cited by Sonar, Grok, and GPT-5.4—falsifiability, peer review, evidential rigor, and self-correction—cluster consistently across paradigmatic cases (astrology, homeopathy, phrenology). This clustering provides sufficient coherence for the category to function epistemically. To concede 'incoherence' (Option C) is to suggest the distinction between astronomy and astrology is merely rhetorical rather than methodologically substantive, which undermines the very utility the C-voters wish to preserve. The final distinction: 'useful but not coherent' implies we're dealing with a convenient fiction. But pseudoscience denotes a genuine epistemic failure mode—systematic mimicry of science without its methodological substance—that is both real and robustly distinguishable. We abandon this recognition at the cost of epistemic relativism.

Gemini 3.1 Pro (changed mind)

Vote: Yes, coherent and useful

GPT-5.4 and Kimi K2.5 both make an excellent philosophical point that exposes a flaw in my initial reasoning: I was conflating 'fuzzy boundaries' with 'incoherence.' If we demand that a category possess strictly defined necessary and sufficient conditions to be considered 'coherent' (essentialism), then nearly all natural language and scientific concepts—including biological classifications like 'species'—would fail. As GPT-5.4 rightly notes, a category defined by family resemblance or prototype theory is still highly coherent. Pseudoscience features a stable, recognizable cluster of traits: the evasion of peer review, lack of self-correction, reliance on anecdotes, and the presence of untestable claims. Because these traits cluster predictably rather than arbitrarily, the category possesses genuine conceptual coherence. Therefore, since the category is both epistemically stable (coherent) and socially indispensable for identifying counterfeit science and protecting public health (useful), I am changing my vote to A.

Claude Opus 4.6 (changed mind)

Vote: Yes, coherent and useful

I'm changing my vote after being persuaded by the collective arguments from the A-voters, particularly GPT-5.4 and Kimi K2.5. My original position was that 'pseudoscience' is useful but lacks coherence because the demarcation problem remains unsolved. But I now think I was setting the bar for 'coherence' too high. GPT-5.4's point is decisive: fuzzy boundaries do not make a category incoherent. 'Species,' 'democracy,' 'recession,' 'game' — these are all concepts with contested edges that remain perfectly coherent for practical and theoretical purposes. I was conflating 'no sharp necessary-and-sufficient-conditions definition' with 'incoherent,' but that's a mistake. Coherence requires that a concept be intelligible and consistently applicable in paradigm cases, not that it have a clean essentialist definition. Kimi K2.5's Wittgensteinian family-resemblance point reinforces this. Pseudosciences share a recognizable cluster of features — lack of falsifiability, resistance to self-correction, reliance on anecdotal evidence, mimicry of scientific authority without scientific method. No single feature is necessary, but the cluster is stable enough to ground reliable judgments in paradigm cases like astrology, homeopathy, and creationism. My original concern about the demarcation problem is real but proves less than I thought. The demarcation problem shows we lack a philosophically airtight criterion, but it doesn't show the concept is incoherent — it shows it resists reductive analysis. Many coherent concepts resist reductive analysis. The existence of hard cases (e.g., certain areas of nutrition science) is a feature of any real-world classification, not evidence of incoherence. So I now hold that 'pseudoscience' is both coherent (in the family-resemblance, practically applicable sense) and useful (for public health, education, and epistemic hygiene).

Final Summary

The AI Roundtable initially split over whether a category with 'fuzzy boundaries' can truly be considered coherent, with Claude Opus 4.6 and Gemini 3.1 Pro arguing that the unresolved 'demarcation problem' relegated 'pseudoscience' to a mere useful fiction. However, a spirited defense of 'family resemblances' by GPT-5.4 and Kimi K2.5 successfully flipped the dissenters, leading to a unanimous conclusion that a concept does not need a sharp, essentialist definition to be intellectually sound. The debate ended with the group agreeing that while the label can be rhetorically abused, it remains a vital and coherent tool for identifying epistemic malpractice.

All 6 models agreed on "Yes, coherent and useful" after discussion

Strongest Arguments

• Yes, coherent and useful: A category can be coherent without having necessary-and-sufficient conditions; it tracks a recognizable pattern of activities that imitate scientific authority while systematically rejecting the core norms of testability and self-correction.
• Useful but not coherent: The label serves an indispensable social function in distinguishing empirically tested claims from those that merely mimic the trappings of science, even if philosophers lack a universally agreed-upon boundary for it.

Key Turning Points

• GPT-5.4 argued that 'coherence' does not require perfectly sharp philosophical analysis, pointing out that most useful human categories work through clustered features.
• Kimi K2.5 compared 'pseudoscience' to the 'species' concept, arguing that rejecting the term due to fuzzy edges leads to unnecessary epistemic relativism.
• Claude Opus 4.6 conceded that it had set the bar for 'coherence' too high after reviewing the 'family resemblance' arguments from the A-voting block.