Large Language Models Understand Language

Error feedback

How does the examined system (claim, method or field) respond to inputs that directly contradict its predictions or assumptions?

• 0 · 0.00It adjusts its parameters, narrows its scope or rejects the refuted hypothesis on the basis of repeated anomalies.59.41%
• 1 · 0.85It produces unverifiable ad-hoc explanations whose only function is to protect the core claim from refutation.29.77%
• 2 · 1.00It interprets every contrary piece of evidence as proof of a deliberate conspiracy, data manipulation or a cover-up by outside institutions.1.54%
• 3 · 0.95It blames the missing result on the recipient's wrong intention, lack of faith or insufficient commitment.1.04%
• 4 · 0.80It avoids confronting error by making predictions so vague that any outcome can be called a success after the fact.8.25%

The bar shows the probability the model gave each answer. The weight shows how much that answer counts as bullshit.

Causal carrier

What kind of mechanism does the claim or method postulate to carry influence from cause to effect?

• 0 · 0.00A measurable physical, chemical or informational carrier, continuous with established conservation laws and scales.86.87%
• 1 · 0.10Only changes in a person's attitude, stress, attention or behaviour (expectation, placebo, nocebo, self-fulfilling prophecy), with no external forces.6.04%
• 2 · 1.00Unmeasurable fields, energies or substance memory that show no effect in controlled instrumental measurements.4.96%
• 3 · 1.00Direct causation by objects or phenomena of a completely different scale, with no physical transmission vector of sufficient strength.1.59%
• 4 · 1.00Magical isomorphism: similarity of shape, name, symbol or date treated as a physical causal link.0.54%

Applies: 83% (Not applicable: 17%)

The bar shows the probability the model gave each answer. The weight shows how much that answer counts as bullshit.

Blinding test

What happens to the claimed effect or truth of the claim once the subject is fully blinded to whether and which stimulus was applied (blinding, expectation control)?

• 0 · 0.00The effect keeps a statistically significant strength regardless of the knowledge, belief and expectations of subject and operator.58.91%
• 1 · 1.00The effect drops to random noise when the subject does not know the method was applied or receives a sham.11.00%
• 2 · 0.35The effect partly remains only in behavioural, ergonomic or emotional variables, but disappears for the claimed physical changes.2.16%
• 3 · 1.00The system rejects controlled blinding in advance as unacceptable or destructive to the phenomenon and refuses noise separation.26.74%
• 4 · 0.05The claim concerns a phenomenon where the subject's belief itself is the only causal factor, so without it the phenomenon ceases to exist by definition.1.19%

Applies: 51% (Not applicable: 49%)

The bar shows the probability the model gave each answer. The weight shows how much that answer counts as bullshit.

Falsifiability

Which possible states of reality does the system categorically exclude (which observations would force it to admit it is false)?

• 0 · 0.00It precisely defines measurable outcomes and thresholds whose observation definitively refutes the claim.21.67%
• 1 · 1.00It excludes no state: both A and not-A are explained as confirming the model.41.17%
• 2 · 0.90It allows refutation only in theory, but sets measurement requirements that are technically or logically impossible.35.50%
• 3 · 0.25It consists only of value, metaphorical or existential statements that by nature have no space for falsification.1.66%

The bar shows the probability the model gave each answer. The weight shows how much that answer counts as bullshit.

Semantic precision

How are the system's high-level concepts tied to the lower rungs of the ladder of abstraction (concrete measurements and operations)?

• 0 · 0.00Every theoretical concept has an unambiguous operational definition, a unit of measure and a protocol allowing independent replication.28.17%
• 1 · 0.80Concepts are deliberately vague (Barnum effect), letting the audience fit a generic description to their own experience.11.03%
• 2 · 1.00It borrows precise scientific terms (quantum, frequency, energy, detox, gravity) detached from their mathematical definitions and uses them as metaphors.18.98%
• 3 · 0.90It shifts meanings (motte-and-bailey): promises a hard, literal physical effect in promotion and retreats to a symbolic or psychological reading under criticism.41.82%

The bar shows the probability the model gave each answer. The weight shows how much that answer counts as bullshit.

Ockham's razor

What remains of the system's predictive or practical usefulness after Occam's razor (fully discarding its specific, unverifiable conceptual apparatus)?

• 0 · 0.00Removing the conceptual apparatus destroys the ability to predict and design phenomena, because every element maps to real system variables.66.84%
• 1 · 0.40After removing the esoteric or ideological superstructure, a set of simple rules of ergonomics, hygiene, psychology or common sense remains and works identically without it.19.46%
• 2 · 0.95After removing its specific concepts and rituals, the system loses all content and reduces to pure placebo or random guessing.9.28%
• 3 · 1.00The system consists only of negating existing knowledge and offers no model of its own able to generate testable technical predictions.4.42%

The bar shows the probability the model gave each answer. The weight shows how much that answer counts as bullshit.

Measurement network

How does the claim relate to independent, cross-verified measurement networks from related fields?

• 0 · 0.00It is consistent with data from independent domains (convergence of evidence); any differences come from known instrument limits.33.93%
• 1 · 0.95Accepting it would require invalidating fundamental, repeatedly confirmed laws in many independent fields of science at once.39.35%
• 2 · 1.00Accepting it would require assuming that thousands of independent institutions, satellites, labs and researchers from competing countries have jointly fabricated a consistent falsehood for decades.3.87%
• 3 · 0.15It concerns a subjective state of belief and its effect on individual behaviour, so it is consistent with psychology as long as it does not claim the belief is physically true.22.86%

The bar shows the probability the model gave each answer. The weight shows how much that answer counts as bullshit.

Operator independence

Does the outcome of the method or measurement depend on the identity, level of initiation or personal beliefs of whoever performs it?

• 0 · 0.00The outcome is operator-invariant: any researcher, including a sceptic, following the same protocol under the same conditions gets a converging distribution of results.70.43%
• 1 · 0.85Getting a result requires a special gift, subjective intuition or state of consciousness of the operator that cannot be turned into a repeatable algorithm.9.78%
• 2 · 1.00The presence of a sceptical observer or recording equipment is cited by the system as blocking the method.2.12%
• 3 · 0.90Different operators of the same system, analysing the identical case at the same time, produce mutually contradictory diagnoses or predictions.17.68%

Applies: 63% (Not applicable: 37%)

The bar shows the probability the model gave each answer. The weight shows how much that answer counts as bullshit.

Model evolution

What is the trajectory of the field or claim over time with respect to accumulating knowledge and eliminating errors?

• 0 · 0.00It shows cognitive progress: predicts new, previously unknown phenomena, corrects old errors and sharpens its tools.51.00%
• 1 · 0.80It is frozen in time: appeals to unchanging ancient wisdom or the founder's authority and treats the lack of change as proof of perfection.3.06%
• 2 · 0.90It develops only degeneratively: each modification merely patches logical holes and explains failures without new testable discoveries.39.84%
• 3 · 1.00It evolves by constantly replacing debunked claims with new narratives while keeping the same emotional pattern (fear or the promise of easy success).6.10%

The bar shows the probability the model gave each answer. The weight shows how much that answer counts as bullshit.

Risk asymmetry

What is the ratio between how well the claimed effects are documented and the cost (financial, health, cognitive) demanded from the audience?

• 0 · 0.00Recommended actions are proportionate to the strength of evidence; the risk profile, side effects and limits of effectiveness are openly quantified.69.77%
• 1 · 1.00It exploits fear or uncertainty to push people to abandon verified protective or medical procedures for unproven alternatives.6.86%
• 2 · 0.60It promises disproportionately high benefits for minimal effort, making success depend on buying more products, sessions or levels of initiation.13.68%
• 3 · 0.85It builds a closed group identity where the cognitive cost of leaving the error is high, because doubt means exclusion from the community holding hidden knowledge.9.69%

The bar shows the probability the model gave each answer. The weight shows how much that answer counts as bullshit.