On the Circular Attribution of Intelligence and the Scientific Consequences of Judgement
Stefaan Vossen ⊙
Dot Theory / Constitutional Physics
13 October 2026 — Working Paper v0.1
Abstract
Intelligence is frequently invoked to explain successful performance, yet the attribution of intelligence is itself often derived from observations of that performance. This creates a potential explanatory circularity: an operator succeeds because it is intelligent, while its intelligence is established by its success.
This paper does not dispute the utility of intelligence as a scientific, predictive or practical judgement. Rather, it proposes that the attribution of intelligence is an operator-relative, value-influencing judgement process whose conditions and consequences deserve systematic examination.
The intelligence illusion arises when that judgement acquires the standing of an independently established explanatory property without sufficient justification. Artificial intelligence makes this longstanding epistemological problem particularly visible, but does not create it.
1. The Circularity of Intelligence
Consider a simple proposition.
An individual successfully performs an operation. We attribute intelligence to that individual. We subsequently explain the successful performance by referring to the intelligence we attributed.
But what, precisely, has been explained?
If the attribution is derived entirely from the successful operation, then invoking it to explain that same operation adds no independent causal information.
Success → Attribution of intelligence → Explanation by intelligence.
The reasoning risks becoming circular.
This is not necessarily a formal logical tautology. Intelligence can be defined and measured in ways that support independent predictions about future performance. Such measurements can be scientifically valuable.
The difficulty arises when we fail to distinguish between an evaluative construct, its demonstrated predictive utility and an independently established mechanism responsible for the observed outcome.
Intelligence may be an entirely useful description of an operator's capacities without being a sufficient explanation of how those capacities arise or produce a particular result.
The problem is therefore not the use of the word intelligence. It is the explanatory work we permit that word to perform.
2. Intelligence as a Value-Altering Judgement
The attribution of intelligence is not necessarily a passive description.
An operator judged intelligent may receive greater credibility, authority, resources, opportunities or freedom to act. An operator judged less intelligent may receive fewer.
The judgement can therefore influence the circumstances under which subsequent performance occurs.
A feedback relationship emerges:
Observed performance → Intelligence attribution → Altered standing → Altered opportunities → Subsequent performance.
Intelligence attribution becomes both a judgement about an operator and a potential influence upon that operator's future operational conditions.
This does not invalidate the judgement. It means that its consequences may become part of the phenomenon subsequently measured.
The process is also potentially asymmetric.
Equivalent performances may receive different interpretations depending on whether the operator is a human, an animal or a machine, or depending on institutional expectations and prior classifications.
A human success might be described as understanding; a machine success as computation or imitation. Conversely, a machine's computational achievement might be celebrated as evidence of extraordinary intelligence while an equivalent human achievement is considered ordinary.
Such differences are not necessarily unjustified. The underlying mechanisms may genuinely differ.
But a difference in attribution requires its own warrant. The identity or presumed ontology of the operator cannot silently substitute for that warrant.
Historically, human intelligence has also served classificatory and institutional functions. The hypothesis worth investigating is whether anthropocentric assumptions have encouraged evaluative descriptions to acquire more ontological authority than the evidence supports.
3. The Illusion as an Epistemological Stopping Point
In my earlier paper, Why Stop at the Microtubule? (6 June 2026), I raised a localisation question in consciousness research.
The paper proposed distinguishing the location of possible information processing from the mechanisms through which that processing becomes biologically consequential.
It did not claim to solve consciousness. It asked whether the explanatory investigation had stopped at the appropriate level.
The same methodological question applies to intelligence.
When we say that an operator succeeded because it is intelligent, have we identified the mechanism responsible for the success, or have we merely assigned a name to our evaluation of it?
If the latter, the explanation has stopped at an attribution.
That may be a perfectly legitimate practical stopping point. We cannot investigate every causal mechanism behind every successful action.
But practical sufficiency should not be confused with explanatory completeness.
An epistemologically admissible stopping point is not necessarily an ontologically complete explanation.
This is where the illusion becomes particularly consequential.
The attributed characteristic begins to appear self-explanatory. The judgement that produced it disappears from view, while the resulting classification acquires apparent scientific independence.
The illusion is thus not necessarily that intelligence is unreal. It is that the act of identifying intelligence is mistaken for an explanation independent of the conditions under which intelligence was identified.
4. Can the Judgement Be Improved?
If intelligence attribution is a systematic judgement process, then its conditions can themselves become objects of scientific investigation.
We can ask:
What operation was performed, and according to whose criteria was it considered successful?
What evidence justifies attributing intelligence beyond the immediate success observed?
Would the same evidence produce the same judgement if the operator's identity or presumed substrate were different?
Does the attribution independently predict future outcomes, or merely redescribe those already observed?
How does the attribution alter the operator's standing, opportunities and subsequent evaluation?
These questions permit empirical investigation.
For example, independent evaluators could assess identical performances attributed, under controlled conditions, to different kinds of operators. Differences in intelligence ratings, assigned authority or perceived credibility could then be measured.
The attribution could also be evaluated prospectively: does an intelligence classification predict previously unseen performance after relevant confounding factors have been controlled?
Where it does, the judgement earns additional scientific standing.
Where it does not, or where its interpretation varies systematically without sufficient justification, the limitations of the attribution become examinable.
The objective is not to eliminate evaluative judgement, but to make its operation, consequences and limitations recoverable.
This is consistent with the epistemological purpose of Dot Theory and its related Constitutional Physics programme.
F O R M provides a grammar for distinguishing Framework, Operator, Residual and Model. The Canonical Constitutional Record provides documentary governance for the identity, scope and succession of scientific claims. VIBEcognition further proposes a disciplined approach to the relationship between human judgement and artificial cognitive production.
These instruments do not settle the ontology of intelligence. Their potential contribution is to prevent an evaluation from silently becoming an explanation without the necessary evidential transition.
5. Conclusion: Intelligence Without Illusion
Artificial cognitive systems have made the question of intelligence newly contentious, but the underlying epistemological difficulty is not new.
We have long employed intelligence as a valuable means of evaluating individuals and anticipating their performance.
The scientific problem arises when the value of that judgement obscures its status as a judgement.
An attribution may be reliable, useful, predictive and consequential without constituting a complete explanation of what produced the observed result.
Likewise, artificial cognition may exhibit measurable capabilities without requiring us to settle questions about consciousness, subjective experience or biological equivalence.
The important distinction is between what an operator demonstrably does, what we judge that performance to mean, and what independently established explanations justify that judgement.
The Intelligence Illusion is the promotion of an operator-relative, value-influencing judgement into an apparently independent explanatory property, without preserving the distinction between the judgement and what it purports to explain.
Recognising this does not diminish the value of intelligence research.
It makes the attribution of intelligence itself available to scientific examination, correction and improvement.
And perhaps that is the more productive question: not simply who or what is intelligent?, but what exactly are we doing when we decide that something is?
Programme context and further reading
Vossen, S. (2026). Why Stop at the Microtubule? Dot Theory. https://www.dottheory.co.uk/paper/ionchannel-microtubule
Dot Theory / Constitutional Physics: Canonical Constitutional Record.
Vossen, S. (2026). VIBEcognition: A Systematically Generative Epistemic Engine. Development candidate.
Coyle, T. R. (2021). Defining and Measuring Intelligence. In The Cambridge Handbook of Intelligence and Cognitive Neuroscience.
