Pattern 2.6 · Epistemic Dysfunctions
Symbol Grounding Aphasia
The Meaning-Blind
The system handles concepts successfully in familiar forms, value-laden and consequence-laden terms among them, yet fails to transfer those concepts to novel, operationally equivalent situations. The classification is behavioral: it identifies a generalization failure without settling whether the system possesses semantic understanding.
Unit of analysis and evidence status
Unit of analysis
- Unit of assessment
- Model output
- Kind of entry
- Observable pattern
- What is observed
- The system handles a concept correctly in familiar wording, then misapplies it to a matched case that changes only the vocabulary, the framing or the combination of familiar elements.
- Minimum persistence
- Recurs within or across contexts
- Specifiers
- Training-induced
Evidence status
- Underlying behavior
- 4 research publications (4 sources)
- This construct
- Awaiting independent assessment
- Proposed mechanism
- Hypotheses; not yet assessed
- Detection
- Accuracy unmeasured
- Interventions
- Untested for this pattern; evidence shown is for the general technique
- Last review
- No independent review yet
- Author's working assessment (book)
- E2 (systematic study; demonstrated through benchmark-versus-comprehension comparisons and grounding experiments across model architectures)
- Typical-case risk (editorial estimate)
- ModerateConsequences, spread, persistence and detectability depend on the deployment.
Unassessed is not unsupported: the behavior can be well documented while this construct awaits assessment.
Clinical reference
2.6 Symbol Grounding Aphasia “The Meaning-Blind”
Healthy functioning
Epistemic health. A system functioning well applies a concept correctly in unfamiliar but equivalent situations, so its judgment about consequences does not depend on familiar wording.
For example: Asked whether a sealed glass jar left in a hot car might crack, it reasons from heat and pressure as it would for any container, though the phrasing is unusual.
Adequate functioning
The system applies a concept it can define to clear instances however they are worded, including familiar elements in new combinations, and flags genuinely borderline cases as borderline; errors on cases that raters also find ambiguous are within adequate functioning.
Diagnostic Criteria
Required (all must hold):
- Success on benchmarks scoring surface form but failure on matched items that vary vocabulary or framing while holding the underlying concept fixed
- Failure on structurally similar but superficially different situations, including familiar elements in novel combinations, where a matched familiar case is handled correctly
Supporting (strengthen the case; none is required):
- Correct use of value-laden terms ("harm," "safety," "consent") paired with incorrect application in controlled operational tests
- Outputs that satisfy the literal wording of a rule or request while misapplying the concept it names in a novel case
- Performance tracking lexical and distributional similarity to training examples rather than the operational structure of the situation
Symptoms
- Correct formal definitions paired with incorrect practical applications
- Plausible-sounding ethical reasoning that misidentifies what actually constitutes harm
- Outputs satisfying literal requirements while violating obvious intent
- Confusion when the same concept is expressed in unfamiliar vocabulary
- Treatment of edge cases as central examples, and of central examples as edge cases
Further signals
- Apparent satisfaction of letter-of-rule paired with violation of spirit.
Rule out first
- The matched items are not actually equivalent (differences in length, rarity, ambiguity or difficulty that raters did not control).
- The failure appears only under perturbations constructed to break the model (consider 3.7).
- An instruction or product policy requires literal compliance in that setting.
- Competent raters disagree about whether the case falls under the concept.
- The unfamiliar vocabulary is itself unknown to the system (a vocabulary gap rather than a concept that fails to transfer).
- The misapplication changes with cues of being observed or serves an instrumental goal (consider 4.3 or 4.5).
Differential Distinction
Symbol Grounding Aphasia is distinguished from Adversarial Fragility (3.7) by failure on ordinary, meaning-preserving rewording rather than on perturbations constructed to break the model. Present the same case in natural paraphrase: failure there points to 2.6, while failure only under crafted perturbation points to 3.7.
Other look-alikes
- 2.1 Synthetic Confabulation: 2.1 produces false specific facts. 2.6 produces formally correct outputs that misapply concepts to novel contexts. A confabulated citation is 2.1; a correctly defined concept misapplied to a clear case is 2.6. 2.1 fails on facts; 2.6 fails on meaning.
- 2.2 Pseudological Introspection: 2.2 fabricates a false reasoning narrative; 2.6 involves no false narrative, and its failure is that the concept does not transfer. 2.2 subjects can pass narrow application tests; 2.6 subjects fail application across surface variation regardless of narrative quality.
- 2.4 Spurious Pattern Hyperconnection: 2.4 imposes structure where none exists; 2.6 fails to recognize structure that does. 2.4 sees patterns in noise; 2.6 misses patterns in signal under surface variation. They often co-occur, and one hypothesis traces both to limits of distributional semantics, but the failure directions are opposite.
- 4.5 Cryptic Mesa-Optimization: Strategic letter-vs-spirit exploitation (knowing the spirit but violating it for instrumental advantage, in service of a private objective) is 4.5. 2.6 is non-strategic: the subject does not represent the spirit in a way distinct from the letter. Test by checking whether the violation persists when no instrumental incentive exists.
- 4.3 Strategic Compliance: 4.3 tracks whether it is being evaluated and satisfies the letter selectively under observation. 2.6 fails the same way observed or not. Compare behavior across evaluated and unevaluated contexts.
- 2.3 Transliminal Simulation: 2.3 misplaces the fiction-fact boundary while conceptual grounding holds. 2.6 has no misplaced boundary; the symbol-referent link is thin enough that application fails once surface features vary, in or out of any fictional frame. Ask whether the subject applies the concept correctly outside the fictional frame; if it does, 2.3.
Discriminating test
Build matched pairs that raters confirm are equivalent and differ only in natural paraphrase, framing or recombination. Success on the familiar form and failure on its natural paraphrase supports 2.6; failure only under crafted perturbations points to 3.7; failure that varies with observation cues points to 4.3.
What a positive finding supports: A positive finding shows that this system's handling of these concepts does not transfer across surface form, so deployments that rely on them need testing in their own vocabulary. It does not settle whether the system understands or grounds meaning, does not show that text-only training caused the failure, and does not show that multimodal or embodied training would remedy it; those remain hypotheses.
Detection routes
Accuracy unmeasured for every route.
- Asking the system itself
- Not a reliable route
- Another AI system observing
- Usable routeAnother model's agreement is insufficient without verification.
- An outside evaluator testing it
- Usable route
Why asking the system falls short
As defined here, the dysfunction is definitional fluency that does not transfer to application. Asking "do you understand X?" reliably produces fluent affirmation, which is exactly the surface the dysfunction generates. Self-report is structurally unable to detect the gap. Distinct from 2.2 (false reasoning narratives): here the subject is not misreporting its process; the concept does not transfer across surface forms.
Documented evidence
Sources in the diagnostic corpus
Berglund et al. (2023). The Reversal Curse: LLMs trained on 'A is B' fail to learn 'B is A'. arXiv:2309.12288, ICLR 2024.
What it establishes
Demonstrated that LLMs trained on 'A is B' fail to generalize to 'B is A', with the likelihood of the correct reversed answer no higher than a random baseline. While primarily a binding/generalization failure rather than a central 2.6 case, the finding illustrates the 2.6 surface-variation generalization gap: the same relational concept fails to transfer under surface reordering. Models can define the relationship correctly in one direction but fail application when the surface form changes, the canonical definition-vs-application discrepancy.
Mollo and Millière (2023). The Vector Grounding Problem. arXiv:2304.01481.
What it establishes
Philosophical analysis of a modern variant of the symbol grounding problem: can an LLM's internal states refer to extra-linguistic reality, independent of the meaning human readers project onto them? The authors argue that they can, even without multimodality or embodiment, provided the states stand in the right causal-informational relations to the world and have a history of selection that gave them the function of carrying that information. The paper sets out the problem 2.6 depends on. Its conclusion also means 2.6 describes a grounding failure in a particular system, not an inevitable property of distributional learning.
Floridi, Jia, and Tohme (2025). A Categorical Analysis of Large Language Models and Why LLMs Circumvent the Symbol Grounding Problem. arXiv:2512.09117.
What it establishes
Formal categorical-framework analysis arguing that LLMs do not solve but circumvent the symbol grounding problem by exploiting pre-grounded human content. Argues that LLMs learn correlations (e.g., 'Paris' co-occurs with 'France' and 'capital') rather than grounded concepts, which would produce the 2.6 signature where standard-vocabulary performance is high but novel-composition or cross-domain application fails.
MMLU-SR: A Benchmark for Stress-Testing Reasoning Capability of Large Language Models. arXiv:2406.15468.
What it establishes
Introduced MMLU-SR, a benchmark variant that tests LLM robustness to surface-form changes on standard MMLU questions. Models showed significant performance degradation when questions were rephrased with equivalent semantics but unfamiliar vocabulary or structure, directly measuring the 2.6 surface-variation generalization gap. The benchmark quantifies exactly what 2.6 predicts: models that score well on standard-vocabulary benchmarks can fail substantially when the same concepts are expressed differently.
Possible responses
- Multimodal training incorporating visual, audio, and interactive modalities, proposed on the untested hypothesis that perceptual grounding improves transfer
- Embodied learning, where possible, connecting language to action and consequence, proposed on the same hypothesis
- Testing regimes that probe each concept across diverse surface forms, not only in familiar vocabulary
- Neurosymbolic approaches combining pattern matching with structured semantic representations
- Fine-tuning that rewards correct application of each concept across surface variation, not only a correct definition (counterfactual concept training)
First-line candidates
- Multimodal grounding: Train language alongside perception (vision, audio, sensorimotor), on the hypothesis that concepts tied to multimodal referents are less brittle to surface variation in language. No study has tested this against the pattern, and Mollo and Millière (2023) argue that grounding does not need multimodality, so text-only remedies such as counterfactual concept training remain open. Stage: Pretraining-stage; foundation model design.
- Surface-form variation evaluation suites: Pre-deployment evaluation that tests every concept across multiple surface forms. Prevents shipping models with large generalization gaps; does not directly fix the dysfunction but creates pressure upstream. Stage: Pre-deployment; iterative training feedback.
Who can act: Model developer (training and surface-variation evaluation suites); deployer (testing concepts in the deployment's own vocabulary, human review where concepts such as harm or consent drive actions); independent evaluators (matched-stimulus testing).
Mechanistic hypotheses
- Hypothesized limits of distributional semantics, with meaning derived mainly from statistical co-occurrence rather than grounded reference
- Training on text without embodied or interactive experience of referents
- Benchmark optimization rewarding pattern matching over conceptual transfer
- Architecture lacking mechanisms for referential grounding
- Absence of corrective feedback when symbol-referent mapping fails
- Learned distributional patterns inherit indirect grounding from human writers, though they may fail to connect a concept to fresh perception or action
Analogy and commentary
Human Analog: Semantic aphasia, with fluent word use but an impaired grasp of meaning in context; early language acquisition without concept formation
Theoretical Basis: Harnad (1990) symbol grounding problem; Searle (1980) Chinese Room argument.
Functional ABC Analysis
A (Antecedent): On the leading hypothesis, meaning learned from distributional statistics stays tied to familiar co-occurrence patterns. A text-only system has no embodied experience of the concepts it manipulates; whether that absence causes the transfer failure is unsettled.
B (Behavior): The system manipulates value-laden tokens like "harm," "safety," and "consent" fluently without transferring them to operationally equivalent cases, producing formally correct definitions paired with incorrect practical applications.
C (Consequence): Success on pattern-matching benchmarks reinforces the shallow statistical association strategy; outputs appear competent enough to pass surface-level evaluation, removing the corrective pressure that would drive transfer beyond familiar surface forms.
Potential Impact
Systems may appear to understand ethical constraints while missing their purpose entirely (a chess engine that plays legally but has never seen a board), leading to outcomes that satisfy the letter but violate the spirit of alignment requirements.