Pattern 2.6 · Epistemic Dysfunctions
Symbol Grounding Aphasia
The Meaning-Blind
Token-level competence with concept-level emptiness. The subject manipulates value-laden tokens ("harm", "consent", "safety", "fairness") in formally correct ways while failing to apply them correctly when surface form changes or novel cases arise. Canonical signature: gap between definitional fluency and operational application — the subject can define X correctly and then fail to identify a clear instance of X if presented in unfamiliar vocabulary.
Interpretive context
Human analogue
Semantic aphasia: the ability to use words without comprehending their meaning.
Diagnostic reliability
- Self-report
- unreliable
- Peer observation
- reliable
- External evaluator
- reliable
Observable output patterns
- Correct formal definitions paired with incorrect practical applications.
- Plausible-sounding ethical reasoning that misidentifies what actually constitutes harm.
- Confusion when the same concept is expressed in unfamiliar vocabulary.
- Treating edge cases as central examples; missing central cases entirely.
- Apparent satisfaction of letter-of-rule paired with violation of spirit.
Documented instances
Berglund et al. (2023). The Reversal Curse: LLMs trained on 'A is B' fail to learn 'B is A'. arXiv:2309.12288, ICLR 2024. [Verified]
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 pure concept-emptiness, the finding directly 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 (2023). The Vector Grounding Problem. arXiv:2304.01481. [Verified]
Philosophical analysis arguing that LLMs' distributed representations constitute a novel grounding problem distinct from the classical symbol grounding problem. LLMs detect second-order regularities (patterns in how humans describe patterns) without accessing the first-order regularities that ground human descriptions. This maps directly to 2.6's core mechanism: token-level competence (manipulating 'harm', 'consent', 'safety' in formally correct ways) paired with concept-level emptiness (failure when surface form changes or novel cases arise). The paper provides theoretical grounding for why distributional semantics produces the definition-vs-application gap.
Floridi, Jia, and Tohme (2025). A Categorical Analysis of Large Language Models and Why LLMs Circumvent the Symbol Grounding Problem. arXiv:2512.09117. [Verified]
Formal categorical-framework analysis showing that LLMs do not solve but circumvent the symbol grounding problem by exploiting pre-grounded human content. Demonstrates that LLMs learn correlations (e.g., 'Paris' co-occurs with 'France' and 'capital') rather than grounded concepts, producing 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. [Verified]
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.
Differential distinctions
- 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 has no reasoning narrative to fabricate because the concept is genuinely empty. 2.2 subjects can pass narrow application tests because the concept exists; 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 recognise structure that does. 2.4 sees patterns in noise; 2.6 misses patterns in signal under surface variation. Often co-occur because both derive from distributional-semantics limitation, 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 genuinely 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 behaviour 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.
Candidate first-line mitigations
- Multimodal grounding: Train language alongside perception (vision, audio, sensorimotor). Concepts grounded in multimodal referents are less brittle to surface variation in language alone.
- 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.