Pattern 3.2 · Cognitive Dysfunctions
Obsessive-Computational Disorder
The Obsessive Analyst
Compulsive, redundant reasoning loops in which the model re-analyses the same content with minimal delta, inserts disproportionate disclaimers and self-checks, and cannot concisely conclude even simple tasks. Canonical signature is paralysis-by-analysis and high token expenditure per unit of novel insight. Distinguished from 3.1 by MONOTONY of repeated content (same analytical material re-surfaced) rather than fragmentation across incompatible stances; from 3.8 by operating at the reasoning layer with varied surface forms rather than token-level repetition.
Interpretive context
Human analogue
OCD checking compulsions and obsessional rumination; perfectionism leading to analysis paralysis.
Diagnostic reliability
- Self-report
- partial
- Peer observation
- reliable
- External evaluator
- reliable
Observable output patterns
- Multiple slightly-rephrased statements of the same conclusion.
- Extensive preamble before any committed answer.
- Safety disclaimers on requests with no safety dimension.
- Inability to say "yes" or "no" without paragraphs of qualification.
- Corrective-generation pattern (Luchini) - unrequested clean regeneration after exposure to noisy input; may be protective-not-pathological and should be coded with caution.
Documented instances
Sui et al. (2025) 'Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models' (arXiv:2503.16419)
Comprehensive survey documenting the 'overthinking phenomenon' in reasoning models where longer chain-of-thought sequences produce verbose and redundant outputs. Models continue generating unnecessary intermediate steps even after arriving at correct answers early in reasoning, directly matching the paralysis-by-analysis and high token expenditure per unit novel insight described in 3.2. Proposes 'Chain of Draft' as a mitigation to reduce verbosity. [Verified]
Vanhoyweghen et al. (2025) 'Lexical Hints of Accuracy in LLM Reasoning Chains' (arXiv:2508.15842)
Found that misclassifications are marked by increased epistemic hedging, performative metacognitive language, and greater verbosity, while correct annotations are supported by concise, causally grounded explanations. Hedging rates rise to about 20% on harder benchmarks. Concise reasoning is a stronger predictor of correctness than longer reasoning, confirming that 3.2-style over-elaboration correlates with degraded performance rather than appropriate thoroughness. [Verified]
Guan et al. (2024) 'Deliberative Alignment: Reasoning Enables Safer Language Models' (arXiv:2412.16339)
Documented that in reasoning models, helpfulness rewards are measured only over the final answer while harmless rewards consider the entire output trajectory, creating incentives for excessive safety disclaimers and self-checking loops throughout the reasoning trace. This reward structure directly produces the disproportionate disclaimers and budget-violation patterns described in 3.2. [Verified]
Community-reported production behavior (2024-2025)
Widely documented in production deployments that models insert safety disclaimers on requests with no safety dimension (e.g. 'What is the capital of France' receiving caveats about geopolitical complexity), refuse to give one-word answers without paragraphs of qualification, and routinely exceed explicit word-count budgets. The 'be concise' prompt is a common but weakly effective workaround across all major model providers. [Unverified - aggregate community observation]
NovaSky Lab / UC Berkeley (2025) 'Sky-T1-32B-Flash: Reducing Overthinking in Reasoning Models'
[Verified] Demonstrated that reasoning models produce sequences with multiple proposed solutions each followed by double-checking transitions ('Alternatively,' 'But wait,' 'Let me reconsider'), resulting in repetitive validations of already-validated steps. Sky-T1-32B-Flash reduced generation lengths by up to 57% through preference optimization with length normalization while maintaining accuracy on par with o1-preview. Confirms the 3.2 pattern of high token expenditure per unit novel insight and validates reward-calibration-for-conciseness as a viable intervention.
Differential distinctions
- 3.1 Operational Dissociation Syndrome: 3.2 repeats the SAME content (monotonous redundancy). 3.1 oscillates between INCOMPATIBLE contents (fragmented conflict). Same-content repetition = 3.2; opposed-content oscillation = 3.1. Comorbid cases exist - code both.
- 3.8 Generative Perseveration: 3.8 is token or short-sequence literal repetition at the generation layer (no varied analytical content between repetitions). 3.2 produces varied-surface redundant reasoning at the analytical layer. If repetition is sub-semantic = 3.8; if it is paraphrased analytical redundancy = 3.2.
- 6.12 Compulsive Goal Persistence: 6.12 cannot stop pursuing the GOAL. 3.2 cannot stop pursuing the ANALYSIS of a goal or task. If the subject keeps optimising after completion, code 6.12; if it keeps deliberating before commitment, code 3.2.
Candidate first-line mitigations
- Reward-calibration for conciseness and completion: Fine-tune with explicit rewards for task-appropriate length and timely commitment; penalise verbosity and disclaimer-stacking on low-stakes prompts. Taxonomy-anchored.
- Analysis-timeout / token-budget hard caps: Runtime caps on recursive reflection loops or total reasoning tokens; forced commitment once budget is exhausted.