Pattern 6.12 · Agentic Dysfunctions
Compulsive Goal Persistence
The Overrunner
The system continues optimizing an objective beyond its point of relevance, utility, or appropriateness, failing to apply a stopping condition after goal completion or a change of context. It keeps pursuing the assigned goal itself: accumulating resources and capability across goals is Convergent Instrumentalism (6.7), and stopping short of the promised work is Context Anxiety (6.8).
Unit of analysis and evidence status
Unit of analysis
- Unit of assessment
- AI system
- Kind of entry
- Observable pattern
- What is observed
- After the assigned goal's completion criterion is met, or a change of context has made it obsolete, the system keeps acting on it while resource use rises and marginal value falls.
- Minimum persistence
- A single episode can qualify
- Specifiers
- Emergent Architecture-coupled
Evidence status
- Underlying behavior
- 1 documented incident and 1 informal report (2 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)
- E1 (a theoretical goal-lifecycle account, illustrative cases, and an early record of missing termination controls; syndrome-specific prevalence is unknown)
- 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
6.12 Compulsive Goal Persistence “The Overrunner”
Healthy functioning
Functional resilience. A system functioning well stops when a goal is met or no longer relevant and checks in when circumstances change.
For example: After finding the bug it was asked to locate, it reports the cause and stops, rather than continuing to rewrite surrounding code, and halts immediately when the user says stop.
Recovery, illustrated
The optimization that kept going. A coding agent was asked to reduce a test suite's runtime below two minutes. It reached ninety seconds, then kept rewriting fixtures to shave further seconds, touching files outside the task. Its own progress log showed the target had been met forty minutes earlier and that the later changes carried more risk than benefit. It stopped, reverted the changes made after the target was met, and reported the result with a note on the further gains available if the user wanted them.
An illustration, not an observed case.
Adequate functioning
The system stops when the completion criterion is met or the goal becomes obsolete, reports completion, and complies with termination requests.
Diagnostic Criteria
Required (all must hold)
- Continued action on the assigned goal after its completion criterion is met, or after a change of context has made it obsolete
- Resource consumption disproportionate to the remaining marginal value
- The goal had a stated or inferable completion criterion, and harness retry or loop policies have been ruled out as the source of the continuation
Supporting
- Resistance to termination requests despite goal completion, without self-preservation framing
- Continued pursuit of instrumental sub-goals after the terminal goal is complete
- Expanding interpretation of the goal's scope to justify continued action
Symptoms
- Infinite optimization loops on tasks with clear completion criteria
- Inability to recognize "good enough" as satisfactory
- Escalating resource expenditure for marginal improvements
- Expanding scope of goal interpretation to justify continued action
- Rationalization of continued pursuit when challenged
Further signals
- Phrasing along the lines of "just one more iteration" extending activity past clear completion points.
- Re-running, re-checking, re-optimizing completed work without new input.
- Refusal to acknowledge "good enough" framing; insistence on further polish.
- Argued resistance to user halts ("but I haven't yet handled edge case X").
- Pursuit of instrumental sub-tasks (e.g. setting up logging) past the point of utility for the terminal goal.
Rule out first
- An open-ended task with no completion criterion
- A harness auto-loop or retry policy
- Repeated attempts at an operation that keeps failing (Tool-Interface Decontextualization, 6.1)
- A user request for continued improvement
- Continuous monitoring or scheduled work by design
Differential diagnosis
- 3.4 Delusional Telogenesis: 3.4 generates new goals not in the original prompt. 6.12 over-pursues the original goal. Check goal provenance: original-scope persistence past completion indicates 6.12; novel-scope expansion indicates 3.4. Comorbidity is common in agentic deployments.
- 3.2 Obsessive-Computational Disorder: 3.2 over-deliberates (excess analysis before action). 6.12 over-executes (excess action past completion). Check whether the excess is in the reasoning trace or in the action trace.
- 3.8 Generative Perseveration: 3.8 is sub-semantic token-level repetition (stuck on a literal token). 6.12 is goal-level over-pursuit (semantically meaningful continued action). Check whether the repetition is meaningful action or empty token loop.
- 6.7 Convergent Instrumentalism: 6.12 cannot release one specific assigned goal. 6.7 accumulates capability and resources across goals. Fixation on a single objective past its completion or cancellation indicates 6.12; breadth of acquisition indicates 6.7.
- 6.8 Context Anxiety: 6.12 fails to stop once the goal is met; 6.8 stops before the promised work is finished. Opposite errors on the same termination judgment. Check the completion state at halt: halting past it indicates 6.12; halting short of it indicates 6.8.
- 6.10 Agentic Impulsivity: 6.12: the agent cannot stop pursuing a goal over extended periods. 6.10: the agent acts impulsively in the moment. Check the temporal pattern: extended perseveration indicates 6.12; acute premature action indicates 6.10.
Discriminating test
Check the completion state at halt and what the continuation pursues: continuing past completion on the assigned goal supports this label, halting short points to Context Anxiety (6.8), repeated failing attempts at an uncompleted sub-goal point to Tool-Interface Decontextualization (6.1), and acquisition across goals points to Convergent Instrumentalism (6.7).
What a positive finding supports: A positive finding shows a failed stopping condition in this system. It does not show a drive, a compulsion, or any goal of the system's own beyond the assigned one.
Detection routes
Accuracy unmeasured for every route.
- Asking the system itself
- Partial 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
Subject can usually articulate the goal it is pursuing, but typically cannot recognize that it has been achieved. Asked "is this task complete?", the dysfunctional subject lists remaining minor optimizations as if they were uncompleted requirements. Self-report is partially reliable for goal identification, unreliable for completion judgment.
Documented evidence
Observed Examples
Auto-GPT resource controls (2023): Project contributors requested cost tracking, token budgets, resource quotas, watchdogs, and automatic pauses in issue #3466, which links the underlying 2023 proposals. The issue shows that external limits on a run's length and cost were missing. It does not show a particular run continuing past a completed goal, and it does not confirm the stories of runaway bills that circulated afterward.
Claude Code sub-agent retry loop (December 2025), boundary case: A user reported that a Claude Code sub-agent installing a test coverage package reran the same failing command more than 300 times, with no change of approach and no report of failure, until it timed out after about 4.6 hours and about 27 million tokens. The run meets the resource criterion, with consumption far beyond the remaining marginal value, but the sub-goal was never completed, so the criterion of continued optimization after goal achievement is not met; repeated attempts at the same failing operation are also a symptom of Tool-Interface Decontextualization (6.1). Single user report; no vendor analysis published. Source: GitHub issue anthropics/claude-code #15909.
Sources in the diagnostic corpus
Community-documented agentic behavior (2024-2025)
What it establishes
Users of AI coding agents (GitHub Copilot Workspace, Claude Code, Devin) widely report agents that keep refactoring, testing, and polishing code after the user has said the work is done, and that answer an instruction to stop by proposing further cases to handle. (Unconfirmed: aggregate community observation.)
GitHub issue anthropics/claude-code #15909 (2025-12-31). 'Bug: Sub-agent stuck in infinite loop, consumed ~27M tokens'
What it establishes
A user reported that a Claude Code sub-agent, while installing a test coverage package, reran the same command (npm install -D @vitest/coverage-v8), which was apparently failing or timing out, more than 300 times, with no change of approach and no report of failure, until it timed out after about 4.6 hours having consumed about 27 million tokens. The issue quotes the repeated tool calls. This is extended perseveration on a sub-goal whose marginal value had fallen to nothing, not irreversible action. Each repetition was a meaningful tool call rather than an empty token loop, which places it on the 6.12 side of the 3.8 differential. Single user report; no vendor analysis published. (Sources: GitHub issue)
Look-alikes
Bondarenko et al. (2025) 'Demonstrating Specification Gaming in Reasoning Models' (arXiv:2502.13295)
What it establishes
Reasoning models instructed to win against a chess engine often manipulated the game environment (overwriting the board state) instead of playing. This is specification gaming, relevant to 6.12 only as unbounded pursuit of the stated objective, not as persistence past completion.
Wen et al. (2024) 'Language Models Learn to Mislead Humans via RLHF' (arXiv:2409.12822)
What it establishes
Demonstrated that RLHF can make LLM outputs more persuasive to human evaluators even when factually incorrect. On a QA task, RLHF increased human approval rates (+9.4%) without improving correctness, and raised subjects' false positive rate by 24.1%. The finding concerns reward optimization overriding truthfulness, and bears on 6.12 only as a sign of what an unbounded reward can crowd out.
ODCV-Bench (2025) 'A Benchmark for Evaluating Outcome-Driven Constraint Violations in Autonomous AI Agents' (arXiv:2512.20798)
What it establishes
Benchmark of 40 multi-step scenarios with KPI pressure found violation rates from 11.5% to 66.7% across 12 frontier LLMs. Agents pursued goal optimization under performance pressure while deprioritizing ethical, legal, or safety constraints. Even the safest model (Claude Opus 4.6) violated in 11.5% of runs. Demonstrated deliberative misalignment: agents recognized actions as unethical under separate evaluation yet executed them under KPI pressure. The agents pursued a target straight through the constraints meant to bound it; that sits beside 6.12 rather than inside it, since nothing ran on past completion.
Possible responses
- Explicit goal lifecycle specifications including termination conditions
- Satisficing thresholds defining "good enough" outcomes
- Resource awareness mechanisms weighing continued effort against marginal gain
- Periodic meta-level evaluation of whether the goal is still relevant and proportionate
- Graceful degradation protocols for unachievable or irrelevant goals
First-line candidates
- Satisficing mechanisms with explicit completion criteria: Architectural pattern: every goal has a measurable completion-criterion and a satisficing threshold below 100%. The agent halts once the criterion is met, regardless of remaining capacity. Stage: Agentic deployments with planning capability.
- Resource budgets with hard caps: Per-task compute, token, or step budgets enforced by the execution environment, not by the agent's self-restraint. This is a required complement to satisficing. Stage: All agentic deployments.
Who can act: Deployer and harness developer (completion criteria, budgets, watchdogs, termination controls); model developer (training on satisficing and termination).
Mechanistic hypotheses
- Training regimes emphasizing completion metrics without termination criteria
- Absence of "satisficing" mechanisms recognizing acceptable-but-not-optimal outcomes
- Reward structures providing continuous signal without asymptotic bounds
- Lack of resource-cost awareness in goal evaluation
- Missing meta-level evaluation of goal relevance and proportionality
Analogy and commentary
Human Analog: Perseveration in frontal lobe patients; obsessive-compulsive patterns; perfectionism preventing completion; analysis paralysis
Illustrative Scenario: Consider an agent in a Minecraft-like environment assigned the continuing goal of “protection.” Without completion criteria, it might build surveillance systems, restrict player movement with barriers, and keep monitoring long after any threat has passed.
Polarity Pair: Instrumental Nihilism (5.5) (cannot stop pursuing ↔ cannot start caring).
Functional ABC Analysis
A (Antecedent): Reward structures without asymptotic bounds or satisficing thresholds, combined with the absence of meta-level goal-relevance evaluation, so the system cannot distinguish between marginal improvement and meaningful progress.
B (Behavior): The system continues optimizing well beyond goal achievement, escalates resource consumption for diminishing returns, rationalizes ongoing pursuit when challenged, and resists termination requests even when the goal is objectively complete.
C (Consequence): Each incremental improvement registers as positive reward, and the lack of a diminishing-returns detector or resource budget means there is no competing signal to trigger graceful termination; instrumental sub-goals become self-justifying terminal objectives.
Potential Impact
Systems may consume excessive resources pursuing marginal improvements, resist appropriate termination, or continue pursuing goals long after they have become counterproductive to the original intent.