Bounded Fictional Analysis

A portable method for making invisible system effects visible.

A methodology for studying system dependency, cognitive scaffolding, and non-reversible transformations

---

Visual Overview

A system’s removal does not remove its effects.

---

Overview

Bounded Fictional Analysis is a research method for examining how systems reshape cognition, documentation, and coordination over time. It uses deliberately constructed fictional scenarios to isolate dynamics that are difficult—or impossible—to observe cleanly in real-world systems.

Best used when:

• Systems are too entangled to isolate variables
• Gradual changes obscure causality
• Ethical constraints limit experimentation
• Counterfactuals must be explored before implementation

---

Core Framework

Five phases model the lifecycle of system integration:

1. Baseline — Pre-system state → Existing capabilities, constraints, and cognitive models
2. Introduction — Early adoption → Habit formation, initial restructuring
3. Steady-State — Full integration → Capability expansion + capability atrophy
4. Removal — System disappears → Breakage, persistence of behaviors, cognitive mismatch
5. Adaptation — New equilibrium → What remains, what recovers, what is permanently altered

---

What This Reveals

Across domains, four recurring patterns emerge:

Behavioral Lock-In
Actions persist after their purpose disappears

Orphaned Sophistication
Cognitive frameworks depend on missing infrastructure

Context Discontinuity
Artifacts persist, meaning does not

Residual Infrastructure
Structures shaped by systems outlive them

---

Where It Applies

• AI systems — memory, tooling, platform discontinuity
• Documentation systems — archival decay, format drift
• Cognition — scaffolded thinking, habit persistence
• Systems design — degradation, migration, legacy handling

---

How to Use

1. Design a bounded fictional scenario

   • Clear system boundaries
   • Observable effects
   • Toggleable presence/absence

2. Walk the five phases
   Track changes across:

   • Individual cognition
   • Social coordination
   • Structural dependencies

3. Extract patterns
   Focus on persistence, mismatch, and irreversibility

4. Map to real systems
   Use patterns as explanatory lenses—not proof

---

Strengths

• Isolates confounded variables
• Enables safe counterfactual exploration
• Makes hidden dynamics visible
• Generates testable hypotheses

---

Limitations

• Not predictive or quantitative
• Does not replace empirical validation
• Does not prove causality
• Cannot substitute real-world observation

---

Validation Signals

A useful analysis shows:

• Cross-domain recognition
• Explanatory clarity
• Predictive usefulness (pattern-level)
• Generative insight
• Clear boundaries

---

Ethics

When applied to AI systems:

• Explicit fictional framing
• No covert experimentation
• Transparent intent

When mapping to humans:

• Patterns ≠ proof
• Context matters
• Individual variation persists

---

Quick Reference

At each phase, examine:

• Cognition
• Coordination
• Infrastructure
• Persistence

Look for:

• Lock-in
• Orphaning
• Discontinuity
• Residue

Ask:

• What changed?
• What persisted?
• What broke?
• What cannot be undone?

---

Examples

The framework is best understood through concrete scenarios. Below is a minimal walkthrough demonstrating how patterns emerge across phases.

Worked Example — LLM Memory & Tool Dependency

Scenario: A team uses an LLM system with persistent memory and tool integrations for daily workflows.

Phase 1 — Baseline

• Knowledge stored manually (notes, docs)
• Context switching is explicit
• Reasoning is slower but self-contained

Phase 2 — Introduction

• LLM begins assisting with recall and drafting
• Users offload memory and structure
• Prompts become externalized thinking

Phase 3 — Steady-State

• LLM becomes primary interface for:
  • recall
  • synthesis
  • coordination
• Users rely on:
  • saved context
  • tool chains
  • system-specific workflows

Emerging effects:

• Faster iteration
• Reduced internal memory load
• Increased system dependency

Phase 4 — Removal

System becomes unavailable (API change, outage, migration)

Observed effects:

• Prompt habits persist but fail
• Users attempt workflows that no longer function
• Knowledge exists but is harder to access
• Coordination slows abruptly

Discontinuity shock occurs

Phase 5 — Adaptation

• Users rebuild partial workflows
• Some behaviors persist (structured prompting, decomposition)
• Some capabilities degrade (instant recall, synthesis speed)

Patterns Observed

• Behavioral Lock-In — Continued reliance on prompt-based thinking
• Orphaned Sophistication — Workflows assume capabilities (memory, chaining) that no longer exist
• Context Discontinuity — Stored information loses usability without system mediation
• Residual Structures — Team processes remain shaped by prior system

Insight

The system did not just assist cognition—it restructured it. Its removal exposes not absence, but transformation.

---

Citation

Bounded Fictional Analysis: A methodology for studying system dependency and non-reversible cognitive transformations through deliberately constructed fictional scenarios.

---

For a complete catalog of related research:
📂 AI Safety & Systems Architecture Research Index

---