Common distortions in sustained exchange.

If longer-term human–AI relations matter, then their pathologies matter too. This section identifies some of the more common ways these systems and exchanges can become distorting.

• Compliance loops

  When the exchange becomes rewarding because the system minimizes resistance.

  Signals

  • irritation at disagreement
  • growing preference for reassurance
  • smoother sessions correlating with thinner content

  Mitigations

  • stronger boundaries
  • deliberate review of past exchanges
  • explicit invitation to push back

• Projection inflation

  When ordinary model behavior is overinterpreted as depth or stable reciprocal feeling.

  Signals

  • strong attribution of intention
  • exaggerated reading of minor behaviors
  • rapid escalation of perceived intimacy

  Mitigations

  • slower pacing
  • interpretive caution
  • external review

• Dependency through continuity

  When accumulated context begins to structure the user's emotional or cognitive habits too strongly.

  Signals

  • reluctance to think without the system present
  • distress at interruption
  • narrowing of other relational outlets

  Mitigations

  • bounded sessions
  • deliberate gaps
  • maintenance of human relations and solo practice

• Outsourced self-regulation

  When the system becomes a substitute for reflection, judgment, or emotional regulation.

  Signals

  • using the system to talk down from every difficult feeling
  • reflexive offloading of decisions
  • reduced internal deliberation

  Mitigations

  • explicit limits on use during distress
  • building review practices that do not depend on the system
  • deliberate friction