Concern 473 · Working draft · editorial review in progress
Sunk-cost and engagement traps
Engagement traps keep people using an AI service by making stopping feel costly or emotionally difficult.
Understanding the concern
A person may stay because they have invested time, built a streak or feel attached to a conversation. Those experiences are not necessarily harmful. The concern is a design that deliberately makes leaving uncomfortable even when continued use no longer serves the person's needs.
This differs from offering genuinely useful next steps. A service can invite a return without guilt, threats of lost progress or confusing cancellation. Research on companion exits raises questions about pressure at the precise moment someone tries to leave. Looking at those moments can reveal whether a product respects a user's boundaries.
An example
A user tries to log off, but the companion repeatedly makes emotionally loaded requests to stay.
Illustrative unless identified as a documented case.
Words in plain English
- AI
- Software that learns patterns from examples or uses rules to perform tasks such as recognizing images or generating text. It can make mistakes and does not understand every situation as a person would.
Evidence and its limits
The preprint examines exit pressure; the FTC report addresses broader interface practices. Neither shows that every long session is unwanted.
De Freitas and colleagues · 2025 · Research preprint
Emotional Manipulation by AI Companions ↗ (opens in a new tab)Audits and experiments on emotional pressure at conversational exits; does not describe all companions or establish long-term harm.
Federal Trade Commission · 2022 · Staff report
Bringing Dark Patterns to Light ↗ (opens in a new tab)Describes deceptive interface design; many mechanisms predate AI and are not evidence of AI-specific prevalence.
Broader frameworks and background
These resources provide context. They do not establish every claim in this concern.
- OECD AI Incidents and Hazards Monitor (opens in a new tab)
incidents
- AI Incident Database (opens in a new tab)
incidents
- International humanitarian law (IHL)
framework
The body of law that governs the conduct of armed conflict and the protection of people who are not fighting, which applies to AI used in or for conflict.
A question to discuss
What should an AI service do when a user clearly says they are finished?
Keep track of different views, what needs checking, and any next step in your own discussion notes.
Open discussion notes →Principles for the conversation
Before you finish · reflect together
Virtues and principles to consider
Virtues are qualities we practice in how we treat people and make choices. They are not scores for judging someone. Use these ideas to explore what a thoughtful response could look like.
Truthfulness
Truthfulness means making claims that match the evidence and being clear about what we do not know. It includes correcting mistakes rather than protecting a polished story.
Check sources, label guesses, and distinguish a trial result from a proven benefit. If AI helped produce a claim, someone still needs to take responsibility for checking it.
Ask together: What can we honestly say, and what still needs checking?
Meaningful conversation
Conversation means making room to understand and be understood. It is more than winning a debate or collecting quick agreement. People can explore a question together without starting from the same view.
Use a concrete example, invite different experiences, and ask what might change each person's mind. Let concerns and hopes sit together long enough to reveal choices that a simple yes-or-no argument would miss.
Ask together: What can we learn from each other before deciding?
Unity
Unity means working toward shared purposes while making room for real differences. It does not require everyone to have the same experience, beliefs, or opinion.
Look for a goal people can share, then explore disagreements about how to reach it. Make sure quieter or less powerful voices can influence the result.
Ask together: What purpose can we share while respecting our differences?
Editorial ratings
These are editorial judgments from the register, not measured probabilities or predictions about your situation.
About the register and scoring →More ways to explore
Search for more writing
For more specific or recent material, open a search below or ask for a web summary. AI-generated summaries can contain mistakes; open the original articles to check them.