Why it’s still open
Building with AI can be absorbing, productive work — or a loop of re-prompting, near-misses and new projects that never ship. Gambling research offers candidate mechanisms: near-misses under personal control, losses that look like wins, the pull of uninterrupted continuity. None has been tested on AI tools. Candidate markers include whether output ships, how well people predict their own progress, and how they feel after a session.
What would count as an answer
Studies linking tool telemetry (success rate, re-prompts, session length) with shipped outcomes and people’s state afterwards.
Papers that tackle it
- Schull 2012 - Addiction by design — The zone: continuity over winning.
- Clark et al 2009 - Gambling near-misses — Near-misses motivate when you’re in control.
- Stack Overflow 2025 - Developer survey AI — ‘Almost right’ is the top frustration.
- Becker et al 2025 - METR developer productivity — Felt vs. measured progress.
- Ranganathan Ye 2026 - AI intensifies work — Work that expands to fill the day.
- Ghibellini Meier 2025 - Zeigarnik and Ovsiankina effects — The pull to resume unfinished work.
Methods in play
Who’s closest
Placed by the editor, v1 — corrections welcome.
Related questions
Know a paper that belongs here, or think this is framed wrongly? Suggest a paper or a correction (forms on the Atlas home).