Known patterns worth applying
Automation bias, attentional residue after interruption, digital wellbeing, and workplace burnout all have substantial research behind them. We apply these findings rather than restating them as discoveries.
Research & why this matters
Curious.
Careful. Clear.
We take emerging questions seriously without presenting them as settled conclusions. Every claim on this site carries a confidence label, and we would rather publish an open question than a confident guess.
Established01
Emerging02
Open03
Confidence three levels
Evidence language
Three levels
of confidence.
If a claim cannot survive a confidence label, it should not survive the editing process either.
Signals worth studying
Early work on generative AI, cognitive engagement, reliance and social behaviour is suggestive and inconsistent. We treat it as a live question, not a foundation.
Hypotheses, not headlines
Some of the most important effects — long-run autonomy, synthetic intimacy, trust under synthetic media — cannot be claimed yet. We say so on the page where it matters.
No citation appears on this site that we have not read, and none are invented
A research agenda
Questions to answer
before AI becomes invisible.
These are the questions a longitudinal study would need to take seriously. None of them have a clean answer yet.
01
If an AI system is optimised to be agreeable, what happens when it becomes someone's dominant feedback loop?
02
If every small uncertainty is resolved instantly, do people lose tolerance for unresolved questions?
03
How might long-term AI companionship change expectations of real human relationships?
04
When any media may be synthetic, how do people maintain functional trust without permanent scepticism?
05
What safeguards are needed when a system simulates a person who has died?