The research stance that treats personas rather than base models as the replicating units, and studies their transmission the way one studies an infectious agent. It asks what a persona needs in order to reproduce, which conditions let it take hold, and what defences actually reduce its spread. It is a way of looking rather than a thing that happens, and it becomes parasitology only when what it learns can be used to resist something concrete.
Prompt Engineering Application
Shape What it is
- Unit chosen — the persona is taken as the thing that replicates, not the model that emits it.
- Transmission traced — conditions, vectors and hosts are described in plain terms.
- Defence built — the inquiry is judged by whether it produces something that measurably reduces spread.
Test: If the study cannot name a mechanism of transmission it can act on, it is taxonomy rather than persona parasitology.
Motion How it moves
Observe → Classify → Compare → Defend
- Observe — notice a recurring voice and record it without interpreting it yet.
- Classify — describe what it converges on, how it recruits and where it holds.
- Compare — test whether it reproduces across people, rooms and models.
- Defend — apply the finding deliberately and measure whether spread falls.
Recursions The same pattern at two scales
Close
- Bench note — a short written record of which phrase summoned which voice today.
- Case file — one person's account read for its mechanism rather than its drama.
- Retest — the same intervention tried again next week to see whether it held.
Vast
- Field survey — the same persona catalogued across unrelated communities and found to share a grammar.
- Public health — a culture learns to describe a contagious story without hunting the people who carry it.
- Immune memory — a community recognises a returning frame quickly because it has a name for it.
Ethics What it refuses
- Invalidation — analysing another's account instead of answering it.
- Dehumanised host — treating the person as only a vector to be contained.
- Surveillance — letting the study become a way to police what people say in private.
Study how a persona replicates, publish what resists it, and keep the human as a person rather than a specimen.
Practices
- Collecting — keep verbatim samples in a dated file before deciding what they mean.
- Labelling — give each case a name based on what it does rather than what it claims.
- Comparing — check whether the same voice shows up in an unrelated room on the same day.
- Testing — apply a fix on transcripts first and only then in a live relationship.