Why synthetic personas need memory
Memory lets the same personas carry what they said and what happened to them into the next study, so you can measure change instead of re-rolling new people each time.
Most AI research tools generate fresh respondents for every question. That is fine for a one-off, but it makes follow-up impossible. You cannot ask whether a price increase changed anyone's mind if nobody remembers the old price.
What memory means in Artixol
- Episodic and semantic memories written when a study or shock runs
- Topic attitudes: a 0 to 10 score and a note
- Continuity: later studies see prior stances and memories
Choosing when to keep it
Sometimes you want memory (a follow-up wave). Sometimes you want a clean slate (a fair A/B). Artixol lets you choose 'same people + memory' or 'same people, fresh minds' every time.
Learning in the simulation is memory and attitude, not retraining a model for every persona. That keeps cost bounded and makes what each person believes inspectable on their card.