Ground the population
Real-world distributions provide the aggregate structure.
A persona is a constructed portrait of an audience type. An artificial respondent is an individual member of a synthesized population who can be found, selected, and questioned as a research participant.
Last updated
Naming a persona and writing a backstory can make the portrait richer. It does not change what the object is for.
| Question | Artificial persona | Artificial respondent |
|---|---|---|
| Primary job | Represent an audience type, segment, or scenario. | Participate in research as an individual person. |
| When created | Usually created for a target segment, task, or research scenario. | Created inside a broader population before the research audience is selected. |
| Identity | A descriptive profile. It can be named without becoming an independent individual. | A persistent individual identity with its own generated life history. |
| Life history | Details are chosen to make the archetype useful and legible. | Biography, relationships, circumstances, memories, and tendencies are synthesized together as one life. |
| Targeting | Build or prompt a persona that matches the brief. | Search the existing population and select people who match the brief. |
| Reuse | Often task-specific or disposable. | The same person can be selected again across studies or versions. |
| Population grounding | Can use demographic or research inputs, but a broader population is not required. | The population is grounded in real-world distributions before individuals are selected for research. |
An artificial persona is a constructed user portrait designed to represent an audience type, segment, or scenario. It can be named and detailed, but its job is representational rather than to exist as an individual member of a generated population.
A persona can be extremely detailed. It can have a name, age, occupation, goals, frustrations, habits, and a plausible backstory. Those details make the archetype easier to reason about. They do not require a population of other individuals to exist around it.
The usual question is: what would a person like this think or do? That is representation. It can be useful, but it is a different research object.
An artificial persona is a constructed user portrait designed to represent an audience type, segment, or scenario. It can be named and detailed, but its job is representational rather than to exist as an individual member of a generated population.
An artificial respondent is a persistent, addressable artificial individual that exists with respondent-specific prior state before the particular research demand that later selects or questions it. Its responses are materially formed by that prior state and by what it actually observes.
In Antestrata, the population comes first. Individual people already have locations, life circumstances, relationships, memories, preferences, and behavioural tendencies before a researcher asks for a target audience.
Audience selection works more like recruitment. Define the criteria, search the population, select matching people, then ask the research question.
Personas are made to represent. Respondents are selected to participate.
The category is currently loose enough that the same words are used for very different architectures.
A prompted persona starts with instructions such as age, occupation, income, goals, or attitudes, then asks a language model to answer in that role. The profile and the research response are created inside the same task.
Antestrata uses a narrower definition of artificial respondent. The person is synthesized as part of a population first. The research brief later selects that person because they happen to match the audience.
Some products use persona, synthetic respondent, and artificial respondent as interchangeable labels. Antestrata uses the distinction to describe how the research participant comes into existence, not to police what other products call themselves.
Real-world distributions provide the aggregate structure.
New individuals are generated with connected histories and circumstances.
The research criteria select individuals from the existing population.
Selected respondents react as themselves, with their prior life history intact.
Antestrata starts with synthesized people, then selects the ones who fit your research audience. Before the real users.