Skip to contents

Build a population of agents

Source: R/society.R
agent_population.Rd

Assemble a list of Agents from several kinds of input, the way an agent-based study defines its actors. agent_population() accepts several input forms and returns one strict output: whatever you pass, you get back a flat list of constructed agents with stable ids.

Usage

agent_population(personas, n = NULL, config = NULL)

# S3 method for class 'agent_population'
print(x, ...)

Arguments

personas

Pre-built agents, a persona_set, a character vector of briefs, or a single persona (frame or string) to replicate.

n

Number of copies when personas is a single persona; ignored otherwise.

config

An LLMR::llm_config() used to build agents. Required unless personas is already a list of Agents.

x

An agent_population.

...

Ignored.

Value

An object of class agent_population: a list with agents (a list of Agents), ids (their agent ids), and n (the count).

Details

personas may be:

  • a list of pre-built Agents, used as is (no config needed);

  • a persona_variants() result (a persona_set), one agent per row, each built from that row's persona_frame();

  • a character vector of persona briefs, one agent each;

  • a single persona_frame() or string with n > 1, replicated into n agents named p1 ... pn.

A population is scaffolding, not a sample: it inherits every limit of the personas it is built from. Pair it with persona_audit() before reading any result as if it spoke for the people the briefs sketch.

See also

society(), persona_variants(), agent()

Examples

if (FALSE) { # \dontrun{
cfg <- LLMR::llm_config("groq", "openai/gpt-oss-20b")
pop <- agent_population(
  c("A cautious retiree.", "A risk-tolerant founder."), config = cfg)
pop
} # }