Skip to contents

Records, on a run, what its results may be used to argue. The four types, from weakest to strongest: "instrument_pilot" (the run characterizes an instrument, not a population), "theory_probe" (it explores a mechanism or hypothesis), "coding" (it annotates data, with reliability reported), and "calibrated_inference" (it estimates a quantity validated against human data). The strongest type is refused unless a calibration is attached (see agent_calibrate() / attach_calibration()), so a simulation cannot be relabeled as a population estimate without the evidence.

Usage

mark_claim_type(
  run,
  type = c("instrument_pilot", "theory_probe", "coding", "calibrated_inference")
)

Arguments

run

An object accepted by as_agent_run().

type

One of "instrument_pilot", "theory_probe", "coding", "calibrated_inference".

Value

The run (an agent_run), invisibly, with the claim type recorded.

Details

The label flows into report(): prose that would overstate the claim is rewritten or refused for anything short of calibrated inference.

Examples

if (FALSE) { # \dontrun{
run <- as_agent_run(my_deliberation)
run <- mark_claim_type(run, "theory_probe")
report(run)   # prose is scoped to a theory probe, not a population claim
} # }