calibrate
Validate AI-simulated survey results against real CSV answers by checking covariate overlap, per-audience calibration, and an AIPW-corrected rate, then write a trust verdict to calibration.json.
Instructions
Check a run's simulated survey against real answers (absolute path to a CSV with segment,country,gender,age,y where y is yes/maybe/no or 0-1; 50-300 rows is enough): covariate overlap, per-audience calibration, a trust / correct / walk-away verdict and an AIPW-corrected rate. Writes calibration.json into the run folder.
Tool call template
Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.
{
"id": 1,
"jsonrpc": "2.0",
"method": "tools/call",
"params": {
"arguments": {
"run": "<run>",
"real_csv": "<real_csv>"
},
"name": "calibrate"
}
}Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| by | No | ||
| map | No | ||
| run | Yes | ||
| field | No | ||
| real_csv | Yes |