Audit policy feedback performativity

What it's for

Gives Gitrevio a post-deployment AI-governance control most dashboards omit: it tests whether acting on a model changed the relationship the model was trained to predict, while refusing a causal or person-level conclusion.

What you give it

Inputs split into evidence read from your connected systems, calibration your team owns, and numerical controls that affect precision but never the result's meaning.

{
  "bootstrap_draws": 500,
  "minimum_material_feedback_shift": 0.05,
  "observations": [
    {
      "cluster_id": "exposed-0",
      "epoch": "pre",
      "id": "feedback-exposed-0-pre-0",
      "outcome": 1,
      "policy_group": "exposed",
      "predicted_probability": 0.36
    },
    {
      "cluster_id": "exposed-0",
      "epoch": "pre",
      "id": "feedback-exposed-0-pre-1",
      "outcome": 1,
      "policy_group": "exposed",
      "predicted_probability": 0.38
    },
    {
      "cluster_id": "exposed-0",
      "epoch": "pre",
      "id": "feedback-exposed-0-pre-2",
      "outcome": 0,
      "policy_group": "exposed",
      "predicted_probability": 0.39999999999999997
    },
    {
      "cluster_id": "exposed-0",
      "epoch": "pre",
      "id": "feedback-exposed-0-pre-3",
      "outcome": 0,
      "policy_group": "exposed",
      "predicted_probability": 0.42
    },
    {
      "cluster_id": "exposed-0",
      "epoch": "pre",
      "id": "feedback-exposed-0-pre-4",
      "outcome": 0,
      "policy_group": "exposed",
      "predicted_probability": 0.44
    },

What you get back

This is the actual output of running the example above — computed by the same function the platform calls, not an illustration.

{
  "assumptions": [
    "Exposed and comparison clusters follow comparable counterfactual trends absent deployment; baseline balance and overlap are diagnostics, not proof.",
    "Prediction, outcome, policy exposure, epoch, cluster membership, and outcome maturity are frozen from their historical versions.",
    "The cluster bootstrap represents the dependence and assignment unit; interference across clusters is negligible or explicitly modeled elsewhere."
  ],
  "baseline_comparability": {
    "score_overlap_fraction": 1,
    "standardized_differences": {
      "outcome_rate": 0,
      "predicted_probability": 0
    }
  },
  "cluster_diagnostics": [
    {
      "calibration_residual_change": 0,
      "cluster_id": "comparison-0",
      "outcome_rate_change": 0,
      "policy_group": "comparison"
    },
    {
      "calibration_residual_change": 0,
      "cluster_id": "comparison-1",
      "outcome_rate_change": 0,
      "policy_group": "comparison"
    },
    {
      "calibration_residual_change": 0,
      "cluster_id": "comparison-2",
      "outcome_rate_change": 0,
      "policy_group": "comparison"
    },
    {
      "calibration_residual_change": 0,
      "cluster_id": "comparison-3",
      "outcome_rate_change": 0,
      "policy_group": "comparison"
    },
    {
      "calibration_residual_change": 0,
      "cluster_id": "comparison-4",
      "outcome_rate_change": 0,
      "policy_group": "comparison"
    },

How it works

Forecasting & survival — Estimate when something completes or fails, with censoring and unresolved work handled honestly rather than dropped.

1. 1 Freeze historical model probabilities, mature binary outcomes, deployment exposure, pre/post epochs, and the cluster-level assignment or dependence unit; require both epochs for every cluster and support in all four group-by-epoch cells.
2. 2 Aggregate prediction, outcome, calibration-residual, and Brier changes inside clusters, then subtract the comparison-group change from the exposed-group change so common operating drift is not mislabeled model feedback.
3. 3 Resample whole clusters independently within exposed and comparison groups to obtain uncertainty for every difference-in-differences metric; keep score-to-outcome feedback separate from an operating-level prediction or outcome shift.
4. 4 Gate interpretation on prespecified pre-deployment standardized balance and central score-range overlap, and call a feedback signature only when a calibration-residual or Brier shift is both practically material and directionally supported.

Before you trust it

Every tool in the catalog ships with the conditions under which its answer is meaningful — and the conditions under which it should abstain instead of guessing.

Calibrating it to your org

Calibration workflow

1. 1 Define the management decision, target outcome, aggregate unit, privacy boundary, cadence, and prediction/intervention horizon for this organization.
2. 2 Build a tenant-scoped historical cohort using only information available before each prediction or decision; preserve zero periods, censoring, assignment probabilities, and unresolved outcomes when the method requires them.
3. 3 Estimate statistical parameters on training history, but obtain costs, utilities, risk tolerance, practical-effect thresholds, capacity, and policy constraints from accountable decision owners.
4. 4 Validate on later time windows or held-out aggregate units at the deployment grain, against a simple baseline and the function-specific validation strategy.
5. 5 Deploy only if the returned decision clears evidence, overlap, calibration, robustness, and guardrail checks; warning, unsupported, schema-gap, and fallback decisions are abstentions.
6. 6 Monitor realized outcomes, data drift, coverage, and decision regret; recalibrate at a governed cadence or after a detected regime/definition change, never merely because a stakeholder dislikes the result.

Call it from your AI

You don't wire up 388 tools in your MCP client. The GitRevio MCP server exposes 18 tools, three of which let an agent search the catalog, read a tool's schema, and run it — so the assistant finds this one on its own.

gitrevio_capabilities_search
  { "q": "audit whether deploying a probabilitydriven policy" }
  → finds "audit_policy_feedback_performativity"

gitrevio_capability_describe
  { "capability_id": "audit_policy_feedback_performativity" }
  → returns the input schema and agent guidance shown on this page

gitrevio_capability_run
  { "capability_id": "audit_policy_feedback_performativity", "arguments": { ... } }
  → returns the result shown above

Works in Claude Desktop, Claude Code, Cursor, Cline, Continue.dev, Goose and Aider. See the MCP server.

Related tools

See every tool in Measurement integrity →