HCS Score

The Human Confidence Score (HCS) is a single number between 0.0 and 1.0, computed from four models running on a session's 34-dimensional behavioral feature vector. See ML Pipeline for how each model works.

Formula

HCS = 0.55 × M3_score
    + 0.20 × M4_keyboard_score
    + 0.15 × M4_mouse_score
    + 0.10 × M4_touch_score

Thresholds

≥ 0.35human_verified

0.30 – 0.35human_likely

< 0.30synthetic

⚠️

These thresholds were selected via a threshold sweep against the validation set, not a formally published precision/recall curve. Treat 0.35 as the current production setting — if your use case needs a stricter or looser boundary, threshold on the raw hcs value yourself rather than relying on the verdict field alone.

Reading the response

json

{
  "session_id": "550e8400-e29b-41d4-a716-446655440000",
  "hcs": 0.7424,
  "m3_score": 0.8991,
  "verdict": "human_verified"
}

For single-page sessions with no navigation, leave the mouse/scroll/touch/motion/navigation features at exactly zero rather than approximating them — see Behavioral Features for the full schema.