Post-deployment psychometric-biometric drift monitoring

In eyeprocess: Harmonize Eye-Tracking, Pupillometry, Biometrics, and Psychometric Process Data

knitr::opts_chunk$set(collapse = TRUE, comment = "#>", eval = FALSE)
library(eyeprocess)

Drift monitoring links psychometric parameters to response-process and device/stimulus context. A flag is a governance signal for review, not evidence that an item has leaked or been compromised.

spec <- process_drift_spec(
  baseline = "first_batch",
  difficulty_limit = 0.40,
  gaze_validity_drop = 0.10,
  luminance_limit = 25
)

drift <- audit_process_drift(
  deployment_monitor,
  item = "item_id",
  batch = "deployment_batch",
  metrics = c(
    "irt_difficulty", "irt_discrimination", "rt_ms", "dwell_ms",
    "pupil_bc", "valid_gaze_prop", "screen_luminance"
  ),
  spec = spec
)

process_drift_alerts(drift)
plot(drift, type = "trajectory", metric = "irt_difficulty")
plot(drift, type = "delta", metric = "dwell_ms")
plot(drift, type = "heatmap")
plot(drift, type = "control", metric = "valid_gaze_prop")

Stratified helpers support device, site, vendor, and stimulus-version audits:

drift_by_device(deployment_monitor, device = "device_id")
drift_by_site(deployment_monitor, site = "site_id")
drift_by_vendor(deployment_monitor, vendor = "vendor")
drift_by_stimulus_version(deployment_monitor, stimulus_version = "stimulus_version")

eyeprocess documentation built on Sept. 28, 2026, 5:08 p.m.