Why RBT Data Bias Is a Professional Issue

Clinical decisions are made from the data you collect. If your data is biased — even unintentionally — the BCBA may design a program around a problem that doesn't exist, or fail to address one that does. Data integrity is therefore both a measurement issue (covered in the Measurement domain) AND a professional conduct issue.

Common Sources of Observer Bias

  • Expectancy bias: You expect the new intervention to work, so you "see" improvement that isn't there
  • Recency bias: A loud, recent behavior gets over-counted relative to quieter ones
  • Halo effect: A child you like is rated more favorably across unrelated targets
  • Drift: Over weeks, your definition of the target behavior subtly shifts
  • Reactivity: Knowing you're being observed (by the BCBA) changes how you collect data
  • Convenience: Skipping a target you find tedious, then back-filling data later
  • Pressure: Wanting the data to look good for the family or supervisor

How to Minimize Bias

  • Use operational definitions: refer back to the written target definition periodically, not your memory of it
  • Collect data DURING the behavior, not after — memory is unreliable
  • Use mechanical aids (counters, timers, apps) rather than mental tallies
  • Be willing to record "bad" data — bad data are still valid data
  • Request periodic Inter-Observer Agreement (IOA) sessions with a peer
  • When in doubt, document the uncertainty: "Unable to determine if response met target criteria due to obstruction"

Data Falsification

The most severe data integrity violation is falsification — recording data for events that did not occur, or altering recorded data after the fact. This is grounds for termination, BACB sanctions, and (when billing is affected) criminal fraud charges. There is never a justification.

What to Do If You Realize You Have Biased Data

Disclose to your BCBA immediately. Most biases can be corrected going forward without consequence; the violation is in concealment, not in being human. The agency may need to restart baseline data or note the affected period in the clinical record. Honest correction protects everyone.