Spot the Biased Label in the Patient Reminder Model
Identify where model bias can enter through historical labels, missing populations, and average-only evaluation.
The reminder policy goes to the medical directors tomorrow. Where could bias enter this patient reminder model? No-show label used without disruption context Correct. Historical labels can reflect access barriers. Treating a transit-disruption miss as patient reliability can teach the model the wrong mechanism. Excluded walk-in and interpreter-assisted visits Correct. Missing groups create coverage or selection bias. The model may perform worse for patients whose visits were not represented. Only an overall accuracy check Correct. Average performance can hide subgroup failures. Bias review needs sliced metrics tied to the harm the intervention could cause. Shared appointment calendar board Weaker. A…
Sign up free: one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in