Lynn,

> I am preparing a presentation for our K-12 teachers about AI and
> Algorithms and bias.

I often hear people talk about algorithmic bias, but I don't know any
examples of that. However, because machine learning relies on training
data to induce models, biased samples can produce biased predictors.

Still, the underlying classification technique, whether it operates over
neural networks, decision trees, or probabilistic summaries, is not
itself biased, only the models that it uses to make decisions.

This is really no different from problems associated with older, more
traditional statistical models (e.g., logistic regressors) when they
are given nonrepresentative training samples.

I think it's important that you convey this idea to your students and,
ideally, avoid the term "algorithmic bias", which could lead to deep
confusion about the source of the problem.

Best wishes,
Pat Langley