If you've had any exposure to Bayesian statistics, it's very easy to explain the approach. You're basically tokenizing the email by word, and then predicting the probability that P(email is spam|token).
I would expect any data science candidate to have (at least) a basic understanding of Bayesian statistics.
