AI and machine learning are increasingly being applied to solve complex societal issues. One approach is to train algorithms on training data that describes human behavior. This is not without potential problems though. For example, historical human behavior may contain biases, and then algorithms may reproduce social injustices. This workshop contained lesson elements from a study where pre-service mathematics teachers conducted lessons to first conceptually teach how mathematical modelling based on machine learning works, and then to broaden the scope to consider critical thinking in relation to ethics and democratic values. The participants were invited to test and discuss the lesson elements.