CARE-AI Seminar: What, Really, Are Simple Models Good For?
Works on interpretability tend to focus on the limitations of complex blackbox models. It is tempting to assume that simple models can achieve what black-box models cannot. This paper cautions against this temptation.