Teaching

AI models are becoming pervasive in science and management, while our understanding of their capabilities and limitations is still developing. This course prepares graduate researchers in management science to work effectively with AI, not only using it as a powerful assistant, but also recognizing when it fails and how to design systems and methods around its weaknesses. The next frontier is not simply applying AI, but shaping problems, experiments, and decisions in ways that align with its strengths while correcting for its limitations, using mathematical tools from computer science, economics, operations research, and statistics. Topics include the foundations of AI as a research tool; its strengths in pattern recognition and generative modeling via techniques such as deep learning; its limitations and common failures; and methods for building around these weaknesses, such as alignment and control. We also consider approaches to expanding AI's powers through fine-tuning, hybrid models, and structured integration. Syllabus

I designed this course in 2013 and have taught it every winter since.

The aim is to develop the skills needed to turn data and AI technologies into competitive advantage. Students build a rigorous foundation in predictive modeling and machine learning, which is what makes it possible to judge the capabilities and the limits of current AI technologies, including AI agents, and to know when to trust an output and when to override it. The course pays particular attention to the actionable insights that can be drawn from data and to the practical pitfalls of data-driven approaches. It covers statistical modeling, machine learning, and experimental design, with applications spanning advertising, eCommerce, finance, healthcare, marketing, and revenue management. Students work hands-on with real datasets using Python and AI technologies, learning to formulate business-relevant questions and answer them through data analysis. The central theme is that technical capability has to be combined with the domain expertise and business judgment that determine whether a data-driven decision actually works in practice. Students are expected to integrate this material with their existing proficiency in mathematical notation, algebra, probability, and basic statistics. Supported by DataCamp.