Course content
Most managerial decision problems require answers to questions such as “what happens if?”, “what is the effect of X on Y?”, or “was it X that caused Y to go up?” In other words, practical business decision-making requires knowledge about cause-and-effect. While standard tools in machine learning and AI are designed for efficient pattern detection in high-dimensional settings, they are not able to distinguish causal relationships from simple correlations in the data. That means, most commonly used approaches to machine learning fall short in addressing pressing questions in business analytics and strategic management. This creates an important mismatch between the answers that these algorithms can provide and the problems that managers and strategists would like to solve. Which is why several leading companies from the tech sector and elsewhere (among them: Meta, Microsoft, Google, Amazon, Spotify, Zalando, and McKinsey) have started to heavily invest into their causal data science capabilities in recent years.
This course will provide an introduction into the topic of causal inference in machine learning and AI, with a focus on applications to practically relevant, data-driven business cases. The course is meant to be conceptual rather than technical, in order to bridge the gap between data science and management strategy, for better evidence-based decision-making. A variety of hands-on examples will be discussed that allow students to apply their newly obtained knowledge and carry out state-of-the-art causal analyses by themselves. The course will thereby loosely follow the structure of “The Book of Why” by Judea Pearl and Dana Mackenzie, which has ushered a new era of causal thinking in data science and machine learning upon its publication. In particular, students will be put into the position to detect sources of confounding influence factors that threaten valid causal conclusions, understand the problem of selective measurement in data collection, and extrapolate causal knowledge across different business contexts. By developing an overarching framework for causal data science, the course will also cover several standard tools for causal inference, which are often used in empirical research in business and economics (such as instrumental variables, regression discontinuity designs, A/B testing and experiments, etc.). Thus, while not a research methods course as such, this elective will nonetheless be highly relevant for students who plan to conduct a quantitative data analysis as part of their master thesis project.
See course description in course catalogue