Olivier Toubia

Glaubinger Professor Business, Columbia Business School

My colleague Malek Ben Sliman and I have developed a course called Generative AI for Business, which is currently taught to MBA and Executive MBA students in the Business School. The course was created in response to the rise of Generative AI. This course is for students interested in leveraging Generative AI (GenAI) in their professional lives. It aims to provide an in-depth understanding of GenAI, its wide range of applications, and the value it can generate for businesses.

Approach to teaching and learning in the age of AI

We try to follow a balanced approach along at least two dimensions.

First, we discuss both the potential benefits and challenges stemming from business applications of GenAI. We discuss not only the ways in which Generative AI can create value for business and society (“the bright side”), but also ways in which it could be detrimental (“the dark side”).

Second, we balance theory and practice. The theory component of the course is based on the fast-evolving body of knowledge on Generative AI coming out of research in academia and industry. The practice component involves guest speakers who share their experiences with Generative AI, student presentations, and hands-on guided work.

One example of hands-on guided work is an individual-level project we call the “break GPT experiment.” Using the OpenAI API, each student designs, runs and analyzes an experiment to explore potential biases and/or unintended effects of prompt architecture on GPT output. For example, one student might formulate the hypothesis that GPT is biased against people who use the “they” pronoun. They might design an experiment in which they create a job interviewing scenario, with three candidates who vary on multiple dimensions (including pronoun as well as education and years of experience), and ask GPT to choose one of the candidates. Using an experimental design that systematically varies the choices presented to GPT, the student may explore whether candidates with “they” pronouns are more or less likely than chance to be selected (given candidates are balanced on all other dimensions).

Teaching students to leverage AI and develop their AI literacy 

The course relies heavily on the use of APIs (Application Programming Interfaces) as the preferred way to engage with Generative AI in a scalable and robust manner (i.e., beyond standard “chat” interfaces). Although many of our students do not have programming experience, they all learn how to leverage the OpenAI API using Python. They run various experiments and complete various projects throughout the course using the API.

We bring up many ethical and societal issues with AI, e.g., discrimination, labor market implications, copyright issues, and concerns for their consideration. We let students decide how they would like to deal with these issues in their professional lives.

We encourage students to be critical and to evaluate AI output in a thoughtful and thorough manner. 

Lessons learned from teaching and learning with AI 

It is important to stay humble and realize that no single person can claim to know everything there is to know about applications of Generative AI to Business. It is important to develop a collaborative mindset where students and instructors exchange their respective knowledge and expertise.  

To share an example of student-instructor exchange, it is challenging to maintain relevant, up to date readings for the course. In addition, assigning the same readings to all students limits the diversity of perspectives in the classroom. Therefore, this year we plan to assign one common reading for each class, and give an assignment to each student to find another reading related to the topic of the day, and submit that reading with a summary before class.

Advice for colleagues on leveraging AI for teaching and learning

The efforts are considerable as the state of knowledge in this area is constantly evolving and materials need to be updated accordingly. But the payoffs are also significant given the strong interest from students, alumni and the general public. As academics, we have an important role to play in the development of AI, not just on the technical side but also in thinking through its implications for humans, societies, businesses, etc.