AI Expert Spotlight: Georgia Perakis

Kate O'Sullivan

Oct 31, 2024

We asked several MIT experts about their latest projects and what they see as the most exciting—and concerning—aspects of the AI boom.

Georgia Perakis, John C Head III Dean (Interim), MIT Sloan School of Management, Professor, Operations Management, Operations Research & Statistics

focuses her work on pricing and supply chains, using AI to help retailers better predict demand and determine optimal promotional strategies based on different variables.

She and her team have also applied AI to health care—for example, analyzing how to manage an emergency room case load based on different patient conditions, time of day, how many people are waiting for care, and how long treatment would take, among other variables.

How would you describe your work in the artificial intelligence space?

My research is very much in the space of AI (in particular, machine learning and optimization). I focus a lot on pricing and supply chains. Retailers are always trying to decide how best to do promotions for different products: how deep to discount, when to promote, how to promote, and what type of products to promote. We can create a model that predicts demand based on different variables. Currently, we are starting a new project to understand the impact of weather on sales at retailers such as Zara, the clothing retailer.

I also work in health care, where predictive and prescriptive models are important for hospitals. I’ve worked with the emergency department at UMass Memorial Hospital, looking at how to manage their patient load in the emergency department based on the different types of patient conditions, time of day, how many people are waiting for care, and how long treatment will take, for example. I’m also working on a new project with them about burnout for medical staff, trying to allocate patients in a fair way and avoid overworking staff members.

What is the biggest opportunity in working with AI?

There are so many opportunities! AI models allow us to process large amounts of data, run an algorithm, and run it fast. As a human being, you cannot process the same amount of data and determine optimal strategies in complex settings. But you have the experience to interpret it and use it with care. These tools can make your life and your job easier and as a result, enable you to do more meaningful work.

What do you see as the biggest challenge?

AI is a powerful technology with significant ethical implications. It must be developed and deployed with consideration for fairness and non-discrimination, transparency and explainability, privacy and data protection, accountability and responsibility, and both doing good and avoiding harm.

The rapid advancement of AI also risks exacerbating existing digital divides. We must identify ways to ensure fair access to AI technologies and their benefits. We need to prevent

the concentration of AI power in a few countries or companies. And we need to think about building capacity in developing nations in terms of AI education and infrastructure.

Another challenge is the need for accurate and non-biased data. If you provide the wrong data, AI models will give you the wrong recommendations—garbage in, garbage out. They may also provide a biased recommendation. Understanding biases and accounting for them when you build algorithms and use data is very important. As AI algorithms use a lot of data from the open internet, there is the risk of misinformation, among others. Being aware of such issues is the first step to protecting against them.