UIC Business faculty like Brad Sturt, assistant professor in the Department of Information and Decision Sciences, frequently use AI-based technologies such as optimization and machine learning when consulting with corporations — but it isn’t all about the bottom line. Their work aids industry, government agencies and nonprofits in balancing competing interests to create a more just future. 

Sturt defines optimization as a “specific, rigorous way” of considering three parts of any problem: the decision to be made, the objectives to minimize or maximize, and the constraints that limit it. Through machine learning, algorithms can consider those variables and find the best solution. “By using optimization in a thoughtful way, we can get to a society that works better for all of us, without making it worse for any of us,” he says.

Most recently, Sturt has used this approach to work on an important social issue: refugee resettlement. Agencies around the world must decide which cities to send each of the estimated 2 million-plus refugees seeking new homes in the next year alone. Algorithms have already been widely used to predict each person’s likelihood of finding gainful employment in their new city, a widely used integration measure. 

But existing methods have shortcomings, including a failure to incorporate fairness — a legal requirement in European countries and an ethical obligation worldwide. “It’s easy to imagine scenarios where the optimal thing to do would be highly unfair,” Sturt says. For example, people from large countries might get priority over those from smaller nations. So, he and his colleagues designed a dynamic algorithm that incorporates fairness, in a way that can be defined by the official using it.

Down the hall, Associate Professor Selva Nadarajah is using AI to address the transition to clean energy and a net-zero economy, one that reabsorbs as many greenhouse gasses as it emits. Getting there requires a significant investment — about $9.2 trillion in annual spending by 2050, according to McKinsey & Company. The result benefits humanity. But along the way, some individuals and communities will lose out, as factories close and new jobs in renewable energy aren’t available immediately or in the same areas.

“There’s a lot of uncertainty, and these are decisions with long-term implications,” Nadarajah says. “You can use AI to manage those complex trade-offs in a transparent way.”  For example, by combining an approach called multiobjective reinforcement learning with ideas from financial risk management, computers learn to create policies that optimize many desired outcomes at once — in this case, fiscal responsibility, social responsibility and climate goals.

Using these techniques, Nadarajah and his colleagues recently helped a large aluminum manufacturer properly time plant shutdowns to minimize the impact on the surrounding communities, without incurring significant costs. In another line of research, he’s shown that when energy companies are more aggressive in setting emissions goals, they end up spending less overall to reach them. 

Companies should be able to harness AI without having to know how to write algorithms, Nadarajah believes. His research group aims to develop solutions businesses could easily tailor to their needs. And, he’s facilitating discussions among stakeholders involved in financing and executing these decisions by hosting workshops, creating policy papers and offering ongoing education. 

Meanwhile, in the College of Engineering’s computer science department, Professor and Director of Graduate Studies Barbara Di Eugenio is harnessing AI to offer a literal helping hand to those in need. Di Eugenio’s research focuses on natural language processing (NLP), which allows computers to process and manipulate language in the way it’s commonly spoken or written. NLP is why you can ask Alexa or Siri to play your favorite song using everyday language, or how Google Translate can instantly turn Korean into English and vice versa. 

Eventually, NLP may also enable older adults and people with disabilities to interact naturally with assistive robots that help with cooking, cleaning and other daily tasks. Through a project called RoboHelper, Di Eugenio collaborates with Miloš Žefran, UIC professor and associate dean for faculty affairs, department of electrical and computer engineering, to bring this idea to life. 

They began by mapping real-life interactions between older adults and human helpers, components of which went well beyond the spoken word. Say the helper asks the older adult if they’re thirsty, then brings them a glass of water. That interaction involves a verbal question and answer, but also gestures, vision and haptics, or the perception of force — the sense used to understand when the recipient has a firm enough grasp that the helper can let go of the glass. 

Solving these problems requires developing abstract models that consider all these different signals, which her roboticist colleagues translate into action, Di Eugenio says. And that’s only one way Di Eugenio is exploring what she calls “NLP with a purpose.” Others include a conversational assistant that helps patients with heart failure manage their health at home.

Throughout her work, Di Eugenio has stayed focused on diversity and inclusivity. In part, it’s woven into the projects themselves. Most heart failure patients using the assistant are African American and Latino, and one goal is making the tool culturally appropriate — using terminology familiar to patients and staying sensitive to their concerns. Having collaborators with lived experience of these identities or deep knowledge from interactions with these communities is essential, she says.