TOPICS:
Recently published AI Research by Dr. Shavi Cooray focuses on people, not just technology.
Dr. Shavi Cooray, professor, School of Business and Technology, recently published her solo research article, “Socio-Technical Machine Learning - Putting the User in Control” in the Journal of Computer Information Systems published by Taylor & Francis. This journal is considered highly selective within its field, with an acceptance rate of just 3%.
Her research explores how artificial intelligence (AI) tools are designed and asks a simple but important question: Who should shape the AI technology we use every day? While AI systems are usually built by technical experts, Dr. Cooray’s work highlights the importance of involving the people, or end users, who actually use these tools in their daily work.
“Technology design, including AI/Machine Learning (ML) design, is consistently driven and controlled by those with extensive technology expertise, even though the people who will use the technology daily for their operations and decision-making are the end users,” explained Dr. Cooray. “If we want to ensure that the actual needs and problems of end users are met through the technologies they use, then we need users to be more in control of the technology design and development process.”
At the center of the paper is a key research question: “Can end users with limited technical knowledge realistically control the initial phases of AI and machine learning design?” Dr. Cooray explains that this question is critical because AI development is typically led by data scientists who must interpret user needs from the outside. “End-users have the hands-on experience and problem-domain expertise to define the complex socio-technical system within which the ML model should operate,” she said. “Typically, data scientists lead ML design based on their interpretations of the user needs. We argue that this can create a gap between what the user actually wants and what the data scientist believes the user wants. Such a situation can result in AI/ML based information products that do not satisfy end users' and organizations' actual needs, and lead to reduced transparency, trust, and accountability in AI/ML models.”
The paper argues for a more collaborative and socio-technical approach to AI development, one that brings together end users and technical experts from the onset. According to the research, this approach can help ensure that the final information products meet users' actual needs, satisfy organizational goals, and increase transparency and accountability across all phases of the AI/ML design cycle, including the dataset production process.
Dr. Cooray’s published research also aligns with Curry College’s Amplify AI Initiative, which recognizes that artificial intelligence is rapidly changing how people learn, work, and make decisions. As the College works to better understand AI’s potential, benefits, and limitations, the initiative emphasizes using AI responsibly, and in ways that support real human needs.
While many people, including college students, are focused on learning how to use AI tools, Dr. Cooray encourages them to also ask how they can add value in a world driving by AI. “I teach students the technology and business skills needed to not only use AI, but to also be actively involved in the design/development of AI/ML models. I teach students not only to be users of AI but also to be designers of AI, so they can add more value in the modern AI-driven workplace.”