UW–Madison researchers propose an inclusive artificial intelligence system, incorporating voices and perspectives from across Wisconsin.
When most people use artificial intelligence (AI) tools today, they’re interacting with systems trained far from home, built on data pulled from across the internet. Those datasets tend to reflect global conversations, content from platforms such as Wikipedia and Reddit, but not always the lived experiences of smaller communities. And even if they are included, often the inclusion is sparse.
That can mean a cattle farmer in Iowa County, a small business owner in Green Bay, or a member of the Ho-Chunk Nation might not see their lived experiences reflected in the systems that increasingly shape how people search for information, make decisions, and interact with technology.
Information School assistant professors Jacob Thebault-Spieker and Corey B. Jackson are proposing a different approach, which they describe as “AI by Wisconsin, for Wisconsin” — a statewide public AI infrastructure shaped by the people who live and work here, with data gathered locally and transparently. That perspective reflects a broader mission of the College of Computing and Artificial Intelligence, launching in July, to ensure AI is shaped by the people and communities it serves.
Building from local knowledge
Because AI systems are shaped by the data they learn from, creating ways to capture local perspectives becomes essential. Their proposal outlines approaches to gathering and organizing data that reflect Wisconsin’s geographic, cultural, and economic diversity, such as partnerships with communities across the state, whether urban or rural, as well as collaboration with tribal nations.
These efforts would move beyond relying on existing internet datasets by creating new, locally generated sources of information and context, helping ensure that the experiences of Wisconsinites are meaningfully represented in how AI systems are trained and operate. “Our research shows that the data used to train AI systems does not equitably represent Wisconsin today. As a result, current AI tools often fall short in capturing local knowledge and perspectives, especially in rural communities,” explains Thebault-Spieker.
That gap in representation is closely tied to how AI systems are built and understood. Often people don’t have the opportunity to voice how (and what) data is used. “People need to be engaged in the process of developing these technologies,” continues Jackson. “Otherwise, they’re being shaped by something they had no part in creating.” Today, many widely available AI tools operate as a kind of “black box,” where users have little visibility into what data was used, whose perspectives are reflected, or how decisions are made. Opening up that process, both in how data is gathered and how systems are governed, is a central aim of the proposal. Town halls, UW–Madison Extension partnerships, and digital platforms could give residents ways to weigh in on how AI tools are developed and used. This greater transparency could help people better understand how decisions are made.
A foundation for local innovation
Alongside community-driven efforts to ensure data representation, Thebault-Spieker and Jackson envision building the technical backbone to support it. That includes developing a state-of-the-art foundation model (trained on vast amounts of data that is adaptable across tasks) and creating secure infrastructure that both protects privacy and respects data sovereignty for tribal nations and other communities across the state.
Over time, that foundation could support tools tailored to Wisconsin’s needs. Contextually relevant, localized AI could help farmers by incorporating county-level soil characteristics, recent rainfall patterns, and region-specific crop risks rather than relying on national averages. And local governments and schools could explore applications that reflect the communities they serve. Rather than using systems designed elsewhere with data from unknown sources, the idea is to create something that reflects the social, environmental, and regulatory context of Wisconsin.
Centering humans, not technology
The proposal also points to the economic potential of building AI infrastructure locally. Work tied to data labeling, model training, and application development could create new job opportunities across the state. But the broader vision goes beyond jobs or technology. Jackson and Thebault-Spieker see this as a chance to rethink how communities participate in shaping the systems that increasingly shape everyday life.
At its heart, the proposal treats AI less like a product and more like public infrastructure — something that should be built with input from the people it affects. “It’s really about the role of technology and its impact on humans,” Jackson says. “Instead of centering technology, we center humans — and technology is an extension of human goals.”
The team is seeking funding to build proof-of-concept infrastructure and demonstrate how a public, locally grounded approach to AI could work in practice. With university support and potential partnerships across the state, they hope to lay the groundwork for long-term investment. Learn more about the impact of UW–Madison’s federally funded research and how you can help protect it.