Sustainability Hub takes root at MSU Denver
Faculty and students lead National Science Foundation-funded project to democratize access to environmental data for Coloradans.
This story appears in the fall 2024 issue of RED Magazine.
In his former career as a software architect and more recently as an assistant professor of Computer Sciences at Metropolitan State University of Denver, Daniel Pittman, Ph.D., seldom gave much thought to sustainability.
These days, he finds himself leading a three-year, $1.2 million National Science Foundation research grant to develop MSU Denver’s Sustainability Hub — a website powered by machine learning that will support Coloradans seeking an array of economic, ecological and social data. Much of the software development for the project, which launched last October, is being done by MSU Denver students.
“It’s not that I had a bad opinion about sustainability,” he said. “I just never understood how I could apply my skills as a computer scientist to help it, until this opportunity.”
Pittman said the hub will serve community members, policymakers and researchers. The website will make it easier for people to find the great research already being done in Colorado around issues such as air quality or water quality. The key, Pittman emphasized, is making the site approachable.
That’s where machine learning comes in. Pittman envisions a ChatGPT-like interface that allows users to pose queries using natural language and helps them refine their ideas. A user could say, “I live in Denver, and I want to learn about air quality,” and the system would provide a visualization of the data, as well as a natural-language description of what it means.
“That’s the part I’m really excited about,” he said. “I want to give people not only the visualization but natural language: ‘You are seeing air quality and water quality in Denver overlaid. These are the points that we found. This is what’s interesting about the data. Is this what you’re looking for?’ If the user answers, ‘No, I didn’t quite mean that,’ the system will help them further refine the results.”
Student-centered learning
Pittman, who worked at companies such as Lockheed Martin while pursuing his doctorate, joined the MSU Denver faculty full time in fall 2022. He is the administrator of the National Science Foundation grant, which was developed to help Minority Serving Institutions such as MSU Denver expand their research footprint. He credits his colleague Elicia Ratajczyk, Ph.D., a research scientist at Colorado State University, for suggesting that he develop the grant proposal.
The project includes partners at the University of Northern Colorado, CSU Fort Collins, CSU Pueblo and the University of Denver, Pittman said. At MSU Denver, the funding has allowed for the hiring of Project Manager Alyssa Williams, a recent graduate from the University’s Computer Sciences program, and a cadre of students who have taken on various aspects of the website design as full-time interns.
RELATED: AI goes to college
Luke Farchione, a senior majoring in Computer Sciences, is part of the team working on the Sustainability Hub’s machine-learning and natural-language processing components. He contributes to the training of the large language models, or LLMs, that are crucial for developing a chatbot that can “converse” with users. He also assists in cataloging the extensive data sets used in training these models, ensuring accurate and meaningful user interactions.
He had taken several classes with Pittman, who told him about the research opportunity. “I was head over heels for it,” Farchione recalled. “Programmers often find themselves as tiny cogs in a huge corporate machinery. The special thing about the Sustainability Hub is that it’s going to be for the community — the people of Colorado.”
Fourth-year student Jacqueline Hernandez leads the user interface/user experience (UI/UX) team, which has prototyped a design for an informational website to keep the community up to date on the project. Her work in UX design is crucial in ensuring that the platform is intuitive and engaging for users.
“Moving forward, we’ll transition the informational website into the actual application,” she said. The team anticipates that the application will go live in 2025.
Hernandez took a mobile application development class with Pittman and enjoyed his teaching and mentorship. She became interested in working on the project with him and hoped to gain some real-world experience before graduating. As part of the team, she has gotten to work with new programming tools, probe the “black box” of machine learning and learn about data harvesting.
“The learning experience for me has been unprecedented,” Hernandez said. “This has really taken my computer science education to another level.”
Hernandez also has a personal stake in the project’s mission. “Sustainability is something I’m passionate about, growing up in Colorado,” she said. “It has been an important part of my life, going to the mountains and enjoying the outdoors. I had an interest in being involved in a project that’s aligned with my values.”
RELATED VIDEO: Closed-loop compost system paves the way for a sustainable future
Challenges and opportunities
Pittman acknowledges that the ambitious scope of the Sustainability Hub and its reliance on an artificial intelligence tool will require close supervision to ensure the accuracy of its results.
LLMs such as ChatGPT sometimes “hallucinate” and provide wildly inaccurate information. Given that the data provided through the Sustainability Hub might wind up being used by parties in a lawsuit, for example, one concern is whether the model might provide biased output.
“Without proper guardrails, it absolutely will take a side,” Pittman said. “It would be foolish to assume that the large language model we use is not implicitly going to have biases. The real question is: How do we fix it and protect our users? A lot of that comes down to how we prompt the LLM.”
RELATED: How students should — and shouldn’t — use artificial intelligence
But chatbot developers know they can’t fully control the output of their LLMs. The model’s machine-learning capability means it develops its own “understanding” of the data. One way to mitigate this challenge is to ensure that the sources of the data are always shown, and another is to prevent the opportunity for data to be hallucinated when it is displayed, Pittman said. He added that the data will always come straight from the source and only its description will be generated by the LLM.
There are certainly challenges ahead but also opportunities.
Pittman said the system will allow users to build communities of interest on the website around aspects of sustainability that they are passionate about.
“The key phrase that we like to use is ‘We are trying to democratize access to sustainability in Colorado,’” Pittman said. “So anyone who wants to learn can find information on our website. That’s really what we want to build.”