Maria Langworthy, PhD, sees the promise of generative AI on student learning and well-being, while also being aware of potential perils.
“There are a lot of risks and a lot of opportunities with generative AI,” said Dr. Langworthy, a global education thought leader and cofounder and managing partner of CampusEvolve.
On Tuesday, Dr. Langworthy shared her thoughts on generative artificial intelligence with faculty, staff and students during UNMC’s Breakthrough Thinking Conference series, which is designed to challenge and inspire out-of-the-box thinking.
Dr. Langworthy also was joined briefly online by UNMC’s Dele Davies, MD, Emily Glenn and Rachel Lookadoo, JD, each of whom reflected on her comments. Glenn and Lookadoo lead a UNMC AI task force charged with developing a framework for UNMC’s AI guided future.
While the pandemic accelerated the use of technology and forced changes in education, Dr. Langworthy said Chat GPT and Open AI sparked the technical acceleration of these opportunities, which is then driving faster adoption in education. “It’s up to all of us to research and examine how we take advantage and how we mitigate the risks,” she said.
Watch the Breakthrough Thinking presentation with Maria Langworthy online.
Dr. Langworthy explained how “vanilla” GPT (generative pretrained transformer) models allow for general prompts and responses, while generative AI allows for localized content to be pulled into the solution. Highlighting the latter, she demonstrated online how universities can provide university-specific content for such questions as “how do I find career planning help?”
“If you want to get results, you have to clearly define your use case,” said Dr. Langworthy, who, prior to founding CampusEvolve, served as director of worldwide education research at Microsoft, where she led the open education analytics community. She also served as the strategic data officer for education at the Bill and Melinda Gates Foundation.
In education, one use case might be to support advisers with a one-stop advising generative AI, she said, which would allow for personalized student answers across all student services. As a result, students could ask more questions more frequently, while boosting their confidence, motivation and sense of belonging. In turn, advisers would have greater capacity to focus on building personal relationships with students, she said.
Other use cases might relate to admissions and registration, access and academic success. “Generative AI is of most use to advisers and students when it’s grounded in your own data and content,” she said, and includes testing and training with one’s staff. Guardrails can be specified for specific prompts and responses, she said, that anonymize student information, mitigate risks and are monitored by humans.
There are benefits and distinctions to machine intelligence, just as there are with human intelligence, Dr. Langworthy said, highlighting four key points:
- Humans learn through experience, observations and interaction with their environment. Machine learning, a subset of artificial intelligence, requires machines to learn from data.
- Humans can apply knowledge from one domain to another, understand nuances and exceptions and generalize from a few examples, while machine learning often requires large amounts of data to learn effectively and may struggle with generalizations.
- Human learning is inherently creative; machine learning lacks genuine creativity.
- Humans excel at understanding context, which is essential for complex decision-making; machines typically lack such understanding of context beyond what is explicitly programmed or inferred from data.
“I’m at the early stages of understanding how all of this can be used, as are all of us,” she said, emphasizing that greater success will come if “we can have clearly defined use cases carried out in an intentional manner.”