While in graduate school, Dr. James Lester became very interested in applications of artificial intelligence (AI) in education, and in 1994 when he joined North Carolina State University, this was the area he chose to focus on. “There is something inherently interesting about using AI for education,” shares Lester, Goodnight Distinguished University Professor in Artificial Intelligence and Machine Learning, Director of the Center for Educational Informatics, and Director of the National Science Foundation (NSF) AI Institute for Engaged Learning, North Carolina State University. “It’s exciting to be alive during a time where many long-term visions for AI-enabled education can now be realized, and the models are becoming increasingly powerful. These technologies make it possible to re-envision how learning works, how teaching works, and how we think and support both teaching and learning.”
Interested in discovering ways we can transform education with artificial intelligence, Lester’s research focuses on AI-driven narrative-centered learning environments and virtual agents for learning. “We’re exploring ways we can use story-based learning experiences that are driven by AI to create incredibly compelling learning episodes,” shares Lester. “Rather than thinking about a narrative as text, we look at how we can think of a narrative as an immersive unfolding experience. AI is used as an experience manager to dynamically craft that narrative. For example, imagine a middle school science narrative where a student is interacting with a virtual scientist on an expedition. Rather than having a predefined specification of a narrative, that narrative could be planned and replanned, customized and personalized in real time to satisfy different goals at the same time.”
“On the one hand, this helps create effective learning experiences and inquiry-based learning for science,” continues Lester. “We look at how we can introduce students to concepts in the context of active problem-solving while keeping them engaged. Our narrative-centered research explores how to optimize effective and engaging learning simultaneously. We’re also interested in multimodal learning analytics and how to model students’ competencies. With the emergence of powerful models for analyzing multiple data streams, we can develop new insight into how people learn and how computational models can be used to derive those learning experiences and support teachers in the classroom. In addition, we are looking at how AI can be used to analyze student discourse during collaborative learning, specifically, understanding in real time what is going on when students are coordinating their problem-solving activities. We’re interested in understanding how we can use these insights to adaptively scaffold what students are doing in a learning environment.”
Developing AI-Augmented Learning
As director of the National Science Foundation AI Institute for Engaged Learning (ENGAGE AI Institute), Lester leads research on narrative-centered learning technologies and multimodal learning analytics to create engaging collaborative story-based learning experiences. “The NSF is creating centers of excellence around AI in a large variety of disciplines, from physics and meteorology to agriculture and education,” explains Lester. “The ENGAGE AI Institute is driven by a learner-centered vision of AI-augmented learning and we’re looking to develop new technologies that can move the needle in the classroom and create engaging and motivating learning experiences.”
The ENGAGE AI Institute is headquartered at North Carolina State University and has four primary partner organizations, University of North Carolina at Chapel Hill, Indiana University, Vanderbilt University, and the non-profit organization, Digital Promise. “The Institute is a multi-institutional collaboration and brings together people who are experts in machine learning and AI, education, computer vision and natural language processing,” says Lester. “We also have education researchers, learning scientists, educators, and educational technologists. Our research focus includes narrative-centered learning, conversational agents, and multimodal learning analytics. As you can imagine, there are a multitude of complex technical questions in these areas. To find the answers, we’re pursuing a broad range of complementary approaches, with a particular emphasis on those leveraging unparalleled advances in generative AI.”
Combining multimodal learning analytics with foundational AI, the Institute is also looking at the challenging problem of analyzing classroom video. “In comparison to major tech companies, educational settings produce much smaller data sets,” says Lester. “There’s an interesting question of how we can customize standard models that were developed for other purposes and have them work in a data-constrained setting. For example, looking at ways to query a system after a long video about a particular event that took place and see if the system can pick out a segment of that video to answer the question. This could fundamentally change how we conduct education research and will enable us to re-envision qualitative research and education.”
“In addition to applying this approach to AI-enabled learning technologies, we must teach students from K-12 to higher education about ethical issues in AI. Not only will the people who are designing policies and engineering the next generation of machine learning systems need to follow AI ethical guidelines, but our future generations who will become citizens in the next thirty and forty years will need a foundational understanding of the ethical challenges of AI.”
— Vasant Honavar, Ph.D.
Huck Chair in Biomedical Data Sciences and Artificial Intelligence,
Founding Director, Center for Artificial Intelligence Foundations and Scientific Applications (CENSAI), and
the Center for Big Data Analytics and Discovery Informatics, Pennsylvania State University
Creating Scalable AI Technologies
Now three years into the program with over seventy people involved, the ENGAGE AI Institute has come a long way since the team was originally formed. “Getting an AI institute up and running takes forming teams that will work together, as well as getting software development rolling, including creating the tools that will be used to integrate different families of technologies,” explains Lester. “Then we started full scale studies of these technologies in classrooms to understand what is working and how students approach collaborative learning. We looked at general models and classroom specific models and how we could better engineer those practices to work. What is most exciting is being able to investigate how these kinds of technologies are changing learning. How can we help students learn more deeply? Are the learning opportunities effective and engaging? Most importantly, how can we create experiences that compel students to become skilled lifelong learners in the Age of AI?”
Also keeping the future in mind, the Institute is exploring ways to scale these technologies and expand their impact. “We have a goal of working with 25,000 students by the end of the project,” shares Lester. “The Institute is looking at ways to anticipate AI’s role in education and learning and determine a strategy for evolving in a very fast-moving era. This includes developing AI-enabled learning technologies and leveraging AI to improve learning outcomes. We also want to educate students about AI concepts, practices, and technologies so they can become equipped with the in-demand skills of the workforce. We want to help the next generation of students become comfortable with AI, as well as be inspired to become accomplished AI researchers and engineers who will be instrumental in moving the technology forward.”
“Edge believes research, experimentation, and innovation are integral to our progress as a nation, and AI is a vital part of the necessary changes, especially in education and research,” adds Dr. Forough Ghahramani, Assistant Vice President for Research, Innovation, and Sponsored Programs for Edge. “We will continue to raise awareness to research from thought leaders such as Dr. Lester, and available resources, in addition to providing educational programming and events specific to the important role of AI in education and research.”
Establishing Compute Infrastructure
Helping educators leverage AI and understand advanced technologies is an important initiative for many institutions of higher education. “Along with training teachers to be comfortable with using AI-enabled learning technologies, we must also help them be prepared to teach students about AI concepts, practices, and ethics,” says Lester. “Professional development programs must be provided as soon as possible to help get our teachers up to speed on AI. This is a time-sensitive requirement, and we need to develop ways that will ramp up knowledge quickly. While there are certainly challenges around having AI in the classroom, I think we need to shift our perspective toward looking at how this technology can be leveraged to improve education.”
“In looking at K-12 education, there are very heterogeneous systems from the state level down to individual districts,” continues Lester. “We do not have the luxury of imposing a single set of changes that need to be met by a particular deadline. Some districts will have what they need to embrace AI quickly, while others will need a bit more help and encouragement. Many parents are intimidated by the idea of having AI in the classroom, but our approach needs to be introducing AI with guardrails. AI-enabled education will absolutely transform universities, and it will be essential to determine how to best translate these changes across the K-12 community.”
Since AI-enabled learning technologies require very powerful models, a key part of integrating AI is establishing the compute infrastructure. “Many interesting questions arise around infrastructure, like what kinds of data will we use to train these models?” says Lester. “Who is paying for the training and the runtime version of these models? And what kinds of platforms will be used on? In the education marketplace, there will likely be an enormous procurement challenge because buying the right kind of software will require knowledge of both education and AI-enabled learning technologies. We’re going to have to have experts at the district level who can help make informed decisions about buying the right systems.”
“I believe we can architect a future for ourselves that will make a difference in how the next generation lives and will allow us to overcome challenges that have historically been difficult. This is a fantastic time to be a learner of any age, and we have an opportunity to steer technology in a direction that produces a vibrant society and allows us to learn and grow in exciting new ways.”
— Dr. James Lester
Goodnight Distinguished University Professor in Artificial Intelligence and Machine Learning;
Director of the Center for Educational Informatics;
Director of the National Science Foundation (NSF) AI Institute for Engaged Learning, North Carolina State University
Addressing Equity and Ethics in AI
To encourage the ethical development of AI, Fairness, Accountability, Transparency, and Ethics or FATE, must be considered. “Combining privacy with FATE when training the models will give you a good starting point for what it means to have ethical AI,” says Lester. “In addition to applying this approach to AI-enabled learning technologies, we must teach students from K-12 to higher education about ethical issues in AI. Not only will the people who are designing policies and engineering the next generation of machine learning systems need to follow AI ethical guidelines, but our future generations who will become citizens in the next thirty and forty years will need a foundational understanding of the ethical challenges of AI.”
To ensure equity of access to AI and advanced technologies regardless of the school system or institution, policies will need to be put in place. “AI is a data-driven technology,” explains Lester. “That data has to come from somewhere to train the models. The question is, where does that data come from and how can we ensure that these models we’re placing in classrooms will help each individual student? We cannot allow algorithmic bias or bias as a result of improper data. Having strong regulatory policies at the federal level relating to education is incredibly important. For example, we must formulate a policy that sets a clear bar for whether a piece of software should be put in a classroom. If a particular model is making instructional recommendations for a student, we must be fully confident that those models are making recommendations that are reasonable. And the majority of advances we’re seeing that are driven by generative AI and large language models are based on deep learning models, which are inherently not explainable.”
“One of the beautiful things about AI is it is naturally scalable,” continues Lester. “So in principle, if you have a model that follows ethical guidelines and a system that promotes the most positive learning experiences for any individual student, the technologies are scalable. Developing a compute infrastructure that could broadly deliver AI technologies could make a game-changing difference in how we educate K-12 and college students throughout every community.”
The Future of AI
As more countries around the globe continue to explore AI and invest in research and development, Lester says solutions and applications will hopefully be shared on a very broad scale. “I think we can expect to see AI methods, frameworks, and models that were developed in a small number of countries be promoted for international adoption, particularly in the education space. This phased adoption will likely be different in each country, creating a patchwork of approaches to using AI in the classroom. This adoption rate will also depend on the infrastructure and hardware footprints that are required for delivering these technologies.”
Beyond education, AI technologies will offer many benefits that will revolutionize industries across the board. “One of the fundamental attributes of AI is that it’s quite good at personalizing experiences,” says Lester. “We see it online, in social media, and in entertainment, for example. Along with creating AI models that are ethical, companies have the opportunity to use models with deep personalization that helps advance equity and diversity. AI has an economic aspect to it, as well as a strong political dimension that will play out in different ways. But the potential is impressive, it’s baked into the technology, and AI will help unlock possibilities for equity and create deeply engaging experiences.”
With AI transforming the future of human engagement, from students and teachers to customers and organizations, the theories often differ on what tomorrow will bring. “Thoughts about AI are often on a spectrum, varying from fear-based theories to a vision where the world is thriving,” says Lester. “The middle ground is that change will be incremental, whether it’s in the workplace, at school, in the healthcare sector, or at home. This view is that there will be modest changes that make important differences along the way. For those who have a positive outlook, they commonly picture designing and building AI technologies that improve the wellbeing of society, including advances to healthcare, education, and the economy. I believe we can architect a future for ourselves that will make a difference in how the next generation lives and will allow us to overcome challenges that have historically been difficult. This is a fantastic time to be a learner of any age, and we have an opportunity to steer technology in a direction that produces a vibrant society and allows us to learn and grow in exciting new ways.”
