Innovator Interview: Janet Kolodner

Janet KolodnerQ. You began in AI and pioneered case-based reasoning. What is that?

A. Rule-based expert systems were becoming more capable in 1980, but they broke unless you told them exactly what to do, and they didn’t get better with time and experience. My advisor, Roger Schank, and I thought we could realize expertise better by building systems that could have experiences and remember them, reason about those experiences to create new knowledge, and organize those experiences so the new knowledge could be refined with new experiences.

After I finished my Ph.D., I went to Georgia Tech where psychiatrist Rob Kolodner (no relation!) was interested in working with an AI person on software that could do psychiatric reasoning. One of the hardest things psychiatrists had to do was to distinguish depression that has some psychotic components from psychosis that has some depressive components. We examined the way he did that and I saw the ways he used his previous experiences. Organizing cases in memory and finding them later when they could help him figure something out seemed to be the crux of what he was doing. My version of case-based reasoning came from that. I was calling it memory-based reasoning. It was Chris Riesbeck who came up with the name case-based reasoning.

Q. So case-based reasoning (CBR) led to your interest in education?

A. The first CBR system my students built could reason about the Arab-Israeli conflict. It remembered two kids fighting over a candy bar and what the parents did (one child split the candy bar; the other chose his half). It suggested a solution based on that, but we gave the program feedback, telling it that neither side wanted half the region, but rather Israel wanted security rights and Egypt was seeking integrity. It eventually retrieved the Panama Canal solution, where the U.S. had security rights in the area and Panama had land rights, and it suggested that solution to the Sinai. We built several other CBR systems—to mediate labor/management disputes, to plan meals, and more. We learned a tremendous amount about tradeoffs between the reasoning done by the processor at the time of having an experience vs. when it’s trying to remember later, and about indexing strategies for memory (now called “tagging”). I learned a lot through those experiences about learning from experience. Our computer programs were pretty proficient and they seemed to mimic what people did.

Throughout much of this time, my kids were telling me that elementary and middle school science was boring. At the same time, I was running the EduTech Institute at Georgia Tech with the idea that what we know about how people learn should guide what we do with technology. The advisory board consisted of people from around campus who were interested in educating better, including K-12. With its engineering and architecture colleges, design education was the sweet spot at Georgia Tech. Designing things gets people excited about learning. And design requires iterating towards better solutions, matching CBR’s focus.

We thought we would design and build software for helping kids reflect on and learn from their experiences, and we would do a little bit of curriculum development to go along. We didn’t know how hard curriculum development is, and we didn’t know classrooms weren’t ready for software. But we kept on. We had to get the pedagogy right first. We began by working closely with engineers, architects, and middle school teachers, plus Howard Barrows and Paul Feltovich, experts in Problem-Based Learning (PBL). Together, we figured out ways of using the tenets of PBL to guide learning from design in middle school classrooms. Learning by Design, the Georgia Tech approach to middle school science, came from that. It turned out to be a beautiful amalgam of what CBR says about reasoning processes involved in learning from experience and what PBL says about how to manage classroom facilitation and sequencing.

We were so successful that we wanted to publish our curriculum units, but we had two problems. One was that other project-based learning researchers weren’t recognizing us as contributing to their endeavor, so we were not getting airtime at conferences. The other was that we had developed about a semester of Earth science and a semester of physical science, but we did not have a full year of any discipline. We couldn’t get our curriculum units published; publishers wanted a full year.

So I did two things. First, I began specifically calling our approach a project-based approach and pointing out how it goes beyond the conception of project-based that was in fashion at the time. Second, I asked my colleagues to join me in an effort to make a full three-year middle school science curriculum. Our Georgia Tech group teamed up with the project-based science teams at the University of Michigan and Northwestern, and together we developed Project-Based Inquiry Science™ (PBIS), which is used by about 10,000 kids per year around the country. I learned three things from this experience: 1) making a full curriculum that all hangs together coherently is really hard and requires a lot of different kinds of expertise, 2) a good project-based science curriculum can help kids develop the disposition of seeing the world as scientists and seeing themselves as people who can do science, and 3) I have a real strength in envisioning integrations of component parts and making those integrations a reality.

But we never would have gotten as far as we did without two fairy godfathers—Joe Psotka from the Army Research Institute and Gerhard Salinger from the National Science Foundation. They gave my Georgia Tech team the confidence to submit proposals to DARPA and NSF and gave us a tremendous amount of advice.

Q. Why is project-based learning important?

A. Learning by Design (LBD) and then Project-Based Inquiry Science (PBIS) turned out to be project-based inquiry approaches to science learning that really worked with respect to engaging the students and keeping them excited over long periods of time, helping them have experiences they could learn science from, and fostering learning. I tell people that keeping kids engaged long enough to actually learn is between 50% and 80% of fostering learning, and LBD and PBIS are both quite successful at that; they take an approach that helps learners anticipate what comes next so they feel they have a lot of agency. And then we also had all kinds of reflection opportunities built in, so the rest of that 20% to 50% of what’s necessary to foster learning is present in LBD and PBIS as well. In addition, the kinds of challenges we ask kids to address broaden their perspectives, have goals beyond their day-to-day goals, understand better what goes on in the world around them and what it takes to address real-world issues, and what they might want to be and do someday.

So it turns out that project-driven education (that’s what I like to call it nowadays) holds opportunities for both fostering learning of whatever content and skills the powers that be are telling us kids need to learn and also for helping kids develop communication, collaboration, and problem-solving skills, see themselves as problem solvers or designers or disciplinary thinkers who can make things happen in the world, and have the confidence to take on new challenges. Education policy makers don’t usually talk about our need as a democracy and if the U.S. is to remain a world leader to be helping kids develop these capabilities, but I think most of us would agree that these are the qualities we want kids to have when they graduate from high school.

Q. You founded the Journal of Learning Sciences. How did that get started?

A. I started the journal in ’89. We didn’t know what it would become, we only knew there were a lot of people in a lot of different disciplines who wanted to learn about learning. I talked to people in computer science, education, educational technology, science education, math education, cognitive science, and so on and I had a lot of help figuring out what the journal should be. Jim Greeno, Allan Collins, and Bob Glaser were particularly helpful in the beginning. I learned almost everything I know about education from my experience as a journal editor. Reading the submitted articles helped me learn, but even more helpful was reading the incredibly thorough and insightful reviews.

Q. Tell us about your time at the National Science Foundation with the Cyberlearning program.

A. I had a great time at NSF. I appreciated the influence I could have on policy and on the research community around me. I helped found and then run the Cyberlearning program at NSF, but I hope I left a legacy that is bigger than that. I had several beliefs that made their way into the Cyberlearning solicitations and into discussions with others at NSF. First, you can’t be successful at making a difference if what you’re funding is just a little out from the current time. You have to give researchers a chance to design the future they are envisioning. Second, the best insights come from teams with different expertise and a variety of perspectives. Third, the best designs for learning technologies are informed by what is known about how people learn. Research funding for learning technologies should not simply be about designing new products, but should combine designing and building a product with learning from the design and development. So I made sure the Cyberlearning program required all of these things—thinking farther out into the future than a few years, working in interdisciplinary teams, and focusing not simply on products and their evaluations, but also on what can be learned from them about new “genres” of learning technologies, how to design and use them, and on what can be learned about how people learn. My legacy will lie in the way these core tenets are taken up.

Q. What are the next challenges you want to pursue in education and technology?

A. I’ve learned how important integration across projects and approaches is to having impact. I also realized that technology can play two huge game-changing roles. The first is that technology allows learners to have collaborative, immersive experiences with scenarios, phenomena, and processes that are not possible otherwise. Learning about energy’s role in volcanoes is much different, for example, when you read about volcanoes exploding or watch a volcano explode than when you can ride on magma as it is being spewed out of the volcano. Second, the computer allows new kinds of expression that also allow self-evaluation. For example, with paper and pencil, you can write in words the way a process works and draw pictures and schematics. With the computer, on the other hand, you can express the way processes work by programming them, and you get a chance to check your understanding by comparing what you see to what you were thinking you would see. The future of the computer in education is in helping people experience phenomena, processes, and situations they could not otherwise experience and in fostering the kinds of expression that help people refine their understanding.

I’d like to be chief architect on several integrative projects that show what’s possible when technology is used well in education. I’ll be asking researchers, teachers, and software developers around the world to help me imagine, and I’ll be counting on the Concord Consortium’s technological and pedagogical expertise to help make those integrations a reality. And, then, of course, I want to spawn products from those integrations (curriculum projects, technology products, and combined products), learn more about how to use technology well to foster learning, and learn more about how people learn.

Ultimately, I’d like to be able to use those projects and their implementations as inspiration in helping to change the public perception of what education could be. I’ve figured out that a big reason learning sciences and technologies has not had much impact is that most of the public cannot imagine an education system that’s a lot different from what they’ve experienced. I think we each have zones of proximal imagination; we can only imagine a certain distance from what we have experienced. Over the next 10 to 15 years, I want to do something about that!