Disaster is a laboratory for innovation. During a war, or in the aftermath of an earthquake, whole societies mobilize to answer the immediate challenge, while a cadre of researchers looks for a way to transform the crisis into advances that will improve lives, or save lives, in the future.
The global COVID-19 pandemic is such a challenge. For Zachary Pardos, an assistant professor at UC Berkeley’s Graduate School of Education and the School of Information, the crisis brings a persistent question: With tens of millions of students across the world forced to stay home from school, and shutdowns in some areas likely to continue in the fall, how can we assure that they get the best possible education?
Pardos is a specialist in adaptive learning technologies, studying the deep dynamics of student learning and marshaling big data to build user-friendly tools that are both powerful and subtle. He has worked closely with teachers and students at every level to integrate the technology into everyday curricula.
In an interview, he described how these emerging support systems engage students and evaluate their strengths and weaknesses, even when they’re not in the classroom. The systems are not an online course, but rather an online tutor, driven by artificial intelligence, that can assess a student’s strengths and weaknesses and deliver personalized individual instruction.
Such technologies are already in use in undergraduate study, including at Berkeley, and to a limited degree in U.S. high school classes. Today, however, educators are compelled to consider the most effective ways to teach students at home — and that means COVID-19 may open the door to new ideas and new technologies that will endure in the classroom after the disaster subsides.
And while the pandemic and economic disruptions are changing the landscape for future work, Pardos said, adaptive learning technologies have the power to help students pivot, on the fly, toward new careers.
[This interview was lightly edited for length and clarity.]
Zachary Pardos: It’s something different. Those technologies have elements of adaptive learning. Specifically, mechanisms for immediate feedback. Thanks to autograding, you can be given correctness feedback on problems, or even essays, in an online course. But adaptive learning involves more personalization on the part of the technology.
The key components of adaptive tutoring systems have tended to be a model that continually assesses what a student knows, a list of knowledge in the domain being learned and then hints and an adaptive sequencing of content based on what the student knows. An example of that on the Berkeley campus has been the ALEKS system. It’s used by incoming freshmen who aren’t yet ready for college-level math.
But they’re not ready in a variety of different ways. So, it’s not just one short summer course that could remedy that. And having human tutors continuously assess readiness and adapt instruction to each of the hundreds of incoming students is a monumental task that would quickly exceed the resources of a human tutor. But adaptive tutors have been shown to scale very well in this scenario.
Mastering Physics and MATHia, for geometry and algebra, are other examples of adaptive tutors. There are many more. Most major textbook publishers have purchased or developed tutoring systems like these, and there is a wide variety of adaptive learning technologies coming out of industry and academic labs, some of which share the same focus on assessment as the tutoring systems approach.
It has been a weakness of adaptive tutors that they tend to be in limited STEM domains. A challenge going forward is to broaden them. One of the bottlenecks is just how much subject matter expertise is needed to model a new domain. But big data approaches have been promising for overcoming that. A first milestone was achieved when we were able to automatically generate personalized help on the fly in an online course by using AI, which learn from past student interactions with the course.
It’s the immediate personalized help and prescribed practice that can take place. While the process of seeking out help can be beneficial, many students don’t know where to start and look to the course materials and the instructional staff for assistance. Adaptive learning technologies can provide some of this assistance. If a student isn’t up to a level of knowledge to be able to answer a question, instead of continuing to the next lesson, the system would adaptively extend the current lesson, giving help along the way in the form of hints and other activities until the student is prepared to advance.
For limited subjects, yes. The emergency move away from traditional classrooms has caused a reduction in instructor-student contact hours. This is taking place across K-12 and higher ed. A lack of contact hours could be partially compensated for with adaptive technology, where in those moments where students can’t have synchronous learning sessions (with teachers, in real time), they can be interacting with a technology that has the capacity to personalize instruction — a limited capacity, but more so than a video or textbook.
Definitely. Now that so many educators have had to communicate, learn and teach through the online medium, it can’t be ignored as an option to consider in the future, nor can the question be ignored of what tools could be brought to bear to improve the quality of learning in online and place-based environments.
This is an opportunity to reflect on the challenges experienced during the pandemic, challenges such as lack of engagement and lack of a sense of connection with students. How can appropriate application of adaptive technology make online learning experiences whole, from both teacher and student perspectives?
There’s both an access and an orientation issue here. With access, we’ll start to see data SIM (subscriber information module) cards and devices treated like school buses — delivery vehicles expected to be provided for students to get them to the now virtual classroom.
Even when students have access, there is evidence that the way in which they orient toward online learning can lead to an achievement divide. A colleague at Arizona State University was curious how students were navigating materials in his online course and if students who failed the course navigated in a different fashion from those who passed.
Our research into the data from the course showed that going to the quizzes first and then looking for answers in the preparatory material was a dominant pattern among students who failed. Following the prescribed syllabus path was dominant among students who passed. The instructor made a couple of modifications for the next class, offering a small amount of extra credit for those accessing the preparatory materials first and sending an email to those who didn’t, letting them know how important it was to the success of previous students. He saw an increase in grades after making those changes.
The takeaway for the immediate remote instruction situation is that some students will not naturally have a disciplined orientation to online learning. If not through live videoconference sessions and taking attendance, how are teachers maintaining structure and keeping students on track? Incentives and evidence-based personalized communications are options. The rigid personalized sequencing of adaptive learning technologies might be another.
As biological, cognitive beings, we haven’t changed very much. But technology is changing every facet of our life, and I think that is also now happening in education, where teachers are working alongside technology. There is also a hint of the familiar here, since many adaptive learning systems have been inspired by one-on-one tutoring.
We will see technology become easier to integrate with what teachers are trying to accomplish. Machine learning technology, especially natural-language processing, will provide for Socratic pedagogical approaches and better coordinated peer-to-peer learning opportunities.
Adaptive learning will be used in broader contexts. Due to the changing economic landscape, many students may decide to pivot from their intended careers and degrees. They will have prior knowledge gained from the degrees they are pivoting away from and would be seeking personalized curricula that leverage what they have learned to coherently shift toward what they want to now learn.
This is a challenging personalization task, but a scenario similar to those in which adaptive tutors have been shown to excel, only applied at a higher, across-course level. My lab has been developing and piloting an adaptive technology aspiring to this goal here at Cal.
One pitfall is thinking that technology can do everything. People need to learn academic subjects, but they also need to learn to be people. They need to learn compassion, generosity, how to work together, how to share responsibility and credit, and how to maintain relationships, which is certainly a lifelong learning topic. How to become a good citizen who contributes to the conversation on what society should value. You’re not going to have an adaptive learning technology to teach that.
SOURCE: Edward Lempinen
VIA: Berkeley News