Interview with Josh Weiss

June 21, 2023

A portrait of Josh Weiss

“We must be intentional from the outset about what we want AI to help us accomplish.”

Interview with Josh Weiss, Director of Digital Learning Solutions at Stanford University Graduate School of Education, Stanford, California

Generative artificial intelligence (AI) is revolutionizing the world of education, but university administrators must figure out what they are trying to achieve before implementing the technology, says Josh Weiss, Director of Digital Learning Solutions at Stanford University Graduate School of Education. In an interview with IFC, Mr. Weiss explains how AI-generated teacher and student feedback can improve learner experiences and outcomes. He also talks about ways to evaluate the accuracy and originality of students’ work. Mr. Weiss recently participated in an IFC webinar to discuss the future of generative AI and ChatGPT in higher education. 

Which aspects of generative AI in higher education are you most excited about? Can you share an example?

I was a teacher 10 years ago, which feels almost like a century ago given the advancements in AI and what it means for inclusivity. I am excited about making learning more inclusive in ways that would have made a huge difference to my students.

First, access to personalized learning is no longer a matter of income but of being able to ask the right questions. And really, at the root of that is rethinking how we help students engage more deeply in their own learning experiences.

The second is that you don’t have to have a computer science degree to command a computer to do what you want, using human language instead of code.

Many great examples are coming from the Stanford Accelerator for Learning, where seed grants are funding projects that explore the intersection of technology and learning. For instance, a team is working on a project that helps blind and low-vision learners use AI to interpret data visualizations. Another initiative is using image generation as an assistive technology for neurodiverse learners in Bangladesh who otherwise would not have been able to create these types of images independently.

How can ChatGPT and other generative AI improve student experiences and outcomes?

The first thing the technology can do is to provide better feedback. Feedback for instructors might look like reviewing how verbally engaged students were with an AI coach and implementing changes that can improve student experiences and outcomes. This is particularly helpful if an instructor doesn’t have access to a colleague who can provide this feedback.

Generative AI will provide students with a timelier, personalized first round of feedback, which is especially valuable in cases where the student-to-teacher ratio makes it difficult to provide quick responses.

An example is a Stanford program that started during COVID called Code in Place. Two professors in the computer science department wanted to teach coding at scale. But offering feedback was challenging because the free course enrolled over ten thousand students. So, they started using AI to give real-time feedback to students on what was right and what they could do better inside their code. After that first round of feedback, a human volunteer stepped in to lend support.

What steps must administrators take to make the most out of this technology?

Higher education non-profit EDUCAUSE has a great framework to consider the role of AI systems in education. I have adapted their framework to include three of their four “Ds.” The first D stands for “dreaming.” For instance, AI can help us brainstorm ways to freshen up a syllabus with new authors and texts. We can expand our thinking and dream.

Generative AI is also really good for “design,” and by that, I mean building out content, like drafting course materials or ideating a lesson plan — comparable to the support a more junior teaching assistant might provide.

The third category is “drudgery,” tasks like synthesizing reports, summarizing meeting notes, and things an intern could accomplish. This AI tool might accomplish the majority of a task, requiring a human to polish up the remaining part. Offloading some of the “drudgery” frees up time for higher-value tasks.

So, to answer your question about how to make the most out of the technology, we must be intentional from the outset about what we want AI to help us accomplish. AI is being used to help an institution build content, or AI is helping lighten the load on a team so they have fewer of the things they don’t want to do and more time to do the things they like doing more. Once an organization figures out what they are trying to achieve with generative AI, they can have much more productive conversations on how to make that happen.

The other thing is having a solid idea of where the technology is heading, not just knowing what is possible today. But AI is evolving so fast that even technologists find it challenging to keep up with the pace of change. Given this reality, everyone needs to be an intentional consumer of AI and AI literate.

Do you see any dangers with AI when it comes to higher education?

Bias and misinformation are certainly a concern. The internet has wrong and biased information put out there by humans, and AI like ChatGPT is trained on this text. So, the danger is educators and learners accept this content uncritically. That is bad when people are doing a Google search and also when one uses ChatGPT. So digital literacy has become an important skill for students and teachers alike. It’s not just about ensuring information from a website is accurate, it’s about making sure a statement that sounds very confident coming from ChatGPT is true. That can be challenging because, often, ChatGPT is more confident than competent.

The other big danger is devaluing the craft of teaching. Teaching is not just a protocol of question and answer. There are a lot of ineffable points of connection that, as a teacher, can be difficult to explain. Teachers think about how to stack learning strategies for students at the right time so that they have breakthroughs in understanding. Teachers do far more than just transfer information.

What can teachers do to evaluate the quality, accuracy, and originality of students’ work?

First, I would take caution with AI-detecting software. We know that false positives remain a problem with AI detectors, at least for now. Research from Stanford indicates that AI detectors are also biased against non-native English speakers.

It will become more important for teachers to require students to show their work. This might include first and second drafts and reviewing various project iterations. Showing your work means assessing the process more than the product and not waiting for an entire semester to go by before evaluating a student on learning. It’s more important to see the progression of knowledge along the way.

How can this technology prepare students for the future of work? Can you give an example?

Students are practicing using generative AI on their own time and not just for school-related tasks. For instance, they use ChatGPT as an executive assistant to draft low-stakes emails. AI-assisted productivity boosts will only become more commonplace. They are being integrated into Microsoft and Google tools and having prior knowledge of how these solutions work will set apart graduates looking for jobs.

The other way generative AI will prepare graduates for the workforce is by helping them cultivate soft skills employers value, like project management and collaboration. In a way, ChatGPT acts as a team of interns that can do small tasks. It’s an opportunity for young people to practice monitoring the execution of these tasks and adjusting as needed. Higher education aims to prepare students intellectually, but they also need practical skills that will be useful when entering the workforce.

How can generative AI help make higher education more accessible to students in low- and middle-income countries?

The primary way is by lowering the cost of personalized learning. The internet was an earlier example—if one had access to the internet, they had access to the collective sum of human knowledge to an extent. And with ChatGPT and AI bots, people now have access to a sort of research assistant or a tutor that calls upon the sum of all human understanding. Students in low- and middle-income settings can now get farther downstream into the learning material that might otherwise have been out of reach.

Non-experts now also have fewer barriers to entry. If someone is brand new to a field and doesn’t have access to a collaborator, it can feel intimidating to delve in and make sense of a difficult subject. There are now AI apps that can break down explanations of complex topics. For instance, a young person in a rural, lower-resourced setting interested in quantum physics can use generative AI to translate advanced academic texts into everyday language.

This interview has been edited for length and clarity.

Published in June 2023

During his seven years as a teacher, Josh Weiss worked in public and private schools in the United States and Brazil as a literature instructor. In his current position as Director of Digital Learning Solutions at Stanford Graduate School of Education, Weiss serves as the connective tissue between learning and technology. Josh partners with professors, researchers, and practitioners to strategize, design, develop, and iterate digital learning. His role also entails data-gathering and exploratory projects into emerging technologies like AI, XR, and blockchain. Weiss holds Bachelor’s degrees in Political Science and Spanish from Emory University, a Master’s degree in Education from the University of Alabama, and a Master’s degree in Digital Media Design from Harvard.

Please find this interview in Spanish here.

Follow Us