In this post, Dr. Quinn Burke, Director of Computational Thinking Research at Digital Promise, talks with Dr. David Touretzky, Professor in the Computer Science department and the Center for the Neural Basis of Cognition at Carnegie Mellon University. Throughout their conversation, they discuss what education systems are currently in place within K-12 schools and the importance of teaching students about the disruptive technologies that are already present in their daily lives.
- Recent publications on computing standards provide minimal guidance on AI education.
- AI is living up to its name as “disruptive technology” and students need to learn about it.
- Regarding K-12 schools, personnel is more often an obstacle in AI education than finances.
Quinn Burke: Hi, this is Quinn Burke, and I am here with Dave Touretzky of Carnegie Mellon University (CMU). Let us start with introductions. Can you tell me a bit about who you are, including your position at Carnegie Mellon, your background, profession, and education?
Dave Touretzky: I am a Research Professor at the Computer Science Department at Carnegie Mellon University (CMU) in Pittsburgh. I am also the founder and chair of AI4K12.org, a platform that is developing national guidelines for teaching AI in K-12 schools.
QB: Dave, in your role at CMU, you have taught undergraduate and graduate-level students. I would like to hear more about that as well as more about AI with K-12, and more generally about your personal experiences that brought you into this field of study.
DT: Well, I got into the AI4K12.org project sort of accidentally. I had been developing an intelligent robot programming framework for children. I wanted to put real AI into the hands of kids. At the time, we were using the Cozmo Robot. Which is, by the way, an amazing robot made by the company called Anki. I started to think that “I have this thing that could be used to teach AI to K12 students, but would the teachers know how to use it and would they be interested in using it?” So I decided to take a look at the CSTA computing standards. The Computer Science Teachers Association publishes national standards for teaching computing in grades K-12. Some states have adopted those as official state guidelines. Other states have not gone quite that far, but they usually at least acknowledged the CSTA computing standards.
I looked at the most recent set of standards from 2017, and there were two sentences about AI in the entire document. Both sentences were from the 11th and 12th grade band. There was nothing mentioned before these two grade levels. So I realized there was a bigger problem here. Kids are not being taught about AI, though this represents a national priority, not just for the U.S., but for other countries as well. I approached the Association for the Advancement of Artificial Intelligence and CSTA and proposed that we do something to help teachers learn about AI and how to teach it. They both agreed this was important, so we went to the National Science Foundation and got funding, and that is how AI4K12.org was born.
AI Education in a K-12 Environment
QB: Schools have been criticized for being slow to integrate new learning technologies. Meanwhile, we are regularly reading and seeing that AI-based applications are growing at exponential rates. Has there been–in your experience (and your team’s experience)–a rift between K-12’s willingness to adopt AI in classrooms?
DT: Well, we need to distinguish between AI technology for education and educating students about AI. The latter is what I do. AI technology for education–things like intelligent tutoring systems– yes, it is going to take a while for that technology to be widely adopted. But that is not my concern. My concern is teachers need to know enough about AI, so they can help their students understand it because they are living in a world that is permeated by AI technologies now. We know there are major social changes on the horizon, with things like automation of jobs, or AI technologies changing our thoughts about things like privacy, or even the believability of images, with things like deep fakes. AI is finally living up to its promise as a so-called “disruptive technology”, and we need to teach students about it.
But, in short, no, we have not really seen any negative reaction from the schools. They all agree that it is important to teach their students about AI, but we are still feeling the effects of the pandemic. Schools are feeling teacher shortages and resource shortages, so it is not that easy for them to implement new curricular offerings, but we are working with the schools. One thing that has helped is that the pandemic accelerated the adoption of computing technology, so a lot of kids have personal laptops or Chromebooks now that maybe would not have gotten those if it had not been for the pandemic. That is to our benefit because we want kids to be able to run AI demos if they have a computer, and that is the first step.
QB: Based on your look at the landscape and collection of resources and trying to align them with the best principles and trying to put them into logical progressions, especially on the K-8 levels, are you finding that high schools are increasingly having some sort of “landing pad” coursework in terms of an introductory AI offering? I know I certainly see a lot more with CS offerings on the high school level, but do you expect we will see more AI as a standalone course or as an extension of CS coursework?
DT: Yes. I think that is a reasonable expectation. There are a few high schools that are doing things like machine learning electives in high school. But another thing that is happening is Career Technical Education (CTE ). There are already CTE pathways in computing, IT, and computer security. We are starting to see CTE pathways in things like data science, which is not quite AI, but is certainly AI-adjacent. There are CTE pathways in data science and machine learning, and in some cases in AI. All of these courses are starting to sprout up. I think that is a likely way that we see AI education develop in high schools.
Minimizing Obstacles to Equity
QB: What are your thoughts about this persistent digital divide–what some refer to as the “Matthew Effect” in terms of the individual that already has some will only get more, while those with less will only see further deprivation? How does this play with AI?
DT: So AI education is following the path of computing education. The nation figured out that all schools should be offering computing education, but we have not actually delivered on that promise yet. No state has universal computing education. Some of them are halfway there. Some of them are doing a little better. But no state has met the goal yet of comprehensive K-12 computing education. That is something that I think needs to be the highest priority, but then AI education follows from that, right? Now we can say, well, you ought to be including AI education as part of that, but that is gonna require substantial teacher-professional development. So that is one of the key obstacles.
The other obstacle is that you simply need access to a computer. But the computing education drive is going to eventually solve this, right? Kids will eventually have Chromebooks, and you do not really need more than a Chromebook to do AI education–until, maybe, advanced high school courses, where they are doing some serious machine learning. Before then, all the demos that I have developed, in my work, and the demos that are featured in our AI4K12 resource directory run in the browser. It is not that the schools are financially limited in terms of teaching AI. They are largely personnel limited. Their teachers need to get the training to feel comfortable teaching AI, and then they need to make room in the curriculum for it.
What Schools Have Access to Right Now
QB: Getting teachers prepared and getting them more comfortable is going to be a challenge. But what else do you see as a big challenge or setback around AI education and maybe computing more broadly?
DT: Well, so I have a rather idiosyncratic view about this. I will probably give you a different answer than anybody else would, but for me, I think the greatest tragedy for AI education in K-12 was the loss of the Cozmo Robot. Cozmo was a little robot made by Anki. It was incredibly powerful. It had real AI algorithms built in, including computer vision. And it was very cheap. It was $179 at Amazon. I use it in my university courses to this day, but you could also use it with young children. There were programming tools available, one made by me, one made by Anki that was based on Scratch. Kids as young as third grade could program robots to actually interact with the world using computer vision and real AI algorithms. Sadly Anki went bankrupt in 2019.
There was an attempt to try and sell the company’s assets to some large player, but no large player bit. You cannot buy Cozmo robots today. No one has stepped in to correct that. I do not know if we will ever see another Cozmo Robot, but I am fervently hoping that some of the other companies that make educational robots will step into this market and put robotics back in the hands of kids. I mean real AI powered robots, not these crude toys that lack computer vision and cannot move accurately. Those are not suitable for teaching AI.
QB: I appreciate that response. These tools matter. As we wrap up, there are certainly educational challenges out there with AI. But what do you feel especially hopeful about/ encouraged by?
DT: Well, I think AI has begun to permeate everyone’s consciousness now, so it is not just an obscure university subject anymore. Everybody knows what deepfakes are. Everybody talks to Siri and Alexa; everybody seems to be waiting to get their own self-driving car. In fact, there is a whole genre of YouTube videos of people sleeping in self-driving cars now. AI is really becoming a part of the culture, and that means that there will be many more opportunities for teachers to introduce AI into the curriculum. Because people are living with AI, they want to understand it, and they want their kids to understand it.
QB: Dave, is there anything else that you would add that you have been thinking about?
DT: Well, I should probably put in the plug for my software. Even though you cannot buy Cozmo Robots anymore, we have a cloud version of the software called Calypso. If you look up Calypso or go to calypso.software, you can access a cloud version that you can run for free and it runs in the browser.
And last, a brief shout-out to the AI4K12.org co-founders. Besides me, there was a gang of four that made this thing happen. There was me, along with Christina Gardner McCune at the University of Florida, Fred Martin, who is at the University of Massachusetts Lowell (and at the time he was the chair of the CSTA board), and Deborah Seehorn who recently retired but was a major player in leading the CSTA computing standards effort. She was fundamental to our understanding of how to build these guidelines for AI.
Together, we even received the EAAI/AAAI, Outstanding Educator of the Year Award last year (2022) to acknowledge our collaboration. I was glad that all four of us were included in that.