Stanford’s Human-Centered Artificial Intelligence initiative recently released its Global AI Power Rankings. I doubt anyone is surprised that the U.S., China, and the UK topped the overall rankings.
I’ve spent the last two years studying how the U.S. education system is implementing AI rollout and implementation. Two trips to China this year, including one that concluded in early December, gave me a better idea of how China is approaching this difficult task. During my most recent trip, which included multi-day visits to many of the best public schools in Beijing, I gained insight into how the role of the teacher must change, a topic I addressed in an earlier blog for Getting Smart.
In that blog I suggested we adopt the model of the Flipped Classroom: The teacher no longer needs to be the primary vehicle of content delivery. I realized after the visit to China that I didn’t dig deeply enough into the operational details of how a teacher can flip a classroom using a trio of high-leverage AI platforms.
While visiting Moonshot Academy in Beijing I was placed in the audience for two of the best project-based learning presentations I have ever seen. Both student groups relied on generative AI for multiple tasks, among them research, design, and editing.
I asked Executive Head of School Jason Wang, who began working on the creation of Moonshot when he was a teenager, what he thought about the role of the teacher who uses AI for instructional purposes. Wang, rather than looking forward, surprisingly looked to the past for how he envisions this evolution.
“Ideally, teachers should embody the role of lifelong learners in their study, work, and life,” Wang explained. “They are serving as role models to inspire learners how to discover their missions and embark on a fulfilling and meaningful life journey. This reflects a return to the ancient Chinese educator’s role of imparting moral values, inspiring mission, and resolving doubts.”
Jerry Lu, the leader of the digital and natural science center at Beijing New Talent Academy, shared how he approaches the role of a teacher in an AI-enriched classroom.
“I think the teacher should be more like a director rather than an actor in a film,” he told me. “The teacher should prepare the students by telling the story of what they will go through, which means setting goals and context. The teacher should help students to understand their roles, which means sharing best practices in using AI tools. The teacher needs to monitor the learning process to make sure the student is on track. And finally, the teacher must comment on the student’s performance while collecting evidence for assessment.”
During a visit to Beijing Haidan Future Academy, I was pleased to see students participate in an activity in which they examined the differences between using ChatGPT and using a search engine (see image) to find information. This distinction tops my ranks for the most common misuse of AI. A day later at Beijing Navigation School, I watched students using ChatGPT as a search engine with no understanding of the distinction. Net gain from this experience: Zero.
Conversations with teachers at the 11th Education Innovation Conference of China in Xiamen did not produce clarity. Chinese students, unyoked by the 13-year-old age restriction on AI use in U.S. classrooms, were using AI in unstructured ways. Like the vast majority of U.S. schools and districts, AI-acceptable use policies were not in place in China. It is difficult to define the role of a teacher if the use of revolutionary technology is ungoverned.
Frustrated, I returned home determined to create a model that would describe how a teacher can maximize learning efficiently and ethically in a classroom filled with software and hardware powered by AI.
What Makes AI Teaching Tools Different
For purposes of my argument, I will divide the most powerful AI-enabled instructional tools into three categories: 1) tutoring systems; 2) course and teacher chatbots; and 3) textbooks. I’m aware that primitive versions of these tools existed prior to AI and include iterations that are not AI-enabled. More on that in a moment.
Examples of AI-powered tutoring systems include:
AutoTutor: A conversational dialogue-based system that focuses on teaching computer technology through leading questions and adaptive feedback.
Carnegie Learning’s LiveHint AI: A math tutor built on a large language model that helps middle and high school students by leveraging data from millions of students solving math problems.
Khan Academy’s Khanmigo: An AI-powered tutoring system that provides personalized feedback and support to students as they work through Khan Academy’s exercises.
DreamBox: An intelligent educational system that uses advanced algorithms to adjust content based on students’ progress and comprehension, supporting their math skills.
Cognii: An AI-powered tutoring system that provides personalized feedback and support to enhance students’ learning experiences.
Examples of course and teacher chatbots include:
University of Utah: UBot is a virtual tutor that guides students to answers in specific curriculum and course material, assisting them outside traditional tutoring hours.
University of Illinois: The CAII team is developing an AI-powered chatbot that can ingest course materials and function as an effective teaching assistant, answering questions and explaining course content.
University of the People: YUJI is an AI chatbot designed to engage students, answer queries about course materials, and facilitate discussions in online degree programs.
Arizona State University: Sam is an AI-powered chatbot that simulates patient-provider interactions for health sciences students, allowing them to practice motivational interviewing skills.
Harvard University: PS2 Pal was created by lecturer Gregory Kestin and senior lecturer Kelly Miller to support students taking the course Physical Science 2. This experiment resulted in a research paper with interesting results about the efficacy of course-level bots.
Examples of AI-powered textbooks include:
ViewSonic: Provides an “AI Textbook” that offers personalized adaptive learning tools with generative AI features, such as real-time assessments and instant learning diagnostics.
Pearson: Offers AI-powered digital textbooks that allow students to highlight challenging sections and receive tailored, simplified explanations from the AI tool.
McGraw Hill: Launched AI tools for students, including AI Reader in their ebooks, which allows students to request alternative explanations or simpler language for highlighted text.
KITABOO: Provides an AI-powered digital publishing platform that creates, enhances, and distributes immersive learning experiences with features like interactive assessments and customized content creation.
CogBooks: Integrates AI technology with educational content to create adaptive e-books that cater to diverse learning styles and offer personalized recommendations based on individual progress.
What makes these AI-powered tools fundamentally different from earlier versions of the technology is that they are all capable of an adaptive response. Just like a human teacher. These AI tools use their verbal/oral/visual interactions with a student to modify difficulty level, change modality, adjust tone and voice, provide personalized examples, contextualize content, level vocabulary or switch languages, and adjust pacing. Just like a human teacher.
Yes, this process is algorithmically based – I am not suggesting the AI feels empathy nor am I suggesting the AI is sentient. Those characteristics are irrelevant. What the AI does is respond to the behavior (via input) of the learner and provide a level of differentiation and personalization that mimics the techniques that every good teacher uses. The fundamental difference between humans and silicon? The AI never gets tired, overworked, stressed, or distracted. Students operate on a 24-7 schedule so when they need to know something they need to know it.
There is a reciprocal flow of information between the learner and the instructional tool. That was not the case with prior iterations of these tools, although we saw the beginnings of this technology in the widespread use of adaptive assessment systems designed by companies such as Pearson Education, McGraw-Hill Education, Curriculum Associates, CogBooks, Knewton, and Magic EdTech.
The Teacher as Conductor
Let’s think about the role of the teacher in a flipped classroom that has access to the three AI-powered tool groups (tutoring systems, chatbots, textbooks.) I identified earlier. Yes, the teacher is no longer the sole purveyor of content, assessment, support, and differentiation. So what are they doing?
I prefer to use the analogy of the orchestra conductor to describe how the teacher interacts with these new instructional tools (see illustration). The conductor does not play the instruments during a symphony, but she ensures that musicians are following the score to produce a melodious sound. So too will the teacher conduct the AI-powered tools in their classroom.
Over the last couple of years, I have done numerous AI 101 for Teachers workshops in the U.S. and China. The most common response I hear in these sessions goes something like this: “This thing can do all my work for me!” Not quite. Musicians do not put a conductor out of work. The conductor is in fact the most important person in an orchestra.
The 80/20 Approach
One of my favorite guidelines that frames the role of the teacher who uses AI was described in an FAQ posted on the Magic School website: “Use AI for initial work, but make sure to add your final touch, review for bias and accuracy, and contextualize appropriately for the last 20%.” While this dictum relates to the creation of curriculum, it does give a reference point for what humans bring to the learning party.
Math educator Dan Meyer, writing in his Substack section MathWorlds, reversed the 80/20 rule, using data from an informal teacher survey to evaluate the readiness of AI-generated content. Meyer and his followers believe that AI content is only about 20 percent classroom-ready without significant human intervention.
The trifecta of new AI-powered instructional tools (tutors, chatbots, textbooks) are so much more than curriculum creation tools but they are not enough to lead learners in the construction of contextualized skills and knowledge. A human has to be actively involved in the process because learning is a social process, driven in part, for better or worse, by emotion. These tools give information – not meaning.
Regardless of which numerical ratio you choose to define the role (and significance) of the teacher, AI will need a human partner (conductor) to be an effective instructional tool. Don’t plan on early retirement just yet.
The post How Teachers Can Orchestrate a Classroom Filled with AI Tools appeared first on Getting Smart.