OpenAI’s release of the latest iteration of its software, ChatGPT4o, took my thinking about PBL and AI in a new direction.
I’ve been loud and unwavering in my opinion of generative AI: It is the most effective tool for the strenuous task of designing a rigorous and relevant project. I’ve been less clear in my opinion about the relationship between generative AI and student outcomes.
Let’s think back to how current models of PBL design evolved as we explore that relationship.
The origins of the Backward Design Model
Ralph Tyler had a long and illustrious career in education, focusing primarily on assessment. In 1949 he published Basic Principles of Curriculum and Instruction, introducing the term “backward design” to education. It gained wide currency in 1988 when Jay McTighe and Grant Wiggins released Understanding by Design.
Nearly every model of PBL design, especially the standards-based model promoted by the Buck Institute for Education (PBLWorks), suggests that the design process begins with the identification of knowledge and skills the students should master via the project.
Projects that don’t first begin with the identification of student outcomes are much derided in the field and often labeled as “dessert” projects, my least favorite of which is the California Missions Projects, the bane of every 9- and 10-year-old in my state.
Content Is King
Mastering content is the primary goal of any good project. That means the content must be taught, and more importantly, learned. This is where traditional teaching techniques enter into the PBL process. We wrote a blog about the difference between PBL and PBL and the importance of putting the learning at the core. I had to repeat this idea several hundred times a few weeks ago in China: Just because you are doing PBL does not mean you do away with textbooks, lectures, tests, or homework. The project contextualizes those tasks and tools because the students need content knowledge to complete the project.
Do all students master the content in this process? Nope. Do all students master the content in a traditional classroom? Nope.
Tutoring, either with individual learners or small groups, should always be a part of the remediation/reteaching process in a PBL classroom. As all teachers know, that is easier said than done. But now we have a tireless digital friend who can help us with this task.
AI as Tutor
Sal Khan released a new book, Brave New Words: How AI Will Revolutionize Education (and Why That’s a Good Thing), a few days ago. He reveals that his team at Khan Academy began working with the team at OpenAI (the builder of ChatGPT) before the product reached the public in November of 2022. The shared goal was to create an effective, personalized, safe, and ethical tutor for all learners. Khanmingo, now free to all teachers, is the fruit of that collaboration.
Khanmingo, in widespread use for more than a year, has been supercharged by the May release of ChatGPT4o. The conversation speed is real time and the fluency is unerringly humanlike. The software has the ability to decipher emotion and respond to facial expressions via the device’s camera. Now, every student has an endlessly patient tutor who is programmed to help them learn at their speed while offering timely feedback and emotional support.
As you can imagine, Khanmingo is not the only AI tutor that teachers and students can work with. A day after ChatGPT4o dropped, Google released an updated version of its AI Gemini and its similarly capable Project Astra. It too has the ability to be a highly relatable tutor that reads and responds to visual input via the camera and delivers tutoring support in sync with every learner’s needs.
Microsoft offers Reading Coach, an AI-powered literacy tutor that personalizes instruction and support for all learners. As Microsoft explains, Reading Coach has the ability to generate stories and can give learners access to a library of leveled passages from ReadWorks.
There are of course a multitude of other AI-powered tutors, mostly powered by ChatGPT4o, Gemini or Claude, the third of the big three LLMs.
AI and PBL
I teach an online asynchronous course called Designing PBL with Gen AI. Obviously, I believe in the efficacy of generative AI as a tool that eases the design burden of PBL. In that use case, generative AI makes curriculum design easier for the teacher.
As we discussed earlier, that process follows a backward design model that first identifies learning outcomes (both knowledge and skills). Then the burden is passed from teacher to the students, who must master the outcomes while participating in a project.
As we know, not every student learns at the same pace. This can greatly affect the success of a project. The new AI tutors assist the process of learning the content every student needs to know to effectively participate in and complete the project. PBL’s new best friend is AI.
David Ross (@davidPBLross) is the retired CEO of the Partnership for 21st Century Learning and the former Senior Director of the Buck Institute for Education (now PBLWorks). David was an 11th grade American Studies (History and English 11) team teacher. David created curriculum design templates, exemplary projects, rubrics for critical thinking and collaboration, and project management techniques.
