A review of Khan Academy's Khanmigo and the hopes of chatbot boosters for AI in schooling.
This article reminds me of the problem with college professors. Only a small fraction of students become professors. They typically think differently than the other students but they often teach as if all students were like them. Then they wonder why their classrooms dont work well
Oh Dan this is such a needed conversation. Thank you.
This argument is right in synch with Francis Su's positioning of mathematics as a part of the larger project of humanities, and that our teaching (and more importantly, the learning) in our classrooms is sooo much more than the content -- or even the combination of content and pedagogy. Our work is largely a relational engagement, supported by pedagogical and cultural structures, set in a [not *entirely trivial*] context of this beautiful and powerful discipline called mathematics. It just so happens that math is a wonderful context in which to learn all this relational skill, as it provides natural challenges, which begs us to creatively/intentionally engage our students with rich, personal support. I am deeply grateful for this article Dan, and the messages within it -- keep 'em comin'. : )
Nice post. I agree with all of your points except for one.
I do not think the reason that ed-tech-topians are advocating for atomized learning is because it’s how they enjoyed learning when they were students. I think it’s because it scales easily, and scalability is what they care about.
At least when it comes to other people’s kids.
When it comes to their own kids, though, I’m fairly confident that the people banging the drum (and cutting the checks) for AI-based schooling — the execs at Gates, Reach Capital, etc. — have little use for ChatGPSal.
Loved every sentence of this. 👏👏👏
Great article and synthesis of the dirfference between learning algorithm and understanding its place in the real world
Enjoyed this, thanks!
I feel like you are missing a few points here. Sal Khan isn't advocating for Ai to replace teachers but to supplement them, especially in the classroom. Even with a low student-teacher ratio, kids slip through the cracks. He says imagine a classroom with kids doing their work on computers (already happening) where they have access to an Ai tutor. Instead of the teacher getting pelted with questions, the Ai can help kids on the spot and the teacher can follow up. The teacher can also see what kids are asking and the Ai can alert the teacher that half the class has asked the same question in the last 10 minutes. That's really powerful technology and as a teacher, would be a HUGE help. We do this already in the classroom but the Ai can do it faster and help more kids at once. Sal also demoed how Khanmingo can answer those "big questions" as "why is this relevant to me" and "should I wait to go to college"? In reality, guidance counselors aren't answering those questions either but provide a conversation with the student to get the answer themselves. Chatbots will be able to mimic some of that.
As for "revolutionary ed tech" not living up to its promises, again, I don't understand your point. Sal Khan advocated for a flipped classroom and focus on mastery, not grades, which research shows is a very good teaching model. Mastery is all the talk now in the ed world. Many teachers use flipped classroom model and I suspect many teachers would like to if they didn't have time/money/resource/tech barriers. MMOCs did become certification courses. And yes, anyone could have predicted that hosting courses without any "reward" like a grade or certification would fail because most people aren't inherently motivated like that.
Enjoyed reading this. Did you see this article? https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/2804309?guestAccessKey=6d6e7fbf-54c1-49fc-8f5e-ae7ad3e02231&utm_source=For_The_Media&utm_medium=referral&utm_campaign=ftm_links&utm_content=tfl&utm_term=042823
Not about education, but so interesting and so sad. My students didn’t seem to share my disappointment that AI demonstrates more empathy than doctors. I agree with your synopsis about school & relationships, but the empathy piece in the above has me wondering about that. Kids seem to want to have a relationship with the Snapchat AI. Even though it’s not real, could they feel the same about Khanmigo?
Original reflection on the current unoriginal regurgitation coming from AI! Root Cause Analysis is coming in the world of AI, but not here yet. Relevance, Relationships, and Rigor!
I myself have used Bing, not even Khanmigo, for learning and it was immensely helpful and I learned something not nothing.
I think you make many unwarranted assumptions:
1. That the tech won’t evolve, that it will stay the way that it is today frozen in time, which is wrong, very likely it will evolve to be able to handle “big questions”, which I think it already can do with some tweaks.
2. That this is meant to replace teachers, it never was, it is meant to supplement them and to help those who don’t have access to teachers to begin with, in poor countries or communities etc.
Thanks for this perspective and it instantly reminded me of this https://youtu.be/GEmuEWjHr5c
I find interesting the paragraph about Nerds building models for schooling in their image, in their likeness. Nerds who clashed with school and would much rather learn from a book or a computer have been building tools to replace teachers with computers: a model that only helps nerd students and not the general public.
This is the reason they fail, I agree, but not because of the focus on small questions instead of big questions. The reason I think ChatGPT and Khanmigo won’t make a difference is because Nerds (and some Teachers) build tools assuming kids enjoy learning and that all that is missing is quality, accessible, inspiring materials. But average students hate school and are in school against their will so they would resist engaging in any learning, no matter the effort of the teacher and the quality of the material. Why would they want to talk to Socrates when they can watch TikTok. We need to find better ways to motivate to learn.
Their parents may be largely satisfied or actually relieved to delegate their childcare to school). The title of the linked article is misleading and if you read the full article it actually says that 73% of K-12 parents are satisfied with quality of _oldest_ child's education but the public now tilts more dissatisfied (54%) about the quality of K-12 education in general.
I really appreciate your skepticism and historical perspective on how cycles of edtech return to dehumanizing, relationship-obliterating models that don't work for most students and reduce school to information transfer.
But this time is different, I swear! Maybe? I don't know at what point a virtual AI tutor stops being called a "chatbot", but in the future could they...
• Give two possible answers and ask the student which one they think is more likely to be right and why.
• Use voice and handwriting as input, creating more natural interactions.
• Know a student's personal learning path and relate new questions to recent learning and their study plan.
• More to your point about relationships: connect students questions to the knowledge/skills/interests of classmates and teachers. "Good question. Kim was wondering the same thing yesterday – go ask her."
Some beneficial tools may be created along with the inevitable deluge of mediocre content/cheating/shortcuts/distractions.
I appreciate where this is coming from and the need to guard against the latest wrong but popular trend that is going to "disrupt" education. But I disagree with the framing that this is helping people learn nothing. These new tools can take a narrow topic or skill and help you drill. Effective drills and personalized practice are an important thing to be able to offer at scale to everybody, even if they aren't the most important thing.