I think all of that AI for teachers startup energy is heading the wrong direction. I'm sure that generative AI is already able to generate some clever lesson ideas, work samples, etc. That's all fine, but what I'm sure generative AI can't do right now and will struggle with for a long time is coherence. It's really hard to design a curriculum that builds on itself, sets up future lessons, revisits and expands on past representations, teachers high-leverage routines and then makes effective use of them, etc. Any generative AI that is mostly trying to help teachers plan lessons will just result in more patchwork quality. We should have higher standards for curricular coherence than that.
The direction I'd rather AI startups go in is to focus on all of the non-planning tasks that fill teachers' days. Here are some tasks I've done in the last week that I wish AI could have done for me:
Post next week's assignments to Google Classroom
Pull homework completion from DeltaMath to Google Classroom, and then from Google Classroom to SchoolRunner
Write course descriptions for my classes
Pull and organize a list of missing assignments from the last few weeks
Download my students' NWEA data and clean up the spreadsheet so it's easy for me to work with
Find ISBN numbers for a bunch of math books I'm ordering on a grant we got
Write up an agenda for our 7th/8th grade team meetings and email out a calendar invite with the agenda attached
I understand that these are inherently more challenging. They involve multiple platforms and aren't very word-focused so they don't play to generative AI's strengths. But I think this points to a bigger problem. Lots of people want to solve the sexy-sounding challenge of "help teachers plan lessons" even if they don't do a very good job, and no one wants to solve the less-sexy "free teachers from some time-consuming administrative tasks."
I have zero credibility here as a curriculum publisher but the coherence issue resonates strongly with me. Like, yes, I'd prefer an AI platform help me piecemeal productive & engaging daily lesson ideas for my kids over nothing. And I would prefer my school or district to buy a good curriculum even more.
Hey Dylan, hope your school year is off to good start in CO. Good points. There's another path at scale to helping teachers with admin tasks. Combine 2 things common in other sectors, bring to K-12.
First, there's a near unlimited amount of high quality, $5 to $10 hour admin help in Bangladesh and Pakistan et al. Millions of Americans hire them, with all sorts of positive externalities to "trade" with developing world. Few teachers do. It may be a while before AI handles all that.
Second, I've written before that each teacher should control the $10,000 per year that the district spends "for them." Teacher buys any PD, teachers buys any curric (Desmos would do really well in this world), teacher buys services (5 hours per week of help per above would cost $900 to $1800 per year that your district should cover). Teacher controls the "mix" of spend as well. A 2nd grade teacher might put $8k of her $10k into some truly awesome class trips - great.
Thinking AI will solve the education issues out there is similar to the notion of a silver bullet curriculum that will equal the playing field for all learners. Teaching and learning is inherently social, sure having great content is a huge help. Borrowing a phrase heard at a panel discussion, I wish I could remember who said it.... " the curriculum is the science and the educator is the art".
Speaking as an Elementary educator, the development of learners, feeding and flaming curiosity and providing a safe environment where children are able to explore and learn to work together is a huge part of the work, perhaps the most important. As for the content and lesson plans, no matter how well written, it is the delivery and the understanding of how children, teens and even adults learn best that allows that content to resonate and create meaning. Teaching is an art and yes, there are many a routine and inefficient task that can be helped with AI, but what about the art and humanity of the profession. Is AI addressing the craft of teaching as well?
The points you make, Dan, about the most common teacher complaints and political solutions needed to reverse the teacher attrition and attract more people to the profession speak directly to the "artist" behind every effective educator. In actuality, it stings and potentially aggravates the challenges facing the profession with claims that AI can "fix teaching." What we really need is AI to write a monologue touting the impossible paradoxes of life as an educator and have America Ferrera deliver it on the big screen during a major motion picture....a la Barbie!
There are already AI based tools targetted towards reducing the workload of teachers like https://essay-grader.ai. This industry is going to get crowded very soon.
Even to cut down on that 5% of planning/prep would be super useful and empowering for me (high school math teacher of six years). Also, with the classes with more students who need extra support, I tend to dedicate more than 5% of my time to planning/prep. I really think a big possible win here is formatting. Much of the most frustrating/boring time when planning/prepping is formatting. Your idea of integrating something with Google Apps for Ed makes a lot of sense. I think when it comes to formatting, it would be just as useful for math teachers as humanities teachers. One Desmos-related example: I created some Google slides recently for my students from the Algebra 1 Desmos/Amplify curriculum Unit 2, Lesson 3, Equivalent Equations. Here's what it looks like: https://docs.google.com/presentation/d/1HrzseGxxpjEdK7S6mRPEAMlKi3RsDSbgU-1OGwu4ljM/edit?usp=sharing
It's not that pretty, and I feel like AI should be able to take an input, like the PDF of the cards, and when asked, make basic slides, with one card per slide, to function as a kind of anchor chart/lesson synthesis. Basically what I did, minus the red "X"s, green checks, and explanations. It would save some of the work. Also, the AI should be able to add spacing between questions and do other formatting. Even though it's only one of the things teachers have to do, it would help.
This makes me wonder if we should focus AI teacher support efforts on reducing the load for tasks like administrative data collection and analysis of assessment data to free up teacher time for deep thinking about content and planning.
I really appreciate this perspective. Some of the hype around AI and education is rooted in a fundamental misunderstanding of what education and learning is all about.
I think all of that AI for teachers startup energy is heading the wrong direction. I'm sure that generative AI is already able to generate some clever lesson ideas, work samples, etc. That's all fine, but what I'm sure generative AI can't do right now and will struggle with for a long time is coherence. It's really hard to design a curriculum that builds on itself, sets up future lessons, revisits and expands on past representations, teachers high-leverage routines and then makes effective use of them, etc. Any generative AI that is mostly trying to help teachers plan lessons will just result in more patchwork quality. We should have higher standards for curricular coherence than that.
The direction I'd rather AI startups go in is to focus on all of the non-planning tasks that fill teachers' days. Here are some tasks I've done in the last week that I wish AI could have done for me:
Post next week's assignments to Google Classroom
Pull homework completion from DeltaMath to Google Classroom, and then from Google Classroom to SchoolRunner
Write course descriptions for my classes
Pull and organize a list of missing assignments from the last few weeks
Download my students' NWEA data and clean up the spreadsheet so it's easy for me to work with
Find ISBN numbers for a bunch of math books I'm ordering on a grant we got
Write up an agenda for our 7th/8th grade team meetings and email out a calendar invite with the agenda attached
I understand that these are inherently more challenging. They involve multiple platforms and aren't very word-focused so they don't play to generative AI's strengths. But I think this points to a bigger problem. Lots of people want to solve the sexy-sounding challenge of "help teachers plan lessons" even if they don't do a very good job, and no one wants to solve the less-sexy "free teachers from some time-consuming administrative tasks."
I have zero credibility here as a curriculum publisher but the coherence issue resonates strongly with me. Like, yes, I'd prefer an AI platform help me piecemeal productive & engaging daily lesson ideas for my kids over nothing. And I would prefer my school or district to buy a good curriculum even more.
Hey Dylan, hope your school year is off to good start in CO. Good points. There's another path at scale to helping teachers with admin tasks. Combine 2 things common in other sectors, bring to K-12.
First, there's a near unlimited amount of high quality, $5 to $10 hour admin help in Bangladesh and Pakistan et al. Millions of Americans hire them, with all sorts of positive externalities to "trade" with developing world. Few teachers do. It may be a while before AI handles all that.
Second, I've written before that each teacher should control the $10,000 per year that the district spends "for them." Teacher buys any PD, teachers buys any curric (Desmos would do really well in this world), teacher buys services (5 hours per week of help per above would cost $900 to $1800 per year that your district should cover). Teacher controls the "mix" of spend as well. A 2nd grade teacher might put $8k of her $10k into some truly awesome class trips - great.
Thinking AI will solve the education issues out there is similar to the notion of a silver bullet curriculum that will equal the playing field for all learners. Teaching and learning is inherently social, sure having great content is a huge help. Borrowing a phrase heard at a panel discussion, I wish I could remember who said it.... " the curriculum is the science and the educator is the art".
Speaking as an Elementary educator, the development of learners, feeding and flaming curiosity and providing a safe environment where children are able to explore and learn to work together is a huge part of the work, perhaps the most important. As for the content and lesson plans, no matter how well written, it is the delivery and the understanding of how children, teens and even adults learn best that allows that content to resonate and create meaning. Teaching is an art and yes, there are many a routine and inefficient task that can be helped with AI, but what about the art and humanity of the profession. Is AI addressing the craft of teaching as well?
The points you make, Dan, about the most common teacher complaints and political solutions needed to reverse the teacher attrition and attract more people to the profession speak directly to the "artist" behind every effective educator. In actuality, it stings and potentially aggravates the challenges facing the profession with claims that AI can "fix teaching." What we really need is AI to write a monologue touting the impossible paradoxes of life as an educator and have America Ferrera deliver it on the big screen during a major motion picture....a la Barbie!
There are already AI based tools targetted towards reducing the workload of teachers like https://essay-grader.ai. This industry is going to get crowded very soon.
Even to cut down on that 5% of planning/prep would be super useful and empowering for me (high school math teacher of six years). Also, with the classes with more students who need extra support, I tend to dedicate more than 5% of my time to planning/prep. I really think a big possible win here is formatting. Much of the most frustrating/boring time when planning/prepping is formatting. Your idea of integrating something with Google Apps for Ed makes a lot of sense. I think when it comes to formatting, it would be just as useful for math teachers as humanities teachers. One Desmos-related example: I created some Google slides recently for my students from the Algebra 1 Desmos/Amplify curriculum Unit 2, Lesson 3, Equivalent Equations. Here's what it looks like: https://docs.google.com/presentation/d/1HrzseGxxpjEdK7S6mRPEAMlKi3RsDSbgU-1OGwu4ljM/edit?usp=sharing
It's not that pretty, and I feel like AI should be able to take an input, like the PDF of the cards, and when asked, make basic slides, with one card per slide, to function as a kind of anchor chart/lesson synthesis. Basically what I did, minus the red "X"s, green checks, and explanations. It would save some of the work. Also, the AI should be able to add spacing between questions and do other formatting. Even though it's only one of the things teachers have to do, it would help.
This makes me wonder if we should focus AI teacher support efforts on reducing the load for tasks like administrative data collection and analysis of assessment data to free up teacher time for deep thinking about content and planning.
Love reading your stuff Dan. We were in a grad class together (Alan Schoenfeld’s problem solving!) keep up the great work!
I really appreciate this perspective. Some of the hype around AI and education is rooted in a fundamental misunderstanding of what education and learning is all about.