5 Actual Use Cases of AI in Education 🤖 Newsletter # 68
What areas has AI truly impacted educators, learners & workers?
Hi there! Alberto here, joining from Madrid this week.
The Transcend Newsletter explores the intersection of the future of education and the future work, and the founders building it around the world.
We welcome 187 new readers to the newsletter since our last post. If you love reading about the future of education and work, hit the ❤️ button and share it with your friends!
Yes, yes, we all know that AI can transform education.
Articles abound about how it will personalize learning, make it more fun, and even create little robot teachers that will explain things in front of a blackboard (my personal favorite if you ask me).
But a year into the generative AI boom, how is AI being used in the day to day of education?
What you’ll read below is a summary of my observations from seeing hundreds of startups using AI, and the actual use cases that are applied in classrooms and learning spaces all over the world. We are looking for pain killers instead of vitamins, or solutions that solve a really important problem for the education space. It’s not comprehensive, so I want to hear what you are seeing in the post’s comments!
A brief history of AI use cases
AI applications in education have changed every decade, creating different “eras”.
This past decade was the “Uber era” of education. Unicorns like Uber, Amazon, or Google used AI to improve route efficiency and make the most relevant matches between supply and demand. This bled into a movement to Uber-ize education: use AI advancements to predict student paths and match them with the most relevant learning resources.
A 2017 literature review of AI applications in education concluded that the main use cases for AI were predicting student performance and increasing student retention. The underlying hypothesis was that you could identify student patterns using AI, and address them early to increase student success. Turns out students are a little bit more complex than car routes, so this era wasn’t particularly fruitful.
AI applications moved away from student information, and into direct instruction, giving way to the “AI Tutor” era: AI products sought to solve the always elusive 2-Sigma problem originally formulated by Benjamin Bloom, by creating smart tutors available 24/7 for students. We can’t give every student a teacher, so let’s give them a chatbot instead.
Turns out bots aren’t particularly good educators! Students were neither interested in them, nor motivated to learn more. Also this wasn’t a new use case – the idea of teaching machines and personalized learning have been around for a while, without many results.
In this last year, we’ve seen millions of students and teachers implement AI in their classrooms and learning practices. The majority of use cases incorporate AI not as a tutor, but as a complementary guide that can augment the learning process.
This is the “AI Augmentation era”, and it’s one I’m excited about.
JCR Licklyder wrote about the power of computers augmenting human work 60 years ago, and the latest crop of edtech founders building AI for edtech have listened to his advice. Below are 5 great use cases that align with that principle of augmentation, not substitution, and are growing today.
5 Use Cases of AI in Edtech Today
Let’s review five use cases we are seeing “in the wild” in 2023:
1. AI Copilot for educators, managers and leaders
I’ve come to think of AI as a really smart contributor who can’t make decisions on its own, but it can help you with any decision you make.
And this is exactly what a copilot is: someone who’s on your side every hour of the day, making suggestions and pointing out best practices.
The “copilot” use case has been well-documented recently, but we’ve seen it transform over this last year. Initially, there was a lot of excitement around general copilots that cover a wider range of decisions and use cases (inspired by the Github Copilot).
But over the year, as the market has been flooded with AI copilots, copilot startups have had to niche down further: they’ve focused on personas like teachers and managers with communication and coordination needs. They’ve specialized in more specific processes (testing, remembering, creating assessments, etc.), rather than as an all-in-one tool for learning.
Our prediction for the next year is that there will be a great unbundling of AI copilots. Startups will continue to build for more specific personas and processes, with a few general copilots that will look to win a category by offering a portfolio of copilots (such as MagicSchool or SchoolAI in the K-12 space, or Waggle in the management)
There will be some areas where copilots won’t be helpful (check out Dan Meyer’s excellent recent post on the topic) and will see very different products for K-12 compared to higher ed or corporate learning.
Startups in the space
K-12: Magic School (teacher copilot), School AI (teacher copilot), Diffit (leveled resources) Brisk (teacher copilot), AI for Education (AI Training for schools)
Managers & corporate: Waggle (manager copilot), Sparkplan (Sparkplan), Bunch.ai (managers)
2. Flipped Classrooms Chatbots
We thought AI would transform education because it could answer anything you ask it. Instead, it may be its ability to ask great questions that will really impact learning.
AI is able to take into account previous answers and data to inform questions that are adjusted to your level of understanding, which makes it very powerful!
The movement behind “flipping the classroom” sustained that learning shouldn’t happen in a lecture: it should start at home, where you learn the basics, and then you complete the rest through discussions and activities in the classroom.
This type of active learning is exactly what’s behind the idea of “flipping the chatbot”: instead of using it to explain things directly to a student, you can use it to ask you great questions after reading a chapter or lecture, which makes the learning more active.
Ethan Mollick has done a lot of work to prompt AIs to create interesting learning partners, not just one that gives you the answer right away, and I’m excited about this approach to learning. ChatGPTs own prompts are excellent as well!
Startups in the space
3. AI to assess complex answers
This may sound strange, but we are missing out on a lot of great learning insights simply because of our limitation of handling too much data at once.
Think about all the papers you’ve written through school, all the projects you’ve worked on in your career, and all the emails you’ve sent. There are great learnings hidden into those massive amounts of data, but a human would have to spend a lifetime sorting through it to make it through. In my opinion, they are waiting to be collected and sorted by an AI!
Last week, I fed ChatGPT all my newsletters, and asked for advice on which authors my writing resembled the most and how to improve my writing based on what worked for them. It was able to take in all that data and spit out relevant insights and advice for me!
This type of assessment is one that AI can help create in a recurring manner.
Companies like Take2 are able to use it to analyze candidate’s responses to questions, skill assessments and projects they’ve done. The employer can access a neat summary of how all these projects and skills relate!
Startups in the space:
Take2 (skill-based recruiting), Buo (talent mapping), TeachFX (educator feedback), Examind (higher education assessment)
4. AI as a language learning tool
Tired of learning useless sentences that won’t help you learn a language? You are not alone.
Many language-learning startups have been limited by their ability to generate quality learning content, leading to ridiculous examples like the one above.
AI has proven to be an incredible tool for language learning: it helps upgrade the speed and quality of translations, explanations, new content in the target language, and even creates conversational bots (as separate products or by prompting ChatGPT).
Over time, AI’s improvements in speech recognition and responses will make language learning much more interactive and conversational.
Startups in the space:
Speak, ELSA, Linguistic, Voys, Speechy
5. AI to brainstorm ideas
I use ChatGPT every single day to brainstorm new ideas. Whether it’s coining new terms for the newsletter, generating ideas for a curriculum, or making outlines for new content ideas, I find it super helpful as a creative aid. I do this even during meetings (sorry!) when I hear an interesting idea that can be extended while I focus on the conversation.
Would I use ChatGPT to write my own newsletter? Nope, that’d make for pretty boring posts. But would I use it to brainstorm ideas and give feedback? I already do, so it’s ChatGPT’s fault if you don’t like this post.
More specific tools for writing (Lex or Hyperwrite) or educational curricula (Diffit) are now popping up everyday, and will continue to do so through the year to empower more customer personas with creative processes.
Startup ideas: Diffit (for teaching resources), Hyperwrite (writing assistant), Taskade Brainstorm (brainstorming feature), Lex (writing processor and assistant).
Conclusion
As we continue to explore these use cases for AI, it's important to remember that AI should be used to augment human work, not replace it – this will be the main idea driving AI edtech innovation for the next decade.
Are you in the early stages of building AI solutions for education? Join us next week for our Panel Discussion + Open Q&A about Early-stage Fundraising for AI-Edtech Startups!
You will get to hear from 3 awesome founders: Gonzo from Examind AI, Victoria from Linguistic, and Vlad from Difft, who closed rounds this year and are in the early stages of solving critical problems at the Higher Ed, Upskilling, and K-12 levels respectively.
Thank you for reading!
Did you enjoy reading this piece?
Hit the ❤️ button to help us reach more awesome people like you!