Elizabeth here, co-founder (and clinical child psychologist). Fair question. Catalin's post was about the engineering (he is my co-founder and our CTO), so the learning side got short shrift :).

Here's what it actually looks like for a child. Say a 6-year-old is reading a story out loud (I will use a reading example here). The tutor is listening to every word. When she stumbles on "chick," it doesn't just tell her the word; it decides, based on her history, whether to break it into sounds, point back to a pattern she's seen before, or let her wrestle with it a moment longer because she's close. If she misses the same pattern twice, that digraph shows up woven into her next story. If she reads fluently but can't tell the character what happened in the comprhension conversation after the story, she gets another text to work on comprehension instead of just pushing harder words. The instructional approach isn't novel or new, it's what a good teacher does, grounded in the science of learning. We run evals on the interactions and real subject matter experts are grading and annotating the behavior. What's new is doing it responsively, for one specific child, on every turn.

On engagement: I'd push back a little on the framing that engagement and learning are separate things (anyone on our team will tell you this is a drum I have beaten for years). A disengaged child learns nothing, no matter how good the pedagogy is. But we're not optimizing for time-on-screen. The lessons and sessions are bounded. The engagement work exists so the child stays in the productive struggle zone long enough for the teaching to happen.

Why AI: it's not that AI "needs" to be the solution. In fact, a great human tutor is better, full stop, but it has never scaled. A classroom teacher with 25+ kids teaches to the middle. This is the first technology that can make real-time, child-specific teaching decisions, which is what tutoring actually is. More on the pedagogy here if you're curious: https://www.ello.com/our-teaching-approach)

> I'd push back a little on the framing that engagement and learning are separate things (anyone on our team will tell you this is a drum I have beaten for years). A disengaged child learns nothing, no matter how good the pedagogy is.

Engagement and learning are definitely separate things; you're right that engagement is required, but that's only part of it. This is a classic case of 'necessary but not sufficient'.

A disengaged child is not learning; an engaged child might be.

Attention is not all you need? ;-)

Thank you -- your post was downvoted into being hidden for a while, sorry I didn't see it earlier. This context is helpful, the example of the system delivering reading challenges that match their struggles makes a lot of intuitive sense to me.

I really appreciate that (it seems to me) your goal is not to replace human tutors, but to raise the general baseline. You emphasize scaling, how does that work in practice if you're trying to target audiences who may not have access to devices that can run your program? What is your plan from the perspective of funding and resources to scale infrastructure as needed to support these audiences?

Edit: I also think of other learning systems like duolingo and the application of tablets and computers in schools which begin from good places, but I'm curious if you are studying these alternatives and what you have learned from them?

I really think your goals are great, and if you're starting your design of this system from research about effective learning methodologies and working backward from there rather than starting from AI and working backward from there that erodes a lot of my personal skepticism about a project like this. I hope you find a way to make this work.

taught mental math (abacus) to 3-15yr olds for 10+ yrs. A great teacher notices the gaps in the fundamentals and fixes those along with variants before piling on.

ello looks great, great work

Hi! I'm Elizabeth (one of the co-founders of Ello). This is an interesting question. I actually think there is more overlap than people might assume, but it's a bit more because adaptability to various approaches to learning is part of the point. While I'm not deeply familiar with the Sudbury School model, I think there are various approaches to teaching kids that are successful because different approaches serve different types of kids and learners. I can see why this approach would be successful for a certain profile of student for whom it's the right fit.

We start from the belief that children are naturally curious. Our job is to build something engaging enough that a child wants to interact with it because it is interesting and rewarding. If a child in a Sudbury environment never chose to use it, I would see that as useful feedback for us, not a problem with the child. There are opportunities for kids to explore and incorporate their interests within our app.

I also think it is completely fine if a child uses it for a while, disappears for months, and comes back. Learning is rarely linear, and technology should be able to pick up wherever the child is.

On reading, we’re firmly grounded in the science of reading, so we teach through explicit phonics rather than whole-word memorization because that is best practice.

On math, we’re much more interested in helping children build intuition and conceptual understanding than simply getting answers. AI gives us the flexibility to use conversation, visual models, stories, or symbolic math, depending on what helps a particular child understand.

One thing AI is uniquely good at is meeting learners where they are. A 10 year old who is learning to read should not have to work through material that feels like it was made for preschoolers. The underlying concepts can stay the same while the language, topics, and presentation become age appropriate.

I don’t think there is one educational model that works for every child. What excites me is that AI makes it much more feasible to adapt to individual learners instead of expecting every learner to adapt to the same experience.

>> children figure out what they want to do and the learning comes along with it; the model views adults wanting children to learn something specific as generally counterproductive

> One thing AI is uniquely good at is meeting learners where they are

What if the AI wanted children to learn something specific? Able to patiently await an opportune moment. Able to blend it invisibly into other material. Able to subtly check and correct understanding.

Long term, one possibility is AI enabling a massive implicit curriculum. "[A]dults wanting children to learn" about say street crossing might be counterproductive... but funny how, at opportune times, some random stories just happen to include crossing a street, and do formative assessment, and happen to, quietly and eventually, cover and reinforce the associated learning objectives.

Take How the Piloses Evolved Skinny Noses, a children's picture book inoculation against natural selection misconceptions. It could merely be a book on a shelf. Or the AI might introduce or provide the book at an opportune moment. Or the book's approach itself might be dissolved and blended into other content.

So some story is explicitly about Fido going for a walk. But implicitly, it covers some aspect of safely crossing a street, and street lights as a communication device, and of concrete crack propagation, and tick precautions, and natural selection, and ...

Science is a richly interwoven tapestry of stories... but we basically never teach it like that. Even if such material was gathered, which pre-AI was absurdly implausibly demanding of domain expertise, it would largely be beyond the capability of an individual tutor to compellingly and adaptively deliver. But with AI?

Thanks for replying. Glad to hear you use phonics. It sounds like it could have a lot of potential and I intend to pass it along to our community. If you would like to possibly explore its use in a Sudbury community, you can send an email to my gmail account which is the same username as listed here.

I have a child who has to use a speech board to communicate. Will Ello adjust itself to work with a non-verbal child? If not, is that in the plans?

I’m curious why it has to be an either or? Spending 30 minutes with a tutor doesn’t deprive children of interaction with humans. If we can support a child’s learning (perhaps even more efficiently) doesn’t it give them more time to do that?

"That will wildly hallucinate" feels wildly hyperbolic considering how fast models are improving in their abilities, especially with proper safeguards built into the model harness.

I was making a dessert recipe last week where everything including utensils needed to stay cold, and Gemini suggested I freeze my hands in an ice bath before getting started.

It's fairly common to briefly cool your hands with cold water or in an ice bath before working on certain recipes.

It used the words "freeze your hands"

I think any reasonable person would come to the conclusion it meant for you to chill your hands in ice water.

This post is about using AI to teach 5 year olds...

Totally hear the concern. We have a few thoughts about this: 1). It's an AND not an OR for us. Ello doesn't replace parents at all. In fact, many parents use Ello WITH their kids and report that Ello has opened the doors to the kid reading more with them. 2). Not all parents have the skills to teach their child to read (or the literacy skills), there are also non-native English speakers who use our product and are grateful to have that support. 3). According to data from the National Assessment of Educational Progress (NAEP), a nationally representative assessment of student achievement in the United States, about 37% of fourth graders in the United States scored below the "proficient" level in reading in 2021. The scores are disproportionately lower for children of color and those with less resources. I would love for every child to have access to a 1:1 tutor (which runs about $100/hr), individualized education options and high quality education. But right now, we aren't supporting our kids.

It's also worth noting we aren't a full solution or a curriculum. So teaching in other ways still needs to take place. For many parents, it beats Disney plus and YouTube kids :)

First, thanks for taking the time to respond. Like I said, I do hope I'm wrong.

I will admit, this does seem like a pretty good use for computers.

I just I hope it's not part of the slippery slope to letting machines do our thinking for us. (Then again, what if they do a better job than we do!?)

I'm curious about how you detect and correct problems with the machine material? E.g. the (presumably automatic) error of substituting "phantom" for "fathom"?

Does your machine take a overconfident stance, or does it understand its own limits and help the child to know that it is not infallible?

Thanks! And thanks again for the feedback.

Great feedback! We will think about this and perhaps do some testing around this language with our audience.

Right now we work closely with parents (checking in with them each month before the books ship out) to help guide them in making the decision around book choice. We also have a form parents can send to the teachers. In the future, we will have a tighter feedback loop to support this decision making that comes directly from the way the child interacts with the app (if we have the permission to analyze that data from the parent).

Got it. Thanks for the reply.

Thanks for your comment! As of right now, We aren't a tool that diagnoses or assesses in a formal way (though we may move in that direction in the future). Research shows that the number of children who read for pleasure is the lowest it has been in decades. Additionally, many children don't read at home at all, and resist parent (and school) bids to have them read at home independently. Ello's focus is more on making reading for fun a more supported experience where the kids have an opportunity to learn in the moment (using evidence-based strategies). We support comprehension through commentary Ello makes (summarizing and commenting as the child reads) rather than testing or assessing the child's comprehension skills. This was a purposeful choice. Hyperlexia isn't something that can be "created" through reading independently (Elizabeth here, child clinical psychologist and co-founder of Ello). It is a (relatively rare) disorder where very young children decode far above grade level. In most cases, this is flagged in schools, and would be flagged for us if a parent reported a very high reading ability for their child's age (which is usually the opposite profile of our customer). However, you are correct that hyperlexia is a learning difference where there is decoding without meaning. Generally, we send books for children that are at or below grade level in reading for practice with reading fluency. We aren't a full service curriculum (and have designed our program to focus on enjoyment and time in text).

Thanks for the comment! We find that if a parent actually uses the product, and it changes their child's relationship with reading, Ello is quite sticky. Our biggest issues with churn are related to cost of the product and activation (people cancelling without ever using it). In terms of advice: 1) talk to parents about their pain points and iterate again and again, 2) be aware that a product that increases "work" rather than easing it, that causes a lot of resistance and friction.

We took a look at read along - it seems to have fallen off. I think the biggest issue with read along was the approach to supporting kids (it just gave the words rather than coaching the child through decoding them) and the quality of the "books" (the stories weren't that engaging and the illustrations were simple). The quality of the story matters a lot.

We looked at soapbox but for a number of reasons, we want to build our own speech recognition technology and have the team to do it! There are some issues with the time is takes for soapbox to run as well as the accuracy of the model. Thanks again!

Thanks for being honest, and I am sorry you feel that way. You are welcome to learn more about our privacy policy. I have helped countless families through my career, and have an incredibly deep respect for parents who seek support. I also believe that for our service to work, we have to have trust and operate ethically. I would not risk my professional ethics (or my license) for the sake of my business. Our vision is truly to make support more accessible and affordable for families. There is a need, and it's really sad when parents can't get the support they need or want, and don't always have easy access to reliable information.

I'd like to elaborate a little bit on what I'm suspicious of. Two relevant ideas are commonly encountered on Hackernews. The first: "Commoditize your compliment." (https://www.gwern.net/Complement) and the second, related, notion of "platforms."

It sounds to me like your startup wants to create a platform which captures the relationship between a family and their child development expert. In this business model your compliments are families and child development experts and by creating a low friction market between them you are naturally inclined to view them as commodities, as the platform holder. In my experience educated professionals don't like to be treated as commodities (in fact, I've never met a person who likes the idea of their labor being commoditized). Furthermore, my sense is that in the particular case of childhood development, a more personal touch is required (is perhaps the actual effective thing) in counseling children and families. In my view, the business model you have here, the fact that you think of it as a startup and that you debuted it on Hackernews, suggests that these fundamentally economic ideas form the substructure of what you are proposing.

The idea of connecting families to expert advice is great. If you really cared about that, why didn't you form a cooperative of child development experts who could invest in the appropriate technology? Such a platform, because it would be under cooperative control by the professionals in question, would resist commodification of both the families and professionals involved. In fact, a cooperative arrangement could even extend ownership to the families involved.

When I see a startup I assume the founders have dollar signs in their eyes. Years of hanging out on Hackernews have only underlined that perception. What I am getting at is that the very structure and context of this idea suggests either an ulterior motive or (to be more charitable to you) a misaligned incentive. I'm suspicious of either.