Priten: Welcome to Margin of Thought, where we make space for the questions that matter.
I'm your host, Priten, and together we'll explore questions that help us preserve what matters while navigating what's coming.
We talk a lot about technology in American classrooms, but today we'll learn about how one organization in India is responding to the latest technological innovations.
Today I'm speaking with Ratna Gil, who works at Rocket Learning, a nonprofit tackling early childhood education in India through WhatsApp.
That's right WhatsApp.
They deliver bite-size educational content to caregivers and children who might otherwise receive no early stimulation, and they're developing an AI tutor to interact with students while keeping parents involved.
Let's talk about what ethical ed tech looks like when the stakes are different and arguably higher.
Ratna : thank you so much for having me, first of all.
I'm Ratna Gill I currently support the partnerships team at Rocket Learning, which is, an ed tech nonprofit that we will have a chance to dig into.
My background is actually in child safety, so I started my career.
at GaN Trust, which is a nonprofit based out of Mumbai that works across India to prevent serious child harm.
and so I got interested in, forms of child harm that affect children in the Indian context, and in very high risk settings.
I went to policy school to think about child safety from this like, larger policy systemic lens.
And I found myself really gravitating towards classes on early childhood and so at the ed school and at the policy school content on.
early childhood and c child development.
And so that's kind of how I got thinking about these issues and in particular the, the salience of early childhood for safety.
So how, the way we interact with kids when they're really young actually serves as a protective factor throughout their lives.
that's how I got interested in Rocket.
Um, and yeah, I've been here for about two years.
Priten: Awesome.
and what do you do for Rocket?
Ratna : so I work on fundraising for Rocket, basically understanding how different departments work and how they work together to be able to talk to donors about them in a way that is, true.
Priten: the EdTech and nonprofit space is a unique one.
I think oftentimes for the casual person, those are two separate spaces like the nonprofit space and then there's the EdTech space.
what is it like working at the intersection of those two?
Ratna : I think that's one of the things that was really interesting, about Rocket to me when I joined, which is that we have half of the organization
that's like engineers and super techie people, and then you have, folks who have worked at Sesame Street career teachers, education policy people.
and it's so interesting to see how those different.
Work together.
And I think it's also, it makes for a really interesting culture because it feels very much like a startup.
It's, scrappy and fast paced like a startup.
and then it has the people culture of a nonprofit.
So I really like that, because everyone's really nice and also everyone's really efficient and smart.
Priten: Yeah.
Yeah.
Those are two good combos.
For work culture in general.
tell me what the mission for the organization is and what context does the organization work?
Ratna : so in India, 600 million children lack access to early stimulation, which means they're arriving at preschool with, 43% of them not being able to name a single letter of the alphabet.
if we think about what that means once they're in school, it's very hard to build the neuronal connections that they would need, later in life.
by that point, they're so far behind that it's very hard to make that time up.
rocket learning intervenes at that stage before school to make sure that kids are engaging in joyful play-based learning, largely with their
families and caregivers, as well as with the government funded daycare system, to make sure that once they join school, they're ready to learn.
Priten: So tell me a little bit about the government funded daycare system in India.
I, Is this nationally?
Is it the state government program?
What percentage of the students does it reach I would love to learn more about that.
Ratna : Yeah.
So India actually has the largest publicly funded free, daycare system in the world.
The reason you, I mean, you might not have heard of it.
Because Anganwadi workers, like all teachers, are very, overburdened.
They have a lot of competing priorities, including nutrition.
there's been a lot of work, by the Indian government on nutrition.
So to make sure at that age group, as we know, having food to eat is very important for brain development as well.
However, if you are a daycare worker, your job is often seen as making sure kids get that one lunch at school.
and Anganwadi workers don't have or haven't had in the past, a lot of guidance on how they can use that space.
even if it's for play, even if it's for the activities that come so naturally to us, in our context, like rhymes, word, games song, how they can use that space to foster cognitive development.
so it's a national system.
It exists in all states, and we've been working to unlock a higher budget for that system with some success.
So there's actually a lot of momentum in India right now to invest more in those systems.
however, in terms of technical know-how it's hard to roll out like a head start or a curriculum that's used across, a system, even if it's very pervasive.
So some of the work we've been doing is taking cutting edge brain science and overlaying that with an Indian curriculum that's context specific and appropriate.
and making sure that every Anganwadi worker in the country, has that at their fingertips.
rocket's core product is basically.
Bite-sized videos that go out to every Anganwadi worker that we work with.
So we work in 12 states so far, with an idea for a or a two to three minute video for an activity that they can do with their class that day.
So it's kind of taking the guesswork out of what to do.
And it's also not very time intensive.
They didn't have to make a lesson plan.
they don't have to come up with something on the spot.
all of our activities are very, context appropriate and would rely on materials that are likely already to be in the classroom.
And, We have a similar product for parents, which is a short activity they can do with their kids in the evening with something that's likely to already be around their house.
that could be, you know, take two types of lentils and have your child sort them into piles by color and size.
Yeah.
And we know that's working on motor skills, that's working on color recognition, pattern recognition, whatever our learning outcome is for that activity.
but it's really simple to execute on the parent's end.
Priten: Very, yeah.
I'd love to, start with talking about this core product.
do the, workers sign up directly through you all or do the center sign up, and enroll their workers?
Ratna : Anganwadi workers are already in WhatsApp groups for their community if the district signs onto the rocket learning product, rocket learning would enter the chat essentially.
So these are spaces that already exist.
One of rocket learning's, foundational principles is not to create groups or work that doesn't need to be there on top of what, workers is already doing.
So this is a chat she's already in with parents from the community.
parents who sign up to any of our products have an extensive consent process explaining to them what this product is, explaining how it's generated, and then the
rocket bot essentially acts as another teacher who's sending an activity and saying, Hey, if you did this today, can you drop us a message saying that you did it?
Or Can you share, what your child learned from this activity?
Priten: Tell me a little bit about how you contextualize it across India, right?
Like, are all the activities, done in the vernacular or is it predominantly in English?
what does that look like?
I.
Ratna : Yeah.
So the activities are primarily in Hindi.
but recently something that's been very exciting with AI is that it's easier to translate into regional languages.
and as you know, in India, it's not just the language that varies by state, it's cultural references.
It's like the foods that you eat, the holidays that you celebrate.
So we do part of that contextualization, manually.
So, for example, Punjab is one of the newest states that we're working in.
So we would not only translate all of the content into Punjabi, but we would also have, cultural references that are specific to Punjabi culture to make the content more relatable for children.
Priten: Do you know what models the bots are built on?
I'm assuming they're not like standard American frontier models, or am I wrong?
I.
Ratna : I, I tried to ask the tech team and they were like, it's trained on the Indian curricula.
And I was like, I don't think that's the answer that I wanted, but it's okay.
Priten: Okay.
Alright.
That's all right.
I was just, this is just me just genuinely curious because, when I try to get Like Claude or open AI to do anything specific to different cultures.
It, it struggles quite a bit.
So like when my wife and I were getting married, we were trying to like meld the two wedding rituals and.
It had no clue.
It was just like, this is all Indian.
Like it, it struggled to it coming up with like the most distinct things.
And so like Uh, I'm sure they're using something or providing a lot more context to get it to do that content specific stuff, but that's fascinating.
Ratna : Yeah, I'll find that out for you.
As a follow up, I actually asked.
Our tech team about your multilingual question, which was from the all the above podcast, which was, can you speak in Hinglish and have the AI reply in also a mix?
Yeah.
And so they said that it, and, and the third APU video that I shared with you was to show you basically when the kid is re replying in a different language from, uh, apu.
and so Apu still defaults to Apus.
default language.
But, I was told that it's if like every other word let's say is in Hindi and every other word is in English.
it's quite easy for the speech to text to translate that into English and then up upward replies in English.
. So different from what you were asking, but I can find out about the content.
Priten: It's, it's super cool.
I'm curious, pedagogically, Is that on purpose?
Like, there a pedagogical purpose to the, speakers responding in Hindi and then APU being in English, or was that just a proof of concept type video?
Ratna : Yeah, that was just a demo, um, where our chief of staff is talking to APU in Hindi because most of the children that we work with would be more comfortable in Hindi.
And often, like, you know, even if you're not super fluent in a language, you, when something's being asked, like when I was learning
Hindi, if something's being asked and I can understand the question, but I'm gonna respond in the language I'm most comfortable in.
Yeah.
So it was just.
To demo that and to see APU bringing it back.
kindly with Joy, not acknowledging that, oh, you need to speak in English or whatever.
Just, just bringing it back to English.
Priten: Right.
Very cool.
so tell me a bit more about has Rocket learning always been predominantly tech based?
Did it start as an ed tech venture And then, um, is the AI stuff day one or or are these all, things that were adopted as the technology grew?
Ratna : Yeah, so Rocket started during COVID actually when a lot of curricula and educational information were being sent via WhatsApp.
So always intended to be tech based, to have the farthest reach, because to reach, the 600 million children who aren't getting the simulation they need, tech is a really good solution in this context.
we started trying to onboard Anganwadi workers to our own platform that we designed and very quickly realized, and this is the principle I was talking about before, which is rejoice in and work with communities that are already there.
Onboarding to a new tech platform is just like the tech burden, the, skilling up process was too, involved.
So it's always been WhatsApp based.
yeah.
And Um, the idea has always been because WhatsApp is one of the most widely used apps in India because smartphone penetration is quite high in India.
Half a billion people own a smartphone, to have that educational content at their fingertips through tech.
However, all of the work that we do via tech is supplemented with in-person interactions.
So at no point is the rocket products supposed to replace in-person interactions.
In fact, all of the content we send is to facilitate a simple interaction that will happen between a caregiver and a child, whether
that's in the daycare center or at home, and just to provide tools in the right kind of educational framework to support that interaction.
Priten: Very cool.
So Anganwadi workers typically don't come from an education background, right?
You said they're charged with wellbeing, Like they're not coming from an educational context and thinking through curriculum and lesson planning.
you're providing that as an intervention.
Is that right?
Ratna : Yeah.
So workers are charged with a lot.
I, I think they're charged with learning and nutrition.
they often are women from the same community as the children who may or may not have formal education training.
However, something we found is that they have a lot of motivation to be seen as legitimate educators.
Like there's actually, there's a status difference in India between being a real teacher and an Anganwadi worker.
And I think that's partly because you're working with such young kids that societally, there's a wide held belief that what can you learn at that age, like that's the age where you're just rolling around and playing.
And so part of our narrative work has been to say it's really important to learn at that age.
And so this is an educator, not an Anganwadi worker, but an educator who's.
interfacing with your child at a time that's so crucial.
Priten: Yeah.
Very cool.
so now tell me a little bit about what's next.
So you talked about the bot that interfaces directly with the workers and the parents.
and I think we've started talking about a bot that might be interfacing directly with students.
Um, I'd love to hear more about what you're imagining there.
Ratna : Yeah, so APU is, our personalized AI tutor.
It's not been rolled out yet, so it's still in the testing phase.
So it'll go through two rounds of testing, one with.
Our staff, including our content, product and tech team.
teachers will be testing the product and seeing what needs to change.
And then we started some testing with students, supervised by rocket staff.
So, APU will be able to interact directly with students.
The idea there was that we know it's already happening, that parents give their smartphones to their kids when they.
When they need a break or when they might not be thinking about it.
And children from a very young age are using apps like YouTube, unsupervised.
So the idea is that if four 15 to 20 minutes a day, APU could be a substitute for some of those apps where we don't really know what content children might be accessing.
It's a time when they can review the concepts they learned, during the day or they can start to play games where they're learning new.
concepts around phonics, counting, et cetera.
we suggest that APU be used with parent supervision.
That's in the onboarding process, and that's in all of our guidance to parents.
Yeah.
And we've put in place guardrails so that if it's used by a child, APU cannot stray, from the learning outcome that APU is supposed to be teaching.
Priten: So you can't start talking about like the latest Bollywood movie with APU and it won't respond to the student.
Ratna : So APU would say, that's really nice.
Sounds interesting.
Let's go back to counting these apples.
Priten: when you think about the future rollout of it, are there concerns about, equity in terms of like, who gets access to the personal tutor because
the device access for the workers and the parents is one thing, but getting a device in front of the students to me feels at least like one extra step.
I understand that you're saying that the, the parents might be the ones, um, sharing their own devices with the student.
but is that, is it feel pretty comfortable that the parent device access will be one-to-one matched with student access?
Ratna : Right, so we actually have.
aPU time, as we call it, capped at 15 to 20 minutes per day.
usually that time would be, when a parent is home from work or in the evenings when maybe dinner's being cooked, some, you know, kids are playing around.
and they'd be able to use that device.
The reason for the cap is that gamified, but it doesn't really have the features of an addictive game that's really colorful and flashy and engaging.
It kind of looks like a picture book that's come to life.
Um, but we still don't want students to become overly reliant on APU, which is why we've capped that.
Uh, We are confident that in the communities, where we are working and we already know the parent is using their smartphone, they would
be able to share that in the evenings., There are still households where the primary owner of the mobile device is the father or the man.
in these contexts, when the dad is like a day laborer, they may have very long hours, and the child and the, mother or partner may only have access to that device on weekends.
In our core product, one of the things we've done is share more learning materials on the weekend when there's likely to be equitable access to that device.
for apu, we'll have to continue thinking about it.
Priten: Um, is the thought with APU to be a direct to family intervention or, or is there a role with the Anganwadi system?
Ratna : So APU is designed to be a personalized tutor for the child at home, right?
Um, I think for, for the Anganwadi system, we're actually developing an assistant called Shiksha Sathi, which is in even earlier stages than apu, but
where that's where some of the lesson planning, personalization coming up with a new activity, can be so we can provide support to the Anganwadi worker.
In the classroom to ask.
so for example, I'm teaching this word problem.
Can you give me an example That uses cricket?
'cause I know my students like that.
so that will be in the classroom and APU will be primarily at home.
Priten: Very cool.
when you think about the digital literacy of the parents who are making decisions about whether or not to use the apps, how do you, I think you mentioned earlier that a lot of the onboarding material is in the vernacular to help the parents navigate that.
are parents able to understand like what data is being collected, what privacy risks exist?
the dangers of screen time versus not screen time?
Because you all have the right intentions in mind and I don't question that.
but I'm curious as to whether the parents have the resources to make those decisions themselves.
Ratna : Yeah.
So in addition to putting the consent form in the vernacular, we're also making explainer videos to onboard parents and explain some of those concepts to them.
They'll be not dissimilar from the videos that we're sharing with parents on what activities to do with their children.
Yeah.
Um, we, of course, digital literacy remains an issue in these communities.
and so as many ways as there are to onboard parents to this information.
we are making our best efforts to do that.
Parents also have in-person interactions at our Anganwadi centers, regularly
For parents who haven't been as responsive or haven't been engaged with the learning materials for our core product,
that's a place where a Anganwadi worker might say
Ratna : like, Hey, this is really useful for your kid.
You can think about this.
so I think that's a place where we would also socialize.
of course the idea of what consent means is a really important, piece for us to socialize and we're doing that in the language they're most comfortable in.
And, the data that we're storing is.
You know, unique identifier, number, age of the child 'cause that's important for mapping learning outcomes and being able to personalize lessons for them.
So that data is stored for at max 30 days so we can analyze conversations, but there's no, There's no identity revealing information, um, , that's collected.
if consent is not given, we delete them from our database and they wouldn't receive any personalized content anymore.
Priten: There are a lot of ed tech companies, mainly for-profit ones rapidly trying to roll things out in the space, right?
So you're seeing like, Even the, the non-ED tech companies are all trying to become ed tech companies.
and just put out as much as possible in the education space as quickly as possible.
you all are obviously taking it slower.
I already identified two uh, stages and two benchmarks that you have internally.
How do you hold yourselves accountable to that?
Because that, it's hard to be in a space where everybody's moving so fast and then to say no, like, we're gonna take it appropriately slow.
Um, that's commendable.
But I'm curious, what internal conversations help keep it at that pace.
Ratna : Thank you.
So I think for everything that we do, our curricula go through a lot of layers of testing, piloting, and, iterations before they're rolled out to students, including our core product, right?
So if we roll out a tech solution really quickly and children are not learning as a result of that solution, or it's just not feasible in the environment that they're in, that is a net loss for us, right?
So We learn really interesting things about the way that our children learn by piloting and testing our products.
Um, one of the biggest things we're finding with apu, for example, in when we tested it in the field with children, is that the biggest barrier, the biggest.
place where the model is having issues is where there's a lot of background noise, But if you think about the environment where a child is using a product, like this simple speech to
text doesn't work when you have like five kids in the room when there's, when there's like pots and pans being banged, when there might be animals like barking outside or whatever.
So, and children are really soft.
They have really soft, quiet voices.
so it's only enriching to our product to test that over time, see how we can work on that piece, then bring it back and have it go through, the routes of review that it goes through with our teachers.
Again, also, everything that we do is, directly aligned to the Indian curricula for preschool, so if the learning outcomes from that aren't being met,
there's no point, that we just roll out something that's tech-based that we can get out quickly, because we don't reach the the goals that we have.
on actually moving the needle on learning.
Priten: Yeah.
when you think about the efficacy of the tool and parent incentive structures to get their students to.
Um, you know, learn as much as possible as quickly as possible.
'cause for some parents that's, that is fundamentally, how they view education is like a race to, how much, how quickly, how do you plan to justify the 15 to 20 minute
limit?
Like if, if I, APU's so effective, why
Priten: not have my kid like, do it for three hours a day?
Right.
what do you imagine those conversations might look like?
Um, as the rollout increases?
Ratna : I think because a core part of our messaging to parents has always been, even if you didn't go to school, even if you don't know how to read, you are the most transformative tool in your child's education journey.
Like the interactions they're having with you can't be replaced by anything, even if you're just playing a game, singing them a song.
so I imagine that our messaging would be something like.
20 minutes a day.
reviewing the concepts that you've worked on is wonderful, and it's lovely that your child is playing and exploring in this way independently.
If you find Abu useful, how can you continue the lesson at home?
How can you make it into a game or can you do the ABU activity in a different language?
and then recall that where we'd be rolling this out, the parent would also be onboarded to our core product, which is, All based around activities that are very tactile.
They're very, you know, in person analog.
and so if, if parents said, you know, we definitely want more, yes, there are time constraints.
but one of the things we've heard is that in addition to the richness and the, and how helpful the content is, just the reminder and the nudge every day that you get on your phone.
That's like, I should spend 15 minutes right now with my kid thinking about learning.
that's really our core message.
Priten: Yeah.
When did you all realize that the 15, 20 minutes with an AI personalized tutor, would be beneficial?
Right?
So like, Why not provide the AI coach the way you are planning on providing for the under body workers to the parent, and just have the parent do the 15 to 20 minutes.
what is the benefit of the student interacting directly with the ai, I guess is the question.
Ratna : the AI coach for the online body worker is like imagining that this person is gonna use the lesson plan right?
In the classroom, right?
so I think that's a lot to ask of a parent.
So I'm gonna give you an idea.
So this is, generated personalized for your kid.
Now you are going to do the teaching.
So I think it is to address the time constraint that the parent might have.
And, and through our core product, we're sending the parent something they can do.
Then the onus is on them to say, okay, this is what my child understood.
This is what my child didn't understand.
Um, whereas when it is responsive and personalized, APU in the moment can say, Hey, let's revisit that.
we need to go back to the letter M. So it's, it's just providing another way that children can interact with this information.
Priten: most of the information that that APU is currently, the curricular goals that you're currently building it around, are those curricular goals that the parents themselves have already met?
do the parents have the, the, the literacy skills that you're building through apu?
or is there usually a disconnect there?
Ratna : Yeah, I think it is.
what they may not have is like the pedagogical know-how to, to say, let's revisit a concept.
Like they, of course, the parent can count to a certain number or knows three words that start with that letter, but I think it's like the way that in a classroom, it would be playful.
You would revisit things.
You would remember, okay, the kid got this wrong.
But rather than kind of chastising them in the moment, we're gonna come back to it later in a way that's fun.
So I think that's what APU has that a parent wouldn't know.
Priten: That, In some ways that seems like the, the best case scenario, right?
Ratna : Right.
And the other piece on pedagogy is that.
APUS trained to be really joyous, really encouraging, really positive and affirming.
That might be something that the parent hasn't seen in the classroom that he or she grew up in.
So that's something that as a parent, you could also pick up.
we don't know how to talk to our kids in ways that we weren't spoken to often, right?
So when, when Apus, There to get the child excited about learning by, pushing them to be creative and generative and fun.
Um, that's something that might model behavior in a way for the parent.
Priten: Are there things that we haven't covered that you would wanna make sure we cover?
Ratna : so Everything at Rocket is centered around agencies.
You have the option to do the activity, you have the option to send in that you did the activity, but we don't use AI in in like a, normative way.
Uh,
Priten: how do you all track your success metrics?
how do you make sure that the stuff that you're sending out is actually being used?
Ratna : When we share an activity, we encourage parents to send back a response, saying that they did the activity with their child and sharing it could be a photo or a video of their child doing and enjoying the activity.
That is one of the things that we actually saw has led to increased engagement because like behavioral science, peer effects.
If you see that your neighbor's doing the activity, you wanna do it as well and say, oh, my kid really enjoyed it too.
what we wouldn't do is monitor parents, or use AI for surveillance or to make sure that parents are participating in these activities.
So everything at Rocket is framed around human agency and joy.
How can a parent be encouraged and themselves want to partake in this activity.
And then I think similarly for children, the thing that can't be replaced by technology is the very human, intrinsic love of learning, right?
Like, I just want to devour information on this topic.
I wanna read five books about the same thing because I'm so curious.
And that's one of the things that we are trying to build, from a young age at Rocket, because we truly believe that that relationship with learning and with education starts from the youngest age.
Priten: Yeah, that's awesome.
So, when you think about the future of rocket learning, but also just the early childhood space in India in particular, some of this feels so dystopian
to me when folks are talking about like how we can get an AI tutor in everyone's pocket, but you're really speaking to something a little bit different.
So I'm, I'm curious, like in five years is your, is your vision that a, the kids are on their phones all day interacting with this personalized AI tutor?
Like, is that success for rocket learning?
Ratna : Our vision is that every child in India, by the time they get to preschool, is ready to engage, um, with learning.
I think in India we're seeing a lot of investment in the early childhood space.
It's a really exciting time, um, with the new education policy, and budget allocations being allocated to early childhood where und Anganwadi centers are still very.
Brick and mortar analog spaces.
and that's really important, because as you say in your book, higher level skills are built on lower level ones.
So so preschools are, going to be places where we're learning, where children are learning those lower level skills and we wanna make sure that that system.
Is set up to succeed.
in a country as large as India, we see technology as a really important engine to be able to get information out, as it always has been.
to the extent that we can help resource strapped teachers and parents generate more creative, differentiated, personalized learning tools for their students, we think that that will help us arrive at our mission faster.
Priten: Very cool.
Awesome.
Well, thank you so much.
Most of the time I spend my time like talking to people who are being very, very shortsighted about things.
and then you like see mass movements in the US school systems towards The use of technology to kind of subvert the human.
So it's refreshing to hear the use of technology in ways that are actually promising and useful and not just done for the sake of being done.
So, um, I appreciate Yeah, for
Ratna : sure.
No, I, this Team is so intentional, like if the co-founders of Rocket could raise my kids, like I would just give them my kids.
Right?
Like incidentally, like we have five co-founders.
Three of them had a kid in the last year.
it's just been so interesting to like think about what you're putting into young minds for work and then watch it play out at home with your own kids.
And I think I wouldn't work with a team that wasn't really intentional and really human, uh, in that way.
And I truly think they're thinking about it in the right way.
Priten: I really appreciate retina for sharing a thoughtful and intentional use of AI in education.
Retina's work at Rocket Learning reminds us that technology and education doesn't have to just be about schools.
It can also be about teaching children who wouldn't have access to education at all.
Her emphasis on parental involvement, cultural context, and pedagogically sound design shows us what it means to put learning first, even when, especially when you're working with the most resource constrained communities.
Stay with us this season as we continue asking what and who technology and education should really be for and to explore similar questions pre-order my book Ethical Ed [email protected].
Thanks for listening to Margin of Thought.
If this episode gave you something to think about, subscribe, rate, and review us.
Also, share it with someone who might be asking similar questions.
You can find the show notes, transcripts, and my newsletter at priten.org.
Until next time, keep making space for the questions that matter.