# Central Square Foundation: Can PAL help children learn at their Own Pace

An episode of [3 Things](https://invisiblenetwork.ai/podcast/1406730235), 18 Sep 2026, 33 min.

Audio: https://dts.podtrac.com/redirect.mp3/api.spreaker.com/download/episode/75190788/csf_s_03_ep_02_v2.mp3

EdTech can play an important role in making learning more engaging, interactive and accessible for children. In government schools, access to EdTech can help bridge gaps in educational opportunities. And Personalised Adaptive Learning or PAL can provide students with targeted practice while giving teachers information about individual learning levels. As part of our ongoing collaboration with Central Square Foundation, we are excited to bring to you our second episode, where we will be talking to Harish Doraiswamy, Project Director, EdTech, Central Square Foundation, and Abhijeet Singh, Co-Chair, Education, J-PAL. They will both share their insights on how EdTech can help make classrooms more inclusive by catering to different learning needs and abilities. Hosted and Produced by Niharika Nanda Edited and Mixed by Suresh Pawar Link to our first episode with CSF: https://indianexpress.com/audio/3-things/teachers-day-central-square-foundation-beyond-training-investing-in-teachers-for-nipun/10863336/

## Transcript

[0:01] Speaker 4: Thank you.
[0:18] Speaker 1: It can make learning more engaging, interactive, and accessible. As
[0:23] Speaker 1: part of our ongoing collaboration with Central Square Foundation, we
[0:26] Speaker 1: bring to you our second episode, where we will discuss
[0:29] Speaker 1: how EdTech can help make education more personalized for children
[0:34] Speaker 1: at various levels of learning. The conversation will also explore
[0:38] Speaker 1: why this matters particularly more in government school classrooms while
[0:43] Speaker 1: highlighting examples like Andhra Pradesh where the Personalized Adaptive Learning
[0:47] Speaker 1: or PAL program.
[0:49] Speaker 2: Has been used with government school students.
[0:52] Speaker 1: For diving deep into this, we are joined by Harish Duraiswamy,
[0:55] Speaker 1: Project Director, EdTech at Central Square Foundation and Abhijit Singh, Co-Chair,
[1:01] Speaker 1: Education at J-PAL. They will both be sharing their insights
[1:05] Speaker 1: on how EdTech can help make classrooms more inclusive by
[1:09] Speaker 1: catering to different learning needs and abilities.
[1:12] Speaker 2: Harish, Abhijit, welcome to the podcast.
[1:15] Speaker 4: Thanks a lot for having us.
[1:17] Speaker 2: It's all a pleasure. So, let's begin with numbers.
[1:21] Speaker 1: ACER and the National Achievement Survey keep telling us the
[1:25] Speaker 1: same story, that a large share of children in our
[1:29] Speaker 1: schools can't do grade-level maths or reading by the time
[1:33] Speaker 1: they are in middle school. So, Abhijit, Before we get
[1:37] Speaker 1: into any of the solutions, help our listeners understand why
[1:41] Speaker 1: does this gap exist and persist, even as enrollment has
[1:45] Speaker 1: gone up and government spending on education has also grown.
[1:49] Speaker 3: So, I mean, I think we should start with, you know,
[1:51] Speaker 3: what has been a bit of success, which is that
[1:55] Speaker 3: access to education has improved dramatically. At the same time,
[1:58] Speaker 3: I think there's two or three separate factors that affect,
[2:02] Speaker 3: you know, why are we not seeing any learning gains?
[2:04] Speaker 3: One of those is that historically our education system was
[2:08] Speaker 3: really set up as a filtration system. So it was
[2:11] Speaker 3: really intended to teach at the top. It was really
[2:13] Speaker 3: intended to make you do well in class 10th, in
[2:15] Speaker 3: class 12th, go into the civil services, engineering colleges and
[2:19] Speaker 3: the like. It was not the colonial origins of.
[2:22] Speaker 4: Our education system, but not as a mass education system.
[2:25] Speaker 3: The modern legacy of that is that curricula sometimes go
[2:29] Speaker 3: substantially faster than ordinary learning happens. So children fall behind
[2:35] Speaker 3: very early in primary school. And so by the time
[2:38] Speaker 3: they're in class three, class four, they're already so far
[2:41] Speaker 3: behind that they're supposed to be learning fractions, but they
[2:45] Speaker 3: haven't figured out early addition. And the further along you
[2:49] Speaker 3: get in the education system, the distance to curriculum becomes
[2:52] Speaker 3: even larger. And the way that the education system set
[2:55] Speaker 3: up is that each grade has its own syllabus. And
[2:59] Speaker 3: that's what even a really conscientious teacher is really focusing on.
[3:03] Speaker 3: But if you're three, four grade levels behind, which is
[3:05] Speaker 3: what we've now shown in Delhi, in Rajasthan, Ratham's been
[3:09] Speaker 3: showing in multiple states over, well, now almost 20 years,
[3:14] Speaker 3: that that's a fundamental pedagogical challenge. And that's not something
[3:18] Speaker 3: that has been addressed, although there have been recent policy moves.
[3:21] Speaker 4: In that direction.
[3:23] Speaker 2: Right.
[3:23] Speaker 1: And Harish, now coming to you, walk us through it.
[3:27] Speaker 1: India has poured money into classroom technology for years, whether
[3:31] Speaker 1: it's on computer labs or digital content. So where has
[3:36] Speaker 1: it fallen short of moving the needle on actual learning?
[3:40] Speaker 5: So, I mean, for the past 15 plus years, we've
[3:45] Speaker 5: spent hundreds, sometimes thousands of crores every year, right? And
[3:50] Speaker 5: we don't have, as you rightly said, much to show
[3:52] Speaker 5: for gains. And a large part of the reason is
[3:55] Speaker 5: that our expectations out of technology, I feel, are fairly low. Basically,
[4:02] Speaker 5: procuring technology, deploying it in schools, utilizing budgets, that's seen
[4:08] Speaker 5: as an end in itself, right? It's not seen as
[4:11] Speaker 5: a means to promoting learning. Now, if you need technology
[4:15] Speaker 5: to transform learning, then the way that we set technology
[4:21] Speaker 5: up for success needs to be radically different, which is
[4:24] Speaker 5: that it should start with a learning problem. Is there
[4:28] Speaker 5: a specific learning problem that you're trying to solve through technology?
[4:32] Speaker 5: Is there evidence anywhere else in the world that it
[4:34] Speaker 5: can be solved through technology? Are there studies done in
[4:39] Speaker 5: Indian government school systems that show that those gains can
[4:42] Speaker 5: be replicated in the Indian government school system. So I
[4:45] Speaker 5: think just the way that technology deployment has to be designed,
[4:49] Speaker 5: has to be conceived, is different.
[4:52] Speaker 6: And then it comes to procurement.
[4:55] Speaker 5: Ultimately, the emphasis that the procurement effort places on hardware
[5:00] Speaker 5: is purely because hardware costs predominate. 90% of the cost
[5:05] Speaker 5: of procurement is hardware. But hardware causes zero learning, right?
[5:08] Speaker 5: That 5% or 10% of the cost that goes to
[5:10] Speaker 5: software is the one that causes learning. And the way
[5:14] Speaker 5: that software is procured is it doesn't have a solid
[5:18] Speaker 5: technical evaluation framework through which the government can actually judge
[5:23] Speaker 5: whether a particular piece of software is designed to do
[5:26] Speaker 5: the job that it claims that it does. And obviously,
[5:30] Speaker 5: it translates down to the capacity of the governments then
[5:34] Speaker 5: to design, to procure, and then to go through the
[5:39] Speaker 5: rigorous task of implementing this day in and day out.
[5:42] Speaker 5: I mean, the best technology is of no use if
[5:44] Speaker 5: it's not used in classrooms.
[5:46] Speaker 6: Or by students.
[5:48] Speaker 5: And the thing that everybody struggles with is how do
[5:51] Speaker 5: you promote its use? And one of the ways that
[5:54] Speaker 5: you promote its use is obviously monitoring and good data systems,
[5:58] Speaker 5: but also a system that lends enough confidence to teachers
[6:04] Speaker 5: through training, upskilling, and really aligning them on the purpose
[6:09] Speaker 5: of this technology, its benefits, and so on and so forth.
[6:11] Speaker 5: So I think the design of the way technology is
[6:15] Speaker 5: used and procured needs to completely transform.
[6:18] Speaker 4: Right.
[6:19] Speaker 1: And Abhijit, does the research literature back that up? Is
[6:24] Speaker 1: there a pattern globally of edtech expenditure not translating into
[6:28] Speaker 1: learning gains?
[6:29] Speaker 2: And where has it worked?
[6:31] Speaker 1: What distinguishes the kind of edtech that actually moves outcomes
[6:36] Speaker 1: from the kind that doesn't?
[6:38] Speaker 3: So I'd actually like to start with something that Harish
[6:42] Speaker 3: just mentioned. And I think one of the biggest mistakes
[6:44] Speaker 3: that is common in how people think about technology is
[6:48] Speaker 3: that they think of edtech as one thing. They think
[6:50] Speaker 3: of edtech as the input that is going into education.
[6:54] Speaker 4: Whereas the way to think about this is technology is
[6:57] Speaker 4: a delivery mechanism, right? So you need to be very.
[6:59] Speaker 3: Clear about what you're delivering and how is it different
[7:02] Speaker 3: from status quo.
[7:03] Speaker 4: So are you delivering better content?
[7:05] Speaker 3: Are you delivering something in a way that is more
[7:07] Speaker 3: engaging and therefore the student puts in more effort? Are
[7:10] Speaker 3: you putting in something that makes the teacher's life easier?
[7:13] Speaker 3: Or are you putting in something that allows for the
[7:17] Speaker 3: education system to work with more integrity? So if you
[7:20] Speaker 3: think of, for example, computerized exams, what is it that
[7:24] Speaker 3: a computerized exam is going to do? Well, is it
[7:26] Speaker 3: there to make cheating harder? Is it to reduce grading burden?
[7:30] Speaker 4: Is it to do something else? Maybe test more finely
[7:34] Speaker 4: exactly where each child is. And each of these is
[7:36] Speaker 4: a separate use case.
[7:38] Speaker 3: Similarly, when it comes to things that teach, is the
[7:42] Speaker 3: constraint that one is trying to answer really one about
[7:47] Speaker 3: the unavailability of skilled teachers or the unavailability of appropriate
[7:52] Speaker 3: curricula or something around the targeting of instruction.
[7:57] Speaker 4: To the needs of individual learners?
[7:59] Speaker 3: So in general, what seems to be true across a
[8:02] Speaker 3: very large class of interventions globally is that where we've
[8:07] Speaker 3: seen these things work are places where the use of
[8:09] Speaker 3: technology has been far more thoughtful about what is the
[8:12] Speaker 3: problem it's trying to solve.
[8:13] Speaker 4: So in particular, so I.
[8:14] Speaker 3: Guess the early 2000s and maybe even until about 2015,
[8:19] Speaker 3: the way that people thought about technology was if we
[8:20] Speaker 3: can airdrop laptops, kids will figure it out.
[8:24] Speaker 4: And that's now been tried in many, many places. Right.
[8:27] Speaker 3: So there are trials in at least five odd countries
[8:30] Speaker 3: and it's always either a zero or a negative.
[8:32] Speaker 4: Right. So in Peru, one laptop per child, it's zero
[8:35] Speaker 4: in the short run, it's zero in the long run.
[8:37] Speaker 4: And it's very expensive because you're buying a laptop for
[8:40] Speaker 4: every child. Right. And in India, laptops have been distributed
[8:43] Speaker 4: by multiple state governments at different times. Right. So these
[8:46] Speaker 4: are expensive programs. In Romania, it's actually a negative.
[8:49] Speaker 6: Why?
[8:49] Speaker 3: Because these computers were given and students increased their time
[8:53] Speaker 3: on video games and reduced their time on homework. So
[8:56] Speaker 3: it's really about what in the education production function are
[8:58] Speaker 3: you displacing? So time has its own budget. And if
[9:03] Speaker 3: you're spending time on technology, what are you not doing?
[9:07] Speaker 3: So a broad theme out of the entire literature is that,
[9:10] Speaker 3: you know, you have to be very conscious of exactly
[9:12] Speaker 3: why are you using technology? If technology is the answer,
[9:14] Speaker 3: what is the question? And you have to be very
[9:17] Speaker 3: conscious of what are you displacing? So, for example, are
[9:22] Speaker 3: you providing technology as a supplementary thing to be doing
[9:25] Speaker 3: or are you actually substituting time that, you know, the
[9:28] Speaker 3: teacher would have been working on? with the student otherwise.
[9:31] Speaker 3: So more thoughtful uses of technology have typically tended to
[9:34] Speaker 3: work better. There is global evidence on the use of
[9:38] Speaker 3: adaptive learning software. As I said, it's not the availability
[9:42] Speaker 3: of these things, it's whether they're being used, right?
[9:44] Speaker 4: So the input is, is the student working on it?
[9:47] Speaker 3: So for example, the same vendor or the same software
[9:51] Speaker 3: platform Khan Academy now has evidence from many different settings
[9:56] Speaker 3: In Brazil, it's a zero because people didn't use it.
[9:59] Speaker 3: Khan Me Go was a very, very big deal, which
[10:01] Speaker 3: is their Chattanooga bot. Again, it turned out that students
[10:05] Speaker 3: weren't really using it. In India, there's a trial in
[10:08] Speaker 3: government schools in UP where they, in the control group,
[10:13] Speaker 3: students weren't using it. They managed to figure out implementation
[10:15] Speaker 3: protocols that got usage up and they're seeing very large gains.
[10:19] Speaker 3: So it boils down to not thinking of technology as
[10:22] Speaker 3: one thing that you really drop on schools. I think
[10:25] Speaker 3: any version of programs that start that way, I would
[10:28] Speaker 3: be fairly skeptical of from the beginning.
[10:31] Speaker 1: And keeping that in mind, Harish, we just heard what
[10:34] Speaker 1: separates edtech that works from the edtech that doesn't. So
[10:39] Speaker 1: governments have already spent heavily on hardware and ICT infrastructure.
[10:45] Speaker 1: So from the CSF side, how do you actually leverage
[10:50] Speaker 1: what's already in classrooms to get the right kind of
[10:53] Speaker 1: edtech that Abhijit just described?
[10:56] Speaker 5: So I think that the starting point needs to be
[10:59] Speaker 5: what is the problem? We know what the problem is,
[11:02] Speaker 5: which is that children aren't learning at the pace that
[11:04] Speaker 5: they should be learning. And we should have a fair
[11:07] Speaker 5: idea of what is causing that deficiency in learning. Now,
[11:13] Speaker 5: one can point to a zillion reasons, but finally, a
[11:16] Speaker 5: lot of these boil down to the classroom, right? And
[11:20] Speaker 5: in the classroom, what do we see? There's hardly a
[11:22] Speaker 5: state in India, you know, as per the RTE Act,
[11:25] Speaker 5: every state, needs to have 200 days in the year,
[11:31] Speaker 5: needs to be instructional days, 220 for, I think, senior school.
[11:36] Speaker 5: There is hardly a state in this country that remotely
[11:38] Speaker 5: comes close to this number. That's because there is staffing shortages,
[11:43] Speaker 5: teachers are deployed elsewhere, there is student absenteeism, teacher absenteeism.
[11:47] Speaker 5: So all of this results in a child right from
[11:51] Speaker 5: day one. And of course, in primary schools, you have
[11:54] Speaker 5: multi-grade teaching. So a child is falling behind and the
[11:58] Speaker 5: teacher is falling behind the syllabus starting day one. And
[12:01] Speaker 5: the teacher is always in a rush to complete the syllabus.
[12:03] Speaker 5: There's not enough proper instruction time and the students fall behind.
[12:08] Speaker 5: And once they fall behind in grade one, then they
[12:10] Speaker 5: fall further behind in grade two. And this gap not
[12:13] Speaker 5: just persists, but increases over time. Now, India has spent money.
[12:18] Speaker 6: On technology.
[12:19] Speaker 5: So if we have about 1 lakh schools with computer labs,
[12:24] Speaker 5: these computer labs are not designed to be perfect. There
[12:26] Speaker 5: will be some devices there that are not working, but
[12:29] Speaker 5: we have something, right? Now, this constitutes only 25% of
[12:32] Speaker 5: the total number of middle and secondary schools in India, right?
[12:37] Speaker 5: So there is still 75% of schools where students have
[12:42] Speaker 5: no access, have not used a computer when they graduate
[12:46] Speaker 5: from class 12. So there is clearly a need, much
[12:50] Speaker 5: as we have spent money, but there's clearly a need
[12:52] Speaker 5: for us to accelerate this, especially if we want to
[12:56] Speaker 5: get to Vixit Bharat and all that. We cannot dream
[12:59] Speaker 5: of that with this level of access. Secondly, while we
[13:03] Speaker 5: need to expand access, we need to address the fact that,
[13:06] Speaker 5: you know, children are falling behind and children are also
[13:10] Speaker 5: at different learning levels. A class 8 classroom, a grade
[13:13] Speaker 5: 8 classroom will be comprised of students from grade 2
[13:17] Speaker 5: level up to grade 8 level. They've all been promoted
[13:20] Speaker 5: because of RTE, but they are not all at grade
[13:23] Speaker 5: 8 level. Now, the teacher finds it very difficult to instruct, right,
[13:26] Speaker 5: to teach. Personalized adaptive learning handles precisely this problem, right?
[13:32] Speaker 5: It meets every child where every child is. Instead of
[13:36] Speaker 5: hoping that the child is at grade 8 level and
[13:39] Speaker 5: you deliver instruction and hope that the child picks up,
[13:42] Speaker 5: you actually try and address the child to a learner
[13:45] Speaker 5: where the learner is on his or her learning journey.
[13:50] Speaker 5: And Once you do that, you provide learning material, you
[13:56] Speaker 5: diagnose first where the child is, and you pick that
[13:59] Speaker 5: child up, that learner up from that level onwards in
[14:03] Speaker 5: that session. So what this does is that It is
[14:06] Speaker 5: the equivalent of personalized tutoring at scale. Personalization is one
[14:11] Speaker 5: of the big benefits of technology. And if I personalize it,
[14:15] Speaker 5: not just to a student's learning level, but a student's
[14:18] Speaker 5: learning style, a student's learning pace, then I not only
[14:22] Speaker 5: personalize where a student starts from, but also personalize the
[14:26] Speaker 5: path and pace of that journey.
[14:29] Speaker 1: And Abhijit, can you give us some examples from your
[14:32] Speaker 1: context where you have seen PAL has worked and has
[14:36] Speaker 1: been adopted by state governments at nominal scale.
[14:40] Speaker 3: Yeah, so maybe let me start with this. So in general, actually,
[14:43] Speaker 3: the evidence on personalized adaptive learning has been pretty positive.
[14:47] Speaker 3: And going back to what we were discussing earlier, it's
[14:50] Speaker 3: because it addresses a constraint.
[14:52] Speaker 4: That we've known about in the education system for a while.
[14:55] Speaker 3: The fact that kids fall behind and that remedial instruction,
[14:59] Speaker 3: even if present, is actually too spotty to have them
[15:03] Speaker 3: catch up with where the syllabus expects them to be.
[15:05] Speaker 4: Now, I've been working on this for about 10 years now.
[15:08] Speaker 3: Initially, we looked in Delhi with students from Delhi middle
[15:12] Speaker 3: schools who were doing this after school. And we found
[15:15] Speaker 3: really large gains in after school settings. So this is
[15:17] Speaker 3: work with Kartik Muralidharan and Alejandro Kanemian on the software
[15:22] Speaker 3: called MindSpark. The first version was in MindSpark centers. Think
[15:26] Speaker 3: of these as being similar to what private tuition centers
[15:29] Speaker 3: look like, which are super common in places like Delhi.
[15:32] Speaker 3: So it's about 43% of students at that time in
[15:35] Speaker 3: middle schools were going to some version of tuition.
[15:38] Speaker 4: So we found really large gains there. And then the
[15:40] Speaker 4: bigger challenge was how.
[15:41] Speaker 3: Do you make this work in government schools and how
[15:43] Speaker 3: do you make this work in urban and rural areas
[15:46] Speaker 3: across the kind of diversity that India has. So then
[15:49] Speaker 3: we worked for three years in the state of Rajasthan
[15:53] Speaker 3: in four very different districts. So Churu, Junjunu in the north, Udaipur,
[15:56] Speaker 3: Dungarpur in the south and What we find over those
[16:00] Speaker 3: three years is that, you know, I mean, it is
[16:02] Speaker 3: possible to develop protocols where this type of program is
[16:06] Speaker 3: something that is integrated into the school timetable, where teachers
[16:10] Speaker 3: are supported to be able to use it with their students.
[16:13] Speaker 3: And what we find there is that, you know, over
[16:15] Speaker 3: about a year and a half, the productivity of instructional time,
[16:19] Speaker 3: you know, by which I mean, how much does a
[16:21] Speaker 3: child learn in, say, six months or in a year
[16:24] Speaker 3: or in a year and a half, that goes up
[16:26] Speaker 3: about 50 to 66%. So about 50 percent in math,
[16:30] Speaker 3: about 66 percent in Hindi.
[16:31] Speaker 4: And those are big results.
[16:33] Speaker 3: So, I mean, typically in industry, if you manage to
[16:35] Speaker 3: raise the productivity of something by 10 percent, that's a
[16:37] Speaker 3: winner because that's how much better you're doing for.
[16:40] Speaker 4: The same resource.
[16:42] Speaker 3: And in some ways, the binding constraint is really the
[16:45] Speaker 3: time that.
[16:46] Speaker 4: A child spends in the schooling system.
[16:48] Speaker 3: So, I mean, you need to be that time is
[16:50] Speaker 3: fixed because that time is fixed by the education system.
[16:52] Speaker 4: So whether we can improve that. So we find really
[16:55] Speaker 4: large gains there.
[16:56] Speaker 3: That's been in the context of, you know, these other
[16:59] Speaker 3: schools which go all the way from class one to
[17:02] Speaker 3: secondary school. We studied this in class one through eight.
[17:05] Speaker 3: We're finding those kind of gains both in primary levels
[17:08] Speaker 3: and middle school levels. In the short run, you know,
[17:10] Speaker 3: and those are programs that are scaling up in middle
[17:13] Speaker 3: school and in general, PAL solutions are scaling up in
[17:16] Speaker 3: middle schools in lots of different states and different state governments.
[17:20] Speaker 3: There's a separate body of questions about how do we
[17:23] Speaker 3: make this work in primary schools, given that primary schools
[17:26] Speaker 3: typically are smaller, are multi-grade, they don't have labs and
[17:30] Speaker 3: the like. But in middle schools now, it's been done
[17:32] Speaker 3: with various different software providers across many, many state governments.
[17:37] Speaker 3: So this looks very different from five years ago.
[17:40] Speaker 4: Right.
[17:40] Speaker 1: And Harish, the study found that usage mattered. More hours
[17:46] Speaker 1: on PAL led to larger gains. Given that, Can you
[17:50] Speaker 1: give our listeners a picture on how PAL was actually
[17:53] Speaker 1: implemented on the ground and what did it take to
[17:57] Speaker 1: get the students to actually keep on using it consistently
[18:01] Speaker 1: rather than it becoming just another underused lab gathering dust?
[18:06] Speaker 5: Yeah, so as both Abhijit and I have mentioned, usage
[18:10] Speaker 5: is really critical, right? And if you've got to make
[18:13] Speaker 5: a learning return on the investment that the government is making,
[18:17] Speaker 5: then assuming that you have the ability to procure a
[18:20] Speaker 5: well-designed piece of personalized adaptive learning software, assuming that governments
[18:24] Speaker 5: have that capability, then all the energies of the government
[18:28] Speaker 5: need to get focused on driving usage. Now, Andhra Pradesh
[18:33] Speaker 5: adopted personalized adaptive learning in 2020 or 2019 in a
[18:39] Speaker 5: very small scale, and then COVID shut it down, and
[18:42] Speaker 5: then they started the effort again in 2023. And that
[18:46] Speaker 5: has also been independently researched by the Development Innovation Lab
[18:51] Speaker 5: at the University of Chicago. And the findings there were
[18:54] Speaker 5: consistent with what Abhijit and Karthik found in other government
[18:59] Speaker 5: school settings. Except that in Andhra, this was a government
[19:02] Speaker 5: procurement implemented by the government. Typically, what happens is that
[19:07] Speaker 5: when nonprofits or software vendors implement these programs on their
[19:13] Speaker 5: own and research them, you know, they produce good results.
[19:16] Speaker 5: But when it's transferred to government implementation, there is a
[19:20] Speaker 5: loss of fidelity in the way that the government implements programs.
[19:24] Speaker 5: And the same software implemented by a different set of
[19:27] Speaker 5: people could produce different results. That's not because the software deteriorated.
[19:32] Speaker 5: That's because the conditions to promote usage were not maintained. Now,
[19:36] Speaker 5: what Andhra did was, firstly, Andhra was very clear that
[19:41] Speaker 5: they were willing to listen to people and they were
[19:44] Speaker 5: willing to look at evidence and they were willing to
[19:47] Speaker 5: look at other experiments. And the literature pointed to the
[19:50] Speaker 5: fact that usage is critical. So what they did was,
[19:53] Speaker 5: the first thing that they did as schools in Rajasthan
[19:57] Speaker 5: also did, was they marked out clear timetable periods. Second,
[20:02] Speaker 5: what they did was a typical computer lab in a
[20:04] Speaker 5: government school will have, say, seven or eight computers. And
[20:09] Speaker 5: you will have 40 children in class. Now, there's no
[20:12] Speaker 5: way that because, you know, the software is personalized, which
[20:15] Speaker 5: means that it is one on one. It's not designed
[20:18] Speaker 5: to be one on many with eight people watching over
[20:21] Speaker 5: while one child does, you know, the exercises. So this
[20:25] Speaker 5: is supposed to be one on one. So you need
[20:26] Speaker 5: adequate devices. So the choices that states have to make is,
[20:31] Speaker 5: do I buy tablets at one-fourth the cost, or do
[20:35] Speaker 5: I buy laptops? Now, Andhra was very clear. Andhra said, okay,
[20:39] Speaker 5: maybe tablets might get obsolete faster, but what the hell,
[20:44] Speaker 5: in the three or four years that they're going to
[20:45] Speaker 5: be in that lab, they're going to drive learning. So
[20:48] Speaker 5: they took a bold decision to procure tablets so that
[20:52] Speaker 5: there would be enough devices and there would be one-on-one usage.
[20:56] Speaker 6: So that's another thing that they did.
[20:59] Speaker 5: Usually, when government procurement is done for EdTech, once the
[21:03] Speaker 5: hardware is deployed and the software is deployed in the schools,
[21:06] Speaker 5: there's no one point of contact or a person responsible.
[21:11] Speaker 5: So the state actually, in this case, appointed a nodal officer.
[21:14] Speaker 5: And firstly, they appointed a very competent nodal officer. Secondly,
[21:18] Speaker 5: they kept that nodal officer in that same role for
[21:22] Speaker 5: a length of time. So protected him from transfers. and
[21:26] Speaker 5: so on, or resisted the urge to transfer. And that
[21:29] Speaker 5: actually gave some continuity to the program as well. The
[21:32] Speaker 5: state was very willing to look at data. They were
[21:36] Speaker 5: willing to look at the content and look at data,
[21:39] Speaker 5: appoint district nodal officers who could support the schools and
[21:43] Speaker 5: also field management staff who could troubleshoot problems that arose
[21:48] Speaker 5: in school. So I think what they had was a
[21:50] Speaker 5: kind of a 360 degree fail-safe strategy, which they implemented
[21:55] Speaker 5: day in, day out, day in, day out without getting
[21:58] Speaker 5: bored and without being attracted by the next shiny object
[22:02] Speaker 5: that comes around in the area of technology. So I
[22:05] Speaker 5: think they had a constancy of purpose and a constancy
[22:09] Speaker 5: of a cadence of execution. That was really important.
[22:13] Speaker 1: And Abhijit, coming back to evidence, for it to actually
[22:17] Speaker 1: shift how states spend their edtech budgets, rather than just
[22:21] Speaker 1: adding up to a pile of pilots, what needs to
[22:24] Speaker 1: happen next from a research and policy standpoint?
[22:28] Speaker 4: So I would actually split these two things, right?
[22:31] Speaker 3: So For policy, I think we already have a certain
[22:34] Speaker 3: amount of evidence about this class of interventions. And so,
[22:38] Speaker 3: for instance, it seems like the case for things like
[22:42] Speaker 3: smart classrooms, et cetera, is just far less established. And
[22:45] Speaker 3: there's a lot of spending on heads where it is
[22:47] Speaker 3: not obvious what is the input being delivered to whom
[22:51] Speaker 3: and why would it actually translate into learning.
[22:54] Speaker 4: So just mapping that out.
[22:56] Speaker 3: Because it's very easy to say that, you know, governments
[22:58] Speaker 3: should spend more on X, but then you have to
[23:00] Speaker 3: find somewhere else in the budget to actually find that
[23:04] Speaker 3: sum of funds. So as a class of programs, I
[23:06] Speaker 3: think for policy, there's already enough evidence to say that
[23:10] Speaker 3: there should be more money here. But also there needs
[23:13] Speaker 3: to be the kind of attention towards implementation fidelity that
[23:16] Speaker 3: Harish was just talking about. that there needs to be
[23:19] Speaker 3: responsibility in the system for who is implementing what, why
[23:22] Speaker 3: exactly would they do.
[23:23] Speaker 4: It, how would you make it happen in a system.
[23:26] Speaker 3: Where otherwise accountability relationships can sometimes be a bit weak.
[23:30] Speaker 3: So that takes an eye towards a great deal of implementation.
[23:33] Speaker 3: So that's on the policy end. On the research end,
[23:36] Speaker 3: I actually think that the single most important thing that
[23:39] Speaker 3: researchers can do is not about evaluating software A versus
[23:43] Speaker 3: software B. It is really about figuring out, together with practitioners,
[23:48] Speaker 3: together with various implementers, exactly what is the implementation model
[23:53] Speaker 3: that can be taken for implementation conditions that look like
[23:57] Speaker 3: a particular kind. So, I mean, that's been very much,
[23:59] Speaker 3: for instance, the way that I've seen the work that
[24:02] Speaker 3: I've been doing with Karthik, Muralidharan, and others. So for example,
[24:06] Speaker 3: what is it that a thoughtful PAL implementation requires in
[24:09] Speaker 3: middle schools with a lab? What is it that a
[24:12] Speaker 3: thoughtful PAL implementation would require in small classrooms that are
[24:15] Speaker 3: multi-grade and typically the school has 50 kids and two
[24:19] Speaker 3: teachers across five grades?
[24:20] Speaker 4: So how would you implement it there?
[24:22] Speaker 3: How would you implement it where it's effectively meant to
[24:25] Speaker 3: combine a certain amount of after-school usage?
[24:28] Speaker 4: For instance, replacing things that come out as homework. And
[24:32] Speaker 4: very importantly, because we are fundamentally.
[24:34] Speaker 3: A cash-constrained education system, and the most expensive thing that
[24:38] Speaker 3: happens here is the procurement of hardware. How do you
[24:40] Speaker 3: get the hardware used as much as possible during the
[24:43] Speaker 3: school day and outside? So, for example, is it possible
[24:46] Speaker 3: to have a program that is scheduled within the timetable,
[24:50] Speaker 3: within the regular timetable for the regular school, but in
[24:53] Speaker 3: holidays can be used.
[24:55] Speaker 4: To provide remedial instruction? So that's otherwise idle downtime. And
[25:00] Speaker 4: the reason this matters is that suddenly the cost effectiveness
[25:02] Speaker 4: of intervention starts looking very different if you're able to
[25:05] Speaker 4: use it much more. So the school.
[25:07] Speaker 3: Day is depending on state, you know, six hours, seven
[25:10] Speaker 3: hours of the day. if it's possible to extend it,
[25:13] Speaker 3: you know, towards school holidays, possibly after school settings. And
[25:17] Speaker 3: each of these is a tricky implementation problem.
[25:19] Speaker 4: So I think a lot of the thinking around research is.
[25:23] Speaker 3: If you had to fix some of money, what is
[25:25] Speaker 3: the most effective way of putting it in at different
[25:28] Speaker 3: stages of the education system?
[25:30] Speaker 1: Right.
[25:31] Speaker 2: And now my last question is for the both of you.
[25:34] Speaker 1: If you each had to leave our listeners, especially policymakers
[25:38] Speaker 1: and educators, with Three tangible recommendations on personalized adaptive learning.
[25:44] Speaker 2: What would they be?
[25:46] Speaker 5: I think there needs to be very clear policy that
[25:50] Speaker 5: provides a fillip to pal. You know, there are several
[25:53] Speaker 5: choices that need to be made in the deployment of technology, right?
[25:57] Speaker 5: And all of those choices need to be made right
[26:00] Speaker 5: if technology has to deliver proven results. You can't have
[26:05] Speaker 5: two or five decisions taken right and for the results
[26:08] Speaker 5: to be just as good. So the policy should actually
[26:13] Speaker 5: be very clear on A, the fact that we have
[26:16] Speaker 5: a learning problem, B, the fact that we have a
[26:20] Speaker 5: problem of learning heterogeneity, which teachers are not able to
[26:23] Speaker 5: handle in classrooms, C, that technology has a clear role
[26:27] Speaker 5: to play, 4, that personalized adaptive learning is one that delivers,
[26:33] Speaker 5: and the policy needs to make it clear. So if
[26:35] Speaker 5: the policy is going to prescribe or suggest personalized adaptive
[26:40] Speaker 5: learning but leave the choice of devices, the number of
[26:43] Speaker 5: devices to the state, right? And the state is not
[26:46] Speaker 5: maximizing the number of devices in a lab. Even well-designed
[26:49] Speaker 5: personalized adaptive learning software will not deliver its intended results.
[26:54] Speaker 5: So I think the policy should outline all the right
[26:57] Speaker 5: choices that a state could make while implementing PAL and
[27:00] Speaker 5: not just stop at suggesting that PAL could be an
[27:04] Speaker 5: effective class of intervention, right? I think the choices need
[27:07] Speaker 5: to be outlined. Second is, around data. If usage is
[27:12] Speaker 5: the most important metric that state governments can influence, then
[27:17] Speaker 5: they need to actually be able to see usage in
[27:20] Speaker 5: the most granular form. It needs to be up there
[27:24] Speaker 5: on monthly reviews. Today, EdTech is not reviewed by most states.
[27:27] Speaker 5: A small handful of states review their investments in edtech,
[27:31] Speaker 5: but many don't. So one of the things is that,
[27:34] Speaker 5: you know, now many states have this Vidya Samiksha Kendra, which.
[27:38] Speaker 6: Is really a data warehouse. A lot of information comes in.
[27:42] Speaker 5: Now, if data on usage is important, then this data
[27:46] Speaker 5: needs to come in into the Vidya Samiksha Kendra and
[27:48] Speaker 5: be used by bureaucrats, senior bureaucrats and senior officers in
[27:52] Speaker 5: terms of reviews and in terms of course correction. The
[27:56] Speaker 5: third one is, I think that Anything new and anything
[28:01] Speaker 5: which is outside the ordinary will have naysayers as well, right?
[28:07] Speaker 5: And people are not naysayers because they want to be naysayers.
[28:11] Speaker 5: People are naysayers because we haven't taken the effort to
[28:15] Speaker 5: align them and to educate them and to... We can't
[28:18] Speaker 5: just stop at 45-minute training programs, but actually take the
[28:22] Speaker 5: effort to explain what PAL is, why it works... what
[28:27] Speaker 5: the results are, and make it real for teachers and
[28:31] Speaker 5: administrators so that they can have a, if a teacher
[28:34] Speaker 5: is giving away two periods a day to take children
[28:36] Speaker 5: to a computer lab, then he or she deserves to
[28:39] Speaker 5: know more about PAL. And I think that effort needs
[28:43] Speaker 5: to be taken to really make teachers very confident in
[28:47] Speaker 5: the technology so that then they don't have any reservations
[28:50] Speaker 5: about its usage.
[28:52] Speaker 1: Right.
[28:52] Speaker 2: And Abhijit, what would you like to say?
[28:54] Speaker 3: So some of my recommendations will be complementary to what
[28:58] Speaker 3: Harish was just saying.
[28:59] Speaker 4: I think there's three tangible recommendations. The first one is resources.
[29:03] Speaker 3: So this is one of those things that really does
[29:05] Speaker 3: require resources, both in terms of financial resources.
[29:10] Speaker 4: To procure hardware, to procure software. It requires human resources.
[29:16] Speaker 4: So somebody to.
[29:17] Speaker 3: Keep track of the implementation and to keep ownership of this,
[29:21] Speaker 3: both in the education system overall and within schools. And
[29:25] Speaker 3: it requires ways to actually channel these resources to where
[29:29] Speaker 3: they're required. So, for example, as Harish mentioned earlier in
[29:33] Speaker 3: the conversation, most states don't have procedures for being able
[29:37] Speaker 3: to vet different software vendors. There are new software vendors.
[29:40] Speaker 3: vendors all the time. There are new products all the time.
[29:42] Speaker 3: So part of getting resources is to figure out how
[29:46] Speaker 3: would you spend government money responsibly while being in.
[29:50] Speaker 4: The regulations that exist for.
[29:52] Speaker 3: Good reason about how governments spend money. And that's a
[29:55] Speaker 3: big challenge. That's something for policymakers to figure out. The
[29:59] Speaker 3: second one, which is as true for educators in schools
[30:03] Speaker 3: as it is for policymakers in general, is that you
[30:06] Speaker 3: need clarity across the entire system about what the intervention
[30:10] Speaker 3: is supposed to do. And in particular, about the fact
[30:13] Speaker 3: that what you're not trying to do is replace teachers.
[30:17] Speaker 3: So all the evidence suggests that edtech works where it
[30:20] Speaker 3: is complementary to adult human supervision. So it works when
[30:24] Speaker 3: the teacher is brought in, when there is a lab
[30:27] Speaker 3: in charge making sure that students are actually spending time
[30:29] Speaker 3: on it, when their parents are providing some scrutiny for
[30:33] Speaker 3: edtech use at home. So What is it supposed to
[30:36] Speaker 3: do and who is supposed to do it and who
[30:39] Speaker 3: has ownership across it? And that requires, as Harish said,
[30:43] Speaker 3: partly repeated interactions with each.
[30:46] Speaker 4: Of these parts of the education system.
[30:49] Speaker 3: So what is the teacher supposed to do if you're
[30:51] Speaker 3: telling her that, you know, two periods are going to
[30:54] Speaker 3: be taken out of your timetable? What do you expect
[30:57] Speaker 3: her to do with the rest while still finishing the syllabus?
[31:00] Speaker 3: These programs will not work if they make people's lives
[31:03] Speaker 3: harder in schools. So programs that teachers don't like end
[31:08] Speaker 3: up dying a quick death. The last bit is be very,
[31:11] Speaker 3: very open to evidence all the way through. So now
[31:15] Speaker 3: in India, there's good evidence across multiple providers in Rajasthan,
[31:19] Speaker 3: in Andhra Pradesh, in UP. So there's good grounds to
[31:23] Speaker 3: think that this could work broadly. this is going to
[31:25] Speaker 3: vary across softwares, this is going to vary across systems.
[31:29] Speaker 3: So at each point, as governments are investing in doing
[31:32] Speaker 3: the programs, it is worthwhile to invest in the machinery
[31:36] Speaker 3: to see how the programs are doing. And that is
[31:38] Speaker 3: both on the quantitative side, so it's the user's data,
[31:41] Speaker 3: but it's also just learning metrics, And it's also true
[31:44] Speaker 3: in terms of qualitative information about what is it the
[31:47] Speaker 3: teachers are saying about it, what is it the students
[31:48] Speaker 3: are saying about it, how well does that agree with
[31:51] Speaker 3: the quantitative metrics that we're getting. Somebody in the system
[31:54] Speaker 3: overall needs to take ownership of an overview of how
[31:57] Speaker 3: each of these programs goes.
[31:59] Speaker 4: And that's actually one of the things that Andhra Pradesh
[32:01] Speaker 4: did really well. And I suspect that's actually one reason
[32:04] Speaker 4: for the very impressive gains we've seen in that context.
[32:09] Speaker 1: Thank you, Harish. And thank you, Abhijit, for joining me
[32:11] Speaker 1: today for this conversation. It was truly a very insightful one.
[32:16] Speaker 1: And hopefully going forward, we will be able to see EdTech.
[32:20] Speaker 2: Being used in the most fruitful manner possible.
[32:23] Speaker 6: Thank you so much for having us.
[32:25] Speaker 4: Thanks a lot, Harika.
[32:30] Speaker 1: You were listening to Harish Duraiswamy and Avijit Singh on
[32:33] Speaker 1: Express Podcasts. This episode was edited and mixed by Suresh
[32:37] Speaker 1: Pawar and produced by me, Niharika Nanda. You can find
[32:41] Speaker 1: the link to our first episode with CSF in the description.
[32:45] Speaker 1: If you like the show, then do subscribe to us
[32:47] Speaker 1: wherever you get your podcasts. You can also recommend the
[32:50] Speaker 1: show to someone you think will like it. Share it
[32:52] Speaker 1: with a friend or someone in your family. It's the
[32:54] Speaker 1: best way for people to get to know about us.
[32:57] Speaker 1: You can also tweet us at Express Podcasts or write
[33:00] Speaker 1: to us at podcasts at indianexpress.com.

Shows are listed from Apple Podcasts' charts and directory, and Podcast Index where it is used. Every show and episode belongs to its publisher and plays from the publisher's own feed; this site hosts none of them.