Description
Lucie and David continue their discussion on Farmer Research Networks (FRNs), focusing on the idea of embedded scaling and its implications. They explore how scaling out, scaling up, and scaling deep each change the nature of the data and the research itself, and reflect on the challenge of designing systems where farmers collect and use data for their own benefit while also contributing to wider learning and research.
[00:00:07] Lucie: Hi, and welcome to the IDEMS podcast. My name’s Lucie Hazelgrove Planel, I’m an anthropologist and social impact scientist, and I’m here today with David Stern.
We’re gonna continue our discussion about Farmer Research Networks (FRNs). I think David, you were very excited to talk about embedded scaling.
[00:00:21] David: Yes. The embedded scaling piece of the Farmer Research Networks is something I’ve been reflecting on, feeling strongly about, for over a decade. And only just recently have I started to have more clarity on how to communicate why this is so important and why this is so challenging for the researchers.
In the previous episode, we mentioned a bit about the FRN principles, and we explained how in its original conception, FRNs were perceived, or conceived, to both enable things to be more participatory and to have larger N.
[00:01:13] Lucie: Larger N being more farmers or more data points.
[00:01:18] David: This is the key thing. I think at the time, large N was being perceived as being more farmers. But I think what I now understand – and this should have been obvious, because precision agriculture is not about having lots of farmers, it’s about having lots of data points – and so I think the important element is to recognise that large N is lots of data points, not necessarily lots of farmers.
It could be lots of farmers, but that depends on the nature of the scaling. And so, again, we didn’t have the language a decade ago to talk about scaling being up, out or deep, there’s other language which has come in in different contexts as well, which could add further value.
But even just those three, which are quite well accepted and well recognised, they’re already very useful to understand that when you “scale out”, or “broad” as some people call it, then you are working with more farmers, you are working in more contexts, of course bigger diversity of farmers. The N is the farmers in that context, the units or the farms, the units at which you are working.
Whereas the “scaling up” is about the political, so you are embedding in political structures in different ways. You are maybe influencing things, which will then change the structures at scale, which will enable scaling in a different way.
And when you are “scaling deep”, you are not necessarily working with more farmers, but you are working more deeply, or those farmers are thinking about these things in deeper ways. It’s not just what was discussed originally, it’s having implications across more elements of what they do and how they work.
[00:03:05] Lucie: Wait, when we talk about scaling up too, we usually talk about politicians and sort of, well, political deciders in French. But it’s also about funders.
[00:03:12] David: Absolutely. And institutions, organisations. You know, you are affecting societal structure, whatever that may look like. That could be organisational structure, that could be political structure, that could be funders. Scaling up is really about the fact that you are not just doing more of the same, you are changing the structures.
[00:03:33] Lucie: Yeah.
[00:03:33] David: And again, how that relates to data is unclear, but almost always those structures you change lead to different types of data coming in, which have a different nature and a different shape, a different essence to them, one could say.
Scaling deep as well, if you are wanting to do this with the research embedded, again, you’re needing a very different kind of data. For example, in certain contexts where research used to happen at the plot level, and maybe now you need farm level data, or maybe you need landscape level data. So scaling deep within the same community might go from plot level data to farm level data or to landscape level data.
And so that would be an element of how the data changes as you go deeper because the questions you are asking are different in nature and therefore the research on that is different in nature.
[00:04:26] Lucie: But also knowing more about what motivates all of the different decisions that the farmer has made, just in that plot, perhaps.
[00:04:33] David: Yes. Absolutely. There’s so many ways that this could actually take shape. But if we think of that as being what scaling deep is able to do in terms of how it changes the data, what I think is really important is to think, well, what does that do for the research? And more generally, what does embedded scaling actually mean for the research and how do we interpret this?
There will be different opinions on this. My view is that embedded scaling is about embedding research into whatever scaling is happening, be it scaling out or broad, be it scaling deep, or be it scaling up. And what does it mean to embed research in that way?
Well, actually there’s not many good research methods for this. There are people who talk about embedded randomised control trials where you have something which is happening at scale and you embed a research project in that. But that research is often on top of not embedded in the actual scaling process.
And so this idea of embedding research into a scaling process is actually one which really stretches the research methods, and what our role is, in a way which I find extremely challenging intellectually, but even more so practically because there’s a beautiful concept here.
But if we think about the data side of it, this is going into a territory which is sort of unknown. Your standard research is based on statistics. You design a study and you use the design to be able to set yourself up to take advantage of scarce data. But that’s not what we are anymore.
If we are embedding in scaling, data scarcity isn’t the essence. So it isn’t about a rigorous design in the same way to enable us to use a small amount of data to make conclusions about a population. Embedded research is working more like data science. What does data science do? Data science is really born out of data abundance. It’s where you are getting data, which is passively collected and you are using it to be able to make hypotheses.
You can’t use it to determine causation in the same way that experiments can, but you can use it to hypothesise connections. And there’ve been a lot of very successful data science outcomes, you know, machine learning is born out of data science, and so all the AI which is coming up is broadly coming out of that area.
[00:07:46] Lucie: Yeah.
[00:07:47] David: But a lot of this, the research which is happening into data science… You know, when you go to data science events around the world, they all talk about what they would consider “big data”. And you’re never going to get, or you are rarely going to get, to what they would consider big data in these sorts of cases.
The “big data” is data on the scale that your big techs are working with at the moment. And that’s not really what we’re talking about here. We’re talking about data which is abundant, but not big data. That middle ground is actually not that studied.
What could this look like? What if you take the case of Fuma, where they went from farm experimentation on a few plots, to farm experimentation on thousands of farmers’ fields with a lot of choice. Now, there you’ve got an element of experimentation happening at scale. They’re not the only people doing this, as you know, we’ve worked with Tricots who are another group who have done a form of large scale farm experimentation.
[00:08:57] Lucie: Yeah.
[00:08:57] David: And so there are different groups who are looking into that. And there are certain things that you can build in, in different ways, so that the outcomes can be highly scientific.
The interesting point, which is critical to all this is, well, what data is flowing in? So the other thing which is happening is that you have groups like LiteFarm, which is a farm management app to help farmers manage their farm, designed by researchers who are interested in understanding and studying the farms and how people manage their farms.
So this is now an interesting collaboration. The app is designed first and foremost to help the farmers manage their farm, but it is designed in such a way that the research can be embedded. I really have a lot of respect for the researchers behind this, who are thinking about this as I would consider embedded research. So you are embedding the research into a scaling process.
So for them, the scaling process is if lots of people start using their app to manage the farms. Now that’s something where I think there’s still work to be done there, it’s an open source app, and it’s very exciting where it’s got to, but to get it out at scale in the context we work, there’s more work, and that’s fine.
But the concept is there that if lots of farmers use this app just to help them manage their farm, well, then there could be interesting research which happens, where the farmers could choose to share the data to be part of a research study. And the research would then be embedded in the way that the farm is being run and the data is just stored from the running of the farm.
So this is an instance which could be embedded research. And actually this is what, over a decade ago, we were discussing: if we are going to do FRN, how are we going to get the data so that the farmers want to collect the data, not for the research, but for their own benefit? That’s the heart of it.
I believe this is one of the hardest problems to solve. Some of the research on this, which has happened within CRFS, has stated that the farmers are not interested in the data that is collected. And I think that was an obvious answer because the data that was collected was not the data that would be useful to the farmers, it was useful to the researchers.
And so that changing of mindset is what we haven’t done enough work on. To really say, okay, how do we build these systems that would be useful to the farmers, which would also have some information that would be of use to researchers.
[00:11:46] Lucie: Yeah, it’s a difficult balance to achieve, I guess.
[00:11:50] David: Yeah, it’s hard.
[00:11:51] Lucie: And some of the researchers we work with in West Africa, and I call them all researchers, be they researchers or farmer organisations, they are thinking about it, and trying to find that middle pathway perhaps.
[00:12:03] David: Absolutely, and that effort to try and find this is really what is so exciting to me. And it is also so interesting that we’ve made so little progress over the last decade. I mean, it’s not that we haven’t made progress, it’s that because the embedded research was not an FRN principle there hasn’t been a coherent effort to think this through. It is now coming back because of AI and what’s happening with AI, and it might be that actually it was simply ahead of its time a decade ago, that it wasn’t realistic then.
There are ways now that I believe we could be getting data in different formats, which could be doing this, and it could be possible that Digital Green, who we’ve had episodes with in the past, these are things which are ready to, you know…
[00:12:55] Lucie: Expand and be adopted and adapted.
[00:12:57] David: Exactly, yes. Well, no, it’s more than that. They’re reaching over a million farmers in different places and they’re getting data in from them, which is interesting. It doesn’t have research embedded yet, but it could. And we could imagine how that could evolve, and that’s exactly where they’re just about to start a grant with McKnight. And my hope is that one of the things that will come out of that is exactly this question of how do we actually embed research into some of the things that they’re doing. How do we put farmer experimentation as part of what they encourage, and help farmers to do, where they could be doing that and then getting interesting data back and this could be used then as part of these processes?
They’re aiming to have essentially an AI enabled extension agent enhancing what extension agents can do through AI. But as part of that, that could include encouraging the farmers to set up research, mini, simplified research trials, and feed that information back in, so that we actually get this larger scale farm experimentation.
I’m not saying that’s what will happen, I’m saying that’s something that could happen. And these are the exciting things and it does seem that now is a better time to be thinking about these things than it was a decade ago. Because, actually coming back to an example of the AE Hub and the work that we’ve been doing with Beth Medvecky in this, where we have failed to do embedded scaling despite wanting to for a decade.
I think a lot of that is a recognition that we couldn’t get over the hump of getting data which was collected, which the farmers wanted, that was used for themselves. We’ve tried a few different things, my favourite failed effort was to say, well, the farmers themselves aren’t gonna do this, but their kids are at school. Kids got involved in collecting the data and were doing this in an interesting and useful way. And so we used school kids to get involved in that. And that was a lot of work, and the quality of the data was not what we would have hoped.
It was a nice idea and it may not go away, but doing it the way we did required a huge amount of effort from different people. It was all tied in with the competency-based curriculum. But at that point, the competency-based curriculum was only in early primary. So maybe again, we did that too soon, the competency based curriculum has just hit senior secondary, maybe senior secondary students would be a better choice.
[00:15:37] Lucie: Yeah, I remember some of the team struggling because the students didn’t know how to write basically, their literacy levels just weren’t as high as needed.
[00:15:47] David: Yeah again, it might be that it wasn’t yet of its time and now that the competency-based curriculum has gotten further along we could take older students. Now might be another time to retry this. So these ideas are almost certainly not going away, and this relates then to the MSc programme, where we could try and actually get some of these things into the teacher training. Anyway..
[00:16:09] Lucie: And what do you think the importance of technology within that is, like digital technology? Initially you were mentioning AI and things, but I know in this Kenya example, it was handwritten sheets of paper.
[00:16:21] David: I think at least a mobile phone is unavoidable as a means of enabling this. I think that to be able to get the sort of data at scale, which is useful to farmers, it can’t just live on pen and paper because once on pen and paper, it can’t be communicated easily. I think it has to be technology, and digital technology has to be central to this.
And again, the penetration rates in Kenya have continued to increase, and so it’s something which is more realistic now than it was a decade ago. So I guess I’m reflecting on this, if that the route that we’ve taken is not a bad route because there’s a lot of learning which has happened. But I really hope that, as we move forward, embedded scaling starts to be thought of more seriously because it changes and challenges the methods, the nature of the research, how we get data which is of a high quality, how we get data which is, first and foremost, of use to the actual people providing the data, and so it’s their data, how we change data ownership.
These are such hard, each one of these is a hard problem. And yet to think about embedded scaling, we kind of need to think about all of them.
[00:17:39] Lucie: Yeah, and it’s interesting in terms of those, it really takes the idea beyond just the participatory action research that we were talking about in the previous episode too, there’s a whole lot more ideas that come into it.
[00:17:52] David: This was the original concept, which got me so excited over a decade ago. And my excitement for that has not waned. Now, what should we call this? Is this an FRN with embedded scaling, so that we don’t disrupt FRNs as they currently are? But I do wish it’s something we could be thinking about deeply because it’s hard, and I believe it is useful.
[00:18:15] Lucie: Yes. Yeah. The aim of actually having, or enabling farmers to get access to their own data, to monitor their own farms, to improve their own farming, that’s ideal.
[00:18:27] David: Well, and it’s something which exists in high resource environments.
[00:18:32] Lucie: Yep.
[00:18:33] David: Could this exist and be of value in low resource environments? We don’t know because we’ve not put in the hard work that’s needed to find out. This idea of embedded scaling being part of an FRN process and that defining something which is potentially new and more well-defined, I think is something where it will only really be of value if groups outside of CRFS then feel, yes, this is something we need to try out, we need to actually think about this, we need to get involved in this.
And this is exactly where CRFS as a programme now is. It recognises its desire to influence beyond just the communities it’s working within. So I really feel this idea is of its time. And we shall see, let’s see what happens over the next months and years.
[00:19:29] Lucie: Yes. I was gonna say, yeah.
[00:19:31] David: Well, these things take time. It’s taken a decade to get to where we are, it’s not unreasonable that things will take a decade to evolve into what they become. And then it’ll evolve further. And that’s all part of what research is and how these processes work. But I am getting a little impatient. I’d love to really dig into and work on embedded scaling.
[00:19:53] Lucie: Okay. Well, thank you very much, David. Do you have final thoughts?
[00:19:56] David: No, I think something like: “watch this space, I hope learnings will come”. Well, maybe a final thought is that this is not unrelated to our upcoming Float workshop. We’ve had an episode quite recently, which discussed the preparation of that workshop and we will have another one which will discuss the outcomes in a few weeks time.
[00:20:20] Lucie: Yeah.
[00:20:20] David: So looking forward, there is progress that could be happening, on a relatively short time scale.
[00:20:26] Lucie: Yeah. Watch the space.
[00:20:27] David: Thank you.