In the second episode on ACAPS' participation in the Tech to the Rescue AI Bootcamp for Changemakers (aka from ACAPS to AICAPS) Chiara, Ali and Lars Peter discuss the progress made over the past couple of weeks. Since the first episode focus has been on using design methods to come up with a clearer approach to AI. This part of the bootcamp has been facilitated by the legendary design company IDEO.

Ali, Chiara and Lars Peter agree that they have learned and that their thinking has shifted significantly over the course of the bootcamp, but they don't agree on what they want to do or whether ACAPS is in good shape or not when it comes to AI.

00:26

Alia Arbia and Chiara Rizzi, welcome back to Trumanitarian.

Thank you very much Lars Peter.

Thank you, nice to be back.

So, this is the second episode we do on our participation in the Tech to the Rescue AI Bootcamp. We are more or less halfway through, it's a seven-week bootcamp and we've done four weeks. In the last episode, we asked for feedback on...

00:49

what we did. And we got some of the feedback was you need to be more clear on what you're talking about. It sounds like an internal ACAPS talk. So maybe we can just make it clear, for example, when we spoke of Sophia last time. Chiara, just explain to us, what is Sophia?

Sophia is a tool that we have developed in-house at ACAPS to enhance our analysis and data collection capacities.

01:17

So the idea is to have a tool that does not replace our analysts and data collectors but helps them to do a better job by connecting to different sources that they can monitor and helping them streamline the information and analyze it.

I think that's exactly it. Sofía is a very clever AI that helps us quickly find the relevant parts of a report so that analysts don't have to spend so much time looking through a lot.

01:44

Yes, and it also helps us to make sure that we don't miss updated reports and also with translation. So a classic example of how AI can be extremely useful and reduce the time we spend on doing analysis without actually changing the role that humans play in the equation. So that was a great question we got from Mikkel. Thank you, Mikkel, for reaching out and complaining about our lack of clarity. We'd like to encourage any other listeners out there who feel that we are unclear, just want to give us some feedback too.

02:14

to just drop us a mail on info at trumanitarian.org. All right, Ali, last time, let's quickly recap what we spoke about last time. Last time was quite a bit of a bird's-view. I think we were looking ahead, what we think that the journey will be that we started. We did not have a lot of concrete things to talk about. I think at this point, we had developed some initial ideas and that's it.

02:44

Yeah, it was a bit of strategic high-level thinking, but I think that'll change in this episode because for the past two weeks we have been working with ideation, design of different prototypes, and that has been interesting.

Chiara, do you want to share some of your insights when it comes to what we've been doing for the past two weeks?

Yeah, so for the last couple of weeks at the bootcamp they have been led by IDEO, which is our designing company.

03:13

And they have a very fascinating approach to designing a product that is basically human centered. And I think we should say it's not just a design company, it is THE design company. They are so cool, they invented the Apple mouse thing. It's astonishing the number of products that they have created. These guys really rock, they also sound slightly like a cult sometimes, right?

03:39

So they are very, I would say they have some really robust frameworks and methods and they are really confident in them also. And I was very impressed by the wide range of sectors where they operate. They range from technology to medicine to rockets.

Yes, absolutely. Cool. So what did they help us do?

What we did was basically rethink our projects.

04:05

through the lenses of human-centered design. So we basically took a few of our ideas, honestly not focusing so much on the specific ideas, but on the process that led us there. So really interrogating ourself on why we were doing that, what is the problem we are trying to solve. We have discussed a lot also about our target. So to create personas for our use cases, to really understand

04:34

what are their problems and how we can solve them.

What is Personas?

We think quite a lot about our audience when we produce reports and we develop at some point Personas. So kind of ideal types of our readers or consumers of our reports. And I think the idea of Personas here is the same, that it's about who is going to use that invention, that project, and think about

05:04

What do they want from it and how can it be useful for them? And I think really central in the whole human-centered design approach is to empathize with your user and the way we have been working with the personas is we've done a couple of interviews with some of our colleagues and some of our users to figure out what is it actually they want.

I think that was also my biggest lightbulb moment of the last two weeks.

05:33

Because when you think about our users, you often think about the impact our reports have and the people who consume them, but we realized for this project, we also have someone who will use the product we come up with. And so it was a little bit that moment when you're looking everywhere for your sunglasses and you realize you have them on your head, because we basically never really thought about it internally. That was also for me.

06:02

The massive light bulb moment that maybe we know more about our readers than about ourselves. Because we have done quite a lot of interviews and thought a lot about who could actually use what we write and so on. We've been in that space for a while, but I don't think we have taken a critical, close and empathetic look at how we produce analysis ourselves. And that's really interesting. And this was also seen in our, during the boot camp.

06:31

where we were encouraged to think about this through the formulation of the how might we to really try to understand clearly what we are trying to do. And in the end, I think our formulation was indeed focused on how might we enable our analysts to deal with the issues that they have in analysis, so ambiguity and bad information, so that in the end the final product, so their analysis and their reports can be more impactful.

06:59

But really the key is to start from our internal capacity and have tools that can enable our analysts to do better their work. And we should say that the how might we, that's one of the ideal cult's big ideas, is that you whenever you design something, you start out with a statement that begins with how we might make ACAPS a happier place, for example, or

07:26

Something like that, where the statement is meant to guide your design efforts without really controlling them, but also without being so bland and open-ended that you can do anything. So it's a lot of art in how you shape that statement. And it's a very nice formulation because it helps us to think both about what we want to achieve, but then there is also the so-then part. So we also need to think about what's the impact of what we want to do.

07:56

And in this case, our impact is still looking toward the outside and toward our users. But I think we really realize that the key point is starting from inside and improving our own capacity. Do you have our how might we statement there? Let me check. How might we keep ACAPS analysts with the skill set to cope better with ambiguity and shitty information spaces so that they produce more impactful analysis?

08:25

Yeah, excellent. And we would like to say to all our readers that shitty information spaces is a technical term we use quite a bit in ACAPS to basically describe most of the information spaces we work with. Okay, so then we are introduced to this methodology of human centered design. We start thinking personas, we shape our how might we statement, and then we get down to business and start looking concretely at what we want to do. Ali, what do we want to do?

08:55

I think in the process so far, that was also a little bit of a challenge because we had to come up with some ideas to work through them for some of the exercises, but we only did the ideation phase now. Although the way we approached it was very helpful to kind of reset the thinking. What we still are going back to, or we went back to last week was the assistant for doing

09:24

analysis that is actually trained on our work, our reports. So an ACAPS brain, so to say, that can support analysts. Another idea that we're still working with is creative mind that can help with scenario building, that can help a group of experts also to come up with a little bit out of the box scenarios or new scenarios, scenarios we might not.

09:52

think of because of inherent biases. These are the two that spring to mind. I don't know. Do you have anything else Chiara? I see you have the list in front of you.

Well, we still have on the table some machine learning approaches to deal with information gaps, and also we are evaluating potentially tools for only internal usage for very practical things like helping us with onboarding of new people.

10:23

But I have to say that from my point of view, at least, the last couple of weeks have not been so much focused on prioritization and fine tuning of the concepts, rather on reworking the process that led us to have these concepts.

Just unpack that.

Well, we really have to think through what we are trying to achieve and what we are working at right now is to try to disentangle between what is nice to have.

10:51

and something that can really make an impact. And what they're teaching us, it's what's the thought process to go through to understand these.

Yeah, I think what a little bit of a struggle is still also that the whole concept of this bootcamp is that we have a tool, and we are looking for a purpose for the tool, how can we use it, which is probably not the ideal way.

11:20

to go about this. And there is some sense of fear of missing out, I guess.

The young people call it FOMO, Ali. Just for you to know.

Thank you.

Maybe that was TMI for you. IDK.

All right. Yeah, so we're going at it a little bit backwards, but at heart, I'm also a pragmatist. So I think with these two weeks that we had now...

11:51

It still works and it really helped now the way IDEO kind of guided us to think about those ideas. To make sure that we're not just looking for the first purpose that comes to mind for the tool we already have, but that we really think it through. So what are the risks? What do we actually want to achieve?

12:20

would be the impact, what could be unintended consequences, etc. So in that sense, this was really useful to work out a little bit more the details of those ideas that we had.

So I haven't been able to participate in all of the bootcamp meetings these weeks, but I really enjoyed the ones I was at. And it started a whole bunch of thoughts in my head. And I woke up one morning and suddenly I had this picture in my head.

12:49

If you think of it as the humanitarian weather report, if you think of it as us knowing how hot the world is right now in terms of crises and how much hotter or cooler it'll be for the next weeks or months, where is the storm brewing, where is the conflict, where is the risk of hurricanes, that's the front end. That's the...

13:17

That's the weather report. And I think of it just like the app we have on our phones. Very simple. Showing a few data points, fed by millions of data points, but boiled down to something that you can see at a glance and you know whether to bring an umbrella or not when you go out the door. That's what you see from the outside. An app on your phone, which gives you an overview of how big and how bad our crisis around the world. Then if we take the metaphor into ACAPS.

13:47

What we need are some meteorologists. That's our analysts. They sit and analyze the different weather information we get and then produce the weather report. And if you then work backwards, you need some weather stations, you need some thermometers out in the world, giving you information to the meteorologists on a dashboard about how hot is the weather right now. And for me, it was helpful thinking about it like that because what it means is we have to focus

14:17

on creating an interface and app that boils our data down to something very easily ingestible for people. We have to create a dashboard internally that brings all of the data streams we have together so that the analysts have up-to-date weather information that they can base their analysis on. And in order for us to do that, we really need to totally change our game in terms of the data streams we are connected with.

I think for me that...

14:47

That made sense as a use case that ranges from the end user, the decision maker, to our analyst, to the data collection.

I can see happiness on your faces. I'm thinking through the metaphor. So I fully agree with all the steps, what you talked about backwards, but my mind immediately jumped to the forward side of it.

People complain a lot about the weather report because it often gives you...

15:15

probability and then sometimes it still rains or sometimes it does not rain. And people then reflect through that on the weather report that, oh, it's never really true. It's never really exact. And I think there is also something. But how often do you look at the weather app on your phone? I'm not sure if I'm... Chiara, how often do you look at it?

At least daily.

Ali, how often do you look? Probably 10, 20 times a day, easily. And going into...

15:45

the forecast probabilities and seeing how it changes over the day and the radar. And yeah, I like data.

And fair to say that you spent a significant amount of time on a application you just called unreliable.

Oh, no, I did not say it was unreliable. I think, I'm not sure if people are always really equipped to deal with the data they get through the better.

16:14

There is often an element of misinterpretation, what it can say and what it actually does not say. So that's what I was trying to... No, no, I love my weather app. I have actually two on my phone.

16:30

I think there's an issue with this metaphor and is that in the case of the weather report, like the real weather report that we see on our apps about whether it's going to rain or we're going to have sunshine, the summarized information that you see is exactly the single information that people care about. Then, okay, there might be some issues with probabilities that are not fully understood, but what they see is exactly the core of the information that they want. In the case of humanitarian crisis, if we want to come up with a...

17:00

very coincide summary, like where it's hot and where it's not hot, that would be an extreme level of summarization that people are not necessarily so interested in. Then they would keep asking, why do you say that? Which aspect is driving these increasing temperature in this specific place? And this is something not so easy to tackle because then you start going deeper into details and you lose this aspect of synthesis. But in our case, I don't think it's so easy to

17:30

come up with a synthesis that would make everybody happy?

I agree that it's really difficult, but actually what I like about the metaphor is exactly what you pick up on, the simplicity. Because I think the big problem we have today is that as analysts, we love to go more and more granular level data. We really, we love the details. We zoom in and in and in. But how do you zoom out? How do you give a busy decision maker?

17:59

an overview that in 30 seconds give him or her an idea of what not to forget today.

That's really interesting. I think this actually, what you said, Chiara ties in into what I said. I think it's okay to simplify. I think this is the point where we might simplify a lot.

This is the point where we might disagree. But I think what we then need is to help the consumer of that information.

18:29

to understand not only what it says, but also what the limits are. And I think they might be interested in that information, but you need to help them also how to read it and how to consume it.

Absolutely, and you need to help them then figure out where to go next if they want more detail. But how do we provide the global humanitarian weather report that you can read in 30 seconds? That's an interesting challenge.

18:57

Well, we already have quite a lot of experience in products that try to summarize the situation at the global level, such as the informed severity index. I understand that what you're looking for, it's something different that allows also to point out what's happening today. What should I be focusing on today? I want Ali on that app 10 times a day.

I will be, don't worry.

Yeah, I don't know exactly what I'm looking for. I was trying to think about it in a way that brought together both our external.

19:27

design spaces and the internal ones. And for me, it's those three things it boils down to. How do we get the right thermometers in the world? How do we get data that brings us the right information? And how can AI help us get that raw data in? Once we are inside ACAPS, how do we create a dashboard or a cockpit where the analysts can sit and get that overview of...

19:55

create that situation awareness for the analyst that actually brings the right raw data in there. And then finally, how can we empathize enough with a busy decision maker to boil it down to something that can be ingested in 30 seconds?

Maybe another thing that I found really interesting during the bootcamp, another concept that was explored was the concept of analogous inspiration. So basically looking for the same patterns that we see.

20:21

in what we identify as our problem in other sectors and see how they're approaching them and what's the main driver that's really helping them solve this issue and if they're using any particular technological solution. I found that particularly interesting because sometimes I think we are a bit self-centered and it's good to have an outward focus.

20:46

I struggled a bit to find examples of other sectors that would have to deal with similar situations as we are. So basically lack of information, but still need to take action and to make decisions. One example I was thinking about is medicine and the situation where you have maybe the onset of an epidemic. You know nothing about what's going on, but you still need to take action.

21:15

In that case, so if we look, for example, at what happened with COVID, I think one key driver that mobilitated the resources, people, interests, everything, was the strong sense of purpose and impact of action, impact of research. If we develop a vaccine, we can stop the pandemic. So perhaps this is also something that we should explore more internally.

21:41

Because of course we want to create impactful reports, impactful analysis. We always say that. But how can we give to our analysts a better sense of the impact of the work they are doing? This is also something that I think is very important and we could explore which ways technology can help us to achieve that. I was thinking of intelligence services. You have very...

22:09

a very shitty information space to come back to the technical term. And you just need to cobble something together. And, but they then often they come up with probabilistic assessments, which is something that we do not do or not try to do to that position as they do. And the other example I was thinking of, we did a couple of years ago, we talked to insurance companies.

22:38

how they try to assess risks and how they build scenarios. But they have just much, much bigger resources also on the data collection side of things. Yes, and it's very different when you're looking at fields where you have quantitative data, where it's very easy to make probabilistic statements.

Okay, so we've spoken a bit about what we've done in the first four weeks. We have three weeks left.

23:08

What are we going to do in the last three weeks and how does this whole thing end?

The way I understand the upcoming weeks from the organizers, what they want us to do is now really nail the idea to the wall in a certain sense. So we need to work out the details of the idea, the strategy behind it, and then prepare a pitch so we can pitch it to potential...

23:38

partners who could help us with implementing it on the technical side. We've gone from two weeks of learning the basics around AI to two weeks with learning design and focusing on prototyping and so on. And now for the next three weeks, it's really about hammering out both our strategy for AI, it is coming up and fixing on the pitch, doing a killer pitch and then...

24:06

hopefully be matched with a tech company that can help us develop what we come up with.

Yes, and in particular for the next week, we are really looking forward to the sessions led by Netguru, where we should work on the prioritization of our concepts. So to really understand which are the ones that we want to fine-tune better in terms of proposal and then present to tech companies.

24:31

Great. So our pitch day, the big day where we get to show all our stuff is on 1st of August. And so we hope to come back in a couple of weeks with our final pitch and thinking the things we'll be presenting at the end of the bootcamp to see if we can attract a company to work with us on making that reality. Chiara, how do you feel we're doing?

I think we're doing OK. I think the bootcamp has been very interesting for the past couple of weeks.

24:58

It gave us a lot of insights into how to approach the thought process about a project. But at the same time, I don't think we have really moved forward in understanding what we really want out of this bootcamp. What we really want to nail down, what we want our projects to be and why we want them to be our projects. I feel like we're a bit stuck in having the same projects that we had in mind at the beginning of the bootcamp.

25:27

we didn't really change so much the approach. Well, okay, scratch that. We changed a bit the thought process about these concepts, but I don't think we really managed to make this concrete and turn this thought process into actual changes to the project.

So, Ali, what Chiara is saying is repeating past mistakes with increasing levels of confidence. Is that also how you feel?

25:57

I probably had just slightly different expectations in terms of pace. I think we did make progress. You implicitly also said that we changed the way we think about the ideas. I will also look at new ideas now in a slightly different way because I have more tools, more dimensions to think about.

26:27

The four ideas we came in with or the five ideas we came in with have not changed through that process, but I also think with those tools that could still happen when we start thinking about strategy and pitches, I think the last two weeks will be very helpful then to mold it slightly different.

26:53

At the same time, we start running out of time because we're supposed to have our list of projects, our first draft list by Monday and our list of projects completely finalized by the end of next week.

That's interesting. I'm so much more positive than you guys. I'm really, I think we are exactly where we should be. I think we're wrestling with ideas, we are rethinking, we are questioning ourselves. That's what...

27:21

innovation is and it's very messy. And so I think, I think it's been healthy and I really enjoyed thinking with you guys and seeing how your thinking is changing and I can feel my own thinking, shifting and clarifying.

No, we haven't hammered it out yet. We haven't written it down. We don't have that succinct idea yet, but it's right under the surface and it'll be there on the 1st of August.

That will be interesting to see because I agree that innovation.

27:51

is probably often confusing and a messy process. But I think the last session we had, that messy and confusing process hit the concrete wall of hard deadlines and dates that ended at 1st of August. And I think that's what Chiara was hinting at.

All right, tell me next time whether you think I was right or whether you were right. I think we will have a very different feeling when we do the next episode.

Thank you so much, guys. It's been a great conversation and I look forward to the next one.

28:21

Thank you very much Lars Peter, looking forward to the next conversation.

Thank you, it was fun.