Systems Mapping Meets AI

Great. And and yeah, so but before we get into any of that, I would like to start with a quick warm-up. So, I will drop a link to the mirror board here in the chat. Maybe this has already been done, but this will take you directly to this warm-up section. Um and so, if you just scroll down, Fabian, to warm-up section at the bottom. Brilliant. I'll I'll help to admit people in the process. Great. Um so, here you'll see two questions. The first one is are you more of an AI optimist or an AI pessimist or are you undecided, which is also a very very valid position to be. And also, just to admit one thing, personally, depending on the day and sometimes the hour of the day, I fluctuate between each of these three in quite extreme ways as somebody who works in this space. So, you know, no pressure. It's just whatever you feel right at this moment. Um and then the second question below it is have you already applied AI in a systems mapping context? So, thumbs up or thumbs down. Um yeah, no wrong answer, obviously. Uh and and this will really help us to kind of sense where we're at, you know, as as kind of participants in the in this room. Um a lot of you have figured this out already, but you've got this dot voting kind of tile in the middle. If you just click on one of those any color and then just drag it. That that's how you can reflect your answer on the board. So, maybe we'll take another another couple of minutes just to see where we where we stand uh with this one. If anybody has any issues with the Miro or the link or anything, feel free to just submit a message in the chat and see we'll see if we if we can help out. Okay, so initial impression, quite a lot and not decided yet. That's That's totally understandable. We have almost a draw between optimists and pessimists so far. That's uh you know, balanced. And quite a lot of people, yeah, the majority who have applied AI in a in a systems mapping context. That's really good to know. We'll be Yeah, also very interesting to see like how people have been, you know, what tools and what approaches and and how people have found that as well in the in the breakout sessions and the show and later. Okay, so I think in the interest of time, maybe we can just uh you know, take a take take take a moment to reflect on this. So, not decided yet seems to be the majority, and then between the optimistic uh versus pessimistic was slightly more optimistic as as a room, which is good, I guess. Um have you already applied AI in a systems mapping context? Majority definitely have. Uh so, yes, take take take that as as you will, uh Fabian. Um and yeah, on that note, I think I will hand over to you to take us through the five roles that AI can play in systems mapping. >> Great. Thank you for this introduction and uh also good evening uh good afternoon uh good morning, depending on where you are. Um yeah, I'm happy to be here and give you a bit of an overview about what uh we were experimenting with uh in the system mapping academy around um system mapping and AI. And I have a brought with me a presentation, which is basically um yeah, just give you some inspiration on how we applied it. So, these all what you see in the presentation is uh is experiments, I would say. So, the ones are working with AI already know that the developments are so fast and every day there's something new that's changing. So, I consider all the results we gained as a kind of draft or sketches or experiments for the moment. So, there is not a final perfect tool out there which um uh helps us system mapping in a perfect way. But, I think there are a couple of approaches which might be interesting to you um how we have worked with AI till now. And that's um yeah, the purpose of this presentation. And I think it will be around half an hour or 40 minutes at the longest. And I have brought with me some videos and some uh yeah, inspirations, hopefully. Uh just shortly about the system mapping academy um in general. So, we are quite small team of two to three uh consultants which uh focus on design and system thinking. So, we I'm a designer from background. I worked a lot in the innovation field, but somehow then came across the system system world, let's say. And there I also met uh Jian, my co-founder. And he's coming from system modeling and system dynamics and so we joined forces um to make it uh a bit more approachable how system mapping can be practiced and also how system thinking is taught. So, we are doing a lot of trainings and development of tools. And just that you get an idea on what we are working on um I think also some of the uh topics are also there in our examples. so it's quite a various field of application of system mapping from sustainable mutual development, for example, in Peru we worked in the team there. Um city development and health, um also webbing about youth and exploring um this uh smaller context of cities, um or governance innovation, where we work with the German government on uh innovation transfer and uh the general German innovation system. And looking uh on the topics which I would like to go through today with you, um we as a result of our research and our experience in the last year, I would say, around AI and system mapping, we defined five roles for now. So, also here I would say, these five roles are the five roles we found or found useful. Maybe there are more out there, so this would also be something I would be really curious about to hear from you later on. Um so, we have these five roles and we give you an overview about them and uh some examples and key insights, then a bit about benefits and risks what we saw when we're using AI and system mapping. And as already introduced, then we'll work together on a Miro board or discuss in breakouts, but then we have time for questions and [clears throat] the closing. For some housekeeping, I would say, in the beginning, uh just on the definitions of what do we mean with system mapping and what do we mean with AI because these are two quite broad terms. So, if we are talking about system mapping, um from our system mapping academy perspective, we focus a lot on causal loop diagrams. So, um uh on one system map, which you can also see on the right, for example. So, it's about causal relationships, uh feedback loops, and so on. Um I just say this because there are so much mapping methodologies out there, but this is kind of the core um method we are using, and the experiments go in this direction. Uh which you see. Then, what do we mean with AI? And in that sense, um the ma- majority of the tools we use focus on the generative AI, or belong to the generative AI bucket, I would say. So, it's all about creating content like text, images, video, audio, and so on. And um yeah, also using the large language models out there, which are familiar to everyone, I think, like ChatGPT, and so on. So, yeah. And so on and so forth. Um some of these tools may be on the edge of other AI approaches or AI technologies, uh but that's kind of the main um area where we are talking about today. So, then without further ado, let's start into the five roles of system mapping. And actually, before we start here, the question about why are there these five roles? So, one learning for us was that AI can play a different various roles in the system mapping process, and not only one, like uh most of us would imagine we are we are creating system maps with um with AI or these tools, but there are also a lot of more applications before useful. And so, we really came across this notion that uh it depends on which where you are in the process and uh so different tools and different approaches apply. So, yeah, it's more it's not a general purpose tool we will work we are working with. It's more like uh different uh tools which come in depending on where we are. And these five roles um we call them for now uh the knowledge aggregator and researcher. So, the beginning uh where it's mostly about uh bringing informations together and analyzing them. Then the kickstarter and mapper which is all about visualization, critical thinking companion which is about critical uh critical having critical conversations, let's say. The analyst which then looks at leverage points and so on and the storyteller which is about communication. Mhm. >> [clears throat] >> And yeah, so I will now go through all these five roles and give you uh an overview on example what we did here. And you also can map these roles along the system mapping process. So, this is the process we developed which you also the toolkit um which we developed. Um this is only one approach to system mapping. Of course, these processes are also can can look different but for us it's all about framing, exploring um the the system and mapping it out, reflecting on it, and identifying leverage. And there you see kind of the way you can place the different kind of roles in the process. And the underlying principle what was really important for us that we realize that there are some there are of course benefits to using AI but there are also kind of um danger or developments which are maybe not really helpful in this uh thinking world. So um our guiding principle in the end was that AI generated outputs should always be treated as drafts. Like a sketches as a starting point for human judgment and not as a final result. Uh because yeah we also fall into this trap to then go with the result and not checking it and not um uh having the human in the loop. So I think um yeah this was really one important uh thing which maybe is also true for all usage of AI generated outputs. Um but I think especially in system thinking um this is really true. But I come to that uh in the end also. So let's get started. Knowledge aggregator and researcher. Um as already introduced this phase in the beginning is all about collecting data, collecting insights, um analyzing these insights to build a good ground for the system map. And [snorts] here this knowledge aggregation is really a topic where we think AI can help a lot. Um in searching through a vast majority of really big chunk of research articles out there and also managing these uh these topics. And here as an example um we have used Notebook LM which is um based on Gemini as far as I know. So the the model of for the LM Gemini. And this is a quite interesting tool. There are also more tools like this out there, but um this um example is quite interesting because it gives you the opportunity to collect all your research and bring it into one place and then um search for different with different questions and writing prompts about only this research because in the end it's really crucial as we don't know um where the content is coming from if we're only asking uh LLMs to search, let's say in the web, for answers or for um data and to build system apps. This is really interesting because here you can really control um the uh data resources and search them in a really uh practical way. And yeah, like I said, I have brought with me some videos. Um let's just check if the audio is good and this is um video about interesting aspect of um uh Notebook LM which gives you the opportunity to also create podcasts and interact with the speakers um in the podcast so you can have conversations about the topic um your research is based on. So let's have a look. >> On one hand, you have these undeniable, sometimes remarkable uh opportunities for learning, for play, for you know, vital social connection. But the forward in one of the key reports makes it so clear. This digital acceleration, it comes with these pressing, evolving risks that impact everything we're talking about. The development >> Oh wait. So what was the join? Hey, go for it. >> I wonder if the parents' role also is covered in the data. >> That is a fantastic question and honestly, it is the heartbeat of the research we found. >> It absolutely is. Parents aren't just background characters here. They are central to how this data is shaped. >> We're seeing that parental education and even their jobs are used as major indicators for a child's experience. >> Right. Researchers use parental socioeconomic status to predict gaps in how kids use technology and the risks they face. >> And here's where it gets really real for any parent listening. Many feel they have it under control. >> But the data shows a gap between what parents think is happening and what is actually happening. >> So, um I think this example is quite interesting. Also comes from a project which we did around digital technology and children. And also this topic is of course super super big topic in terms of which influences there are. And so this helped us to also at the beginning get into the topic. Have a look at all the data. I can show it here. So, on the left side you see all the studies which we have imported around the topic of technology and children. And this podcast which you just heard is the base on this data. And so I think even though it's of course simplifying and we don't know how the algorithm is processing it. Um I think it's quite an interesting way of get into a topic and get familiar with it. You can also use it to derive variables for example in here in notebook LM already and scan them and then also have them tracked uh to the different kind of studies you have inserted. So, really interesting tool um for the beginning of a phase of system mapping. Then let's move to the next one, which is uh the Kickstarter and mapper. Uh kickstarting, uh we came up with this term because uh here it's also interesting, you become uh or you get a final system map and a lot of tools. So, if you use Claude or um uh ChatGPT or so on, you can also create final system maps and say, "Okay, let me uh create a system map for this or for that." And we tended to say, "Ah, okay, this system map looks okay already and maybe we can use it." But, um after reflecting on it, um for us this is really more like a kickstart. So, these tools which are creating system maps for us, um like also introduced in the beginning, are more like an initial draft, which then actors or stakeholders can work with. So, we use these tools, um which are generating maps, as a starting point or maybe as a or workshop as to find the most important areas in a system and go from there. But, we would rather not use it to create final system maps. And here, um yeah, we have different tools used. The example is from uh Systemic, which is a tool um which is um uh based in Miro, so you can uh add it to Miro. I don't know if everyone is familiar with Miro, but you have just seen it on our uh Miro board, so I think a lot of you already working with the tool um in your in your work. So, um this is kind of an add-on, which you can install in Miro and use it directly there. And this is uh really great, especially for us because as a system thinking academy we work a lot with Miro and also have a toolkit there. But also this is one of the I think the only tool which is really giving you results within a collaborative environment which you then directly can use. And so this was really interesting and [clears throat] here is a an example of that. So if you have installed the application here for example you have inserted a question, really simple question of course. So the prompt would be much more sophisticated but for for this example we just put simple one. So this was about what are the main drivers of deforestation in the Peruvian Amazon. Based on our example we have here. And what is also here quite crucial that it's possible to upload your data. So you see it on the right side. I just uploaded a PDF which is around this topic. And then after some time and restructuring which is of course just spend speed it up here. You get this you get this map which is a really good starting point for further exploration and to discuss it also with other stakeholders. Or just to give yourself a overview about where the topics which might be the topics which are really interesting. So we see here topics like migration which might be surprising that deforestation has a connection to this or it's you also see that it's about illegal logging and so on. So we definitely have found quite interesting topics already. Of course, it's at this stage a quite simple map. So, there's definitely maybe you could add more details to it. And this is also possible. I can also show it to you quickly. I have this example also on Miro board. Um yeah, so the nice thing about this is also that you can select it and also extend this and maybe can say I want to give me more details about how illegal logging is affecting the deforestation and so on. So, you can then work also with the map you already have and then extend it and so on. And you get informations about the variables. Yeah, so there's a it's a simple tool but really powerful um when it comes to getting started. And if you have any questions regarding this tool, Bowen actually uh is the developer and the creator of this tool. So, it's rooted in the community of system innovation. So, if you have any questions around this, don't hesitate to contact him. Uh yeah, really great that we have that now and also curious what yeah, this then also in the future. All right, now we have created our system map. And now it's the it's all at some point always in the process comes this moment where the system map is kind of finished or where we say okay, looks quite sufficient. Um but still the question is how to now get sure and sure that we don't miss something or maybe also get stuck with our biases and our perspective in the process. And here the critical thinking companion is a quite interesting role which could be taken by AI which is more about creating or uh prompting in this in this case Claude in a way that it's conversation and that can have a conversation about a system map and give you some tips and tricks what you maybe missed and also ask particular questions about it. And yeah, so it's actually quite interesting if you are working with this I don't know during the day and then asking and having conversation with Claude um it really feels interesting to to have this companion there and we experience quite good results at some cases also some uh weird results in the middle so it's also again about reflecting on what the AI gives to you of course but um overall um definitely sparked new directions for us and here's an example we uploaded um the system map here on the left which is about fast fashion and if you're looking at the map I've uploaded do you see any stakeholders missing in the current visualization? >> Right. So beyond the garment workers we already talked through one group I don't see anywhere is the second hand receiving countries places like Ghana or Chile that absorb exported textile waste. The map treats waste of textiles in landfills as if it all stays local. Should we bring those downstream communities in? >> Yeah, this might be a good idea. What would be your suggestion on that? >> Here's what I'd suggest. Add a variable like textile waste exported to the global south sitting between amount of discarded clothes and the landfill side of the map. Right now, the map assumes waste stays local, which quietly hides where it actually ends up. But, here's the question that matters for your map's logic. Does adding those receiving countries just document the harm like an externality? Or, do you think there's a feedback loop there where say one of these countries bans imports, and that pressure bounces back to constrain discarding? Which story are you trying to tell? >> So, you see here, what I found quite impressive is that we didn't give any big context or anything to bought in this case. We just uploaded um this picture, which was actually a picture, so no um text data in there. Um data data or something like this. And it refers to concrete variables in the picture, which are already there, and it suggests which uh makes sense, at least on the on the sur- so, uh this was quite impressive um to see, and I think even maybe if it's not perfect uh in the sense that it um yeah, maybe it's not in the detail understanding what is happening in sometimes. Uh really can give inspiration and um new ways of thinking. So, that's definitely uh an interesting role we also will uh implement in our future work to get some inspiration and also get validation. And another one in this kind of critical um critical thinking companion. Another example which uh goes also in the direction of um stakeholders perspectives. Um here we asked her to uh create kind of personas also based on our on the same maps. We gave it the map. And then we asked if we are missing anyone and uh if you if anyone who we are missing, so which perspective is not represented? Kind of the same question, but in the end we asked to create personas um which um speak about the perspective of the stakeholders in this uh system. And also here it's interesting that uh it really realized who is important here and for the one on the one side it was about marketing, so there's a brand executive, there is the customer which buys um the fashion uh and the clothes in the end. And there's a second-hand trader. And also admits kind of the uh yeah, like quotes uh which are really interesting and uh also some challenges and even biography a bit. Of course, again, these results um might be quite biased on the data which is used and also on the data available. So also we have to be careful with especially with the simplification of personas and so on. Um but again, it's quite good to see what uh else what yeah, else could be represented in the in the map. And also these personas could help to communicate um some stories of the map in the end. So [clears throat] again, yeah, uh taken with some caution, but um a really nice outcome here. Uh once more note, the pictures I added later on, so these are just stock photos. Then our first, uh, the fourth analyst's role, which we want to show you here. Um, this is all about analyzing the map and finding leverage in the end. And maybe even brainstorming interventions. So, we have a the foundation of the data, we have created the map, we have we have got some critical feedback on it. Now it's about time to, uh, yeah, think about what we what could be interventions and leverage points. And this is what the where these role comes in. And here again, we are still in the same example of the fast fashion. We basically ask for um to find us some leverage points to use in the map and also uh, where things we could intervene and gives this back to us in a visual manner in that sense, uh, here um uh, graphic. So, the one, two, three bubbles are added um on cloud. And also again, I find impressive um these are more obvious kind of leverage points, but uh, still interesting uh, points here. And if we want to push this further, we could even ask to create uh, dynamic models out of it. Um, just um to show it to you and and really uh, not really on a high level. So, we again we have the the map here inserted as a picture and then we ask it to create a system dynamic model out of it and This is not perfect. So, it's just for example some tutorial reasons here. I wanted to show it to you, but again it shows the potential what AI can have and I think I know that there already really tools system dynamics are developed which really focus on this functionality, but that even with a simple prompts you can create a bit of interactive simulation here and yeah, maybe you see it. Yeah, for example yeah, you can thinking about the user and then about the consumer and down here you see that how the CO2 and how the landfill emissions decrease or increase depending on how the consumption is formed. >> [clears throat] >> Here of course the big question is where does this data come from? How is this developed? What are the algorithms behind that? What is the math behind it and so on. So, this is this is all what we don't know. But so again we would not use this as a final result, but I can imagine really simple feedback loops can easily be created with cloud and especially with this interactive visualizations from the the time graph of a time here. This is a really interesting and clickable functionality. All right. And now, uh we have analyzed our system map. We have found some um leverage, maybe even brainstorm some interventions. Um this part is often under a s- spective, which is about storytelling, because most of the projects there are people who were not part of the uh of the process or the uh audience does know about system mapping and is quite overwhelmed from uh the system map and can't work with it really. And so, uh we emphasize a lot on the topic of uh storytelling. And here we used um kind of a bunch of different tools um to just experiment what could be possible in terms of communicating uh the system stories. And here in the first example, uh we [clears throat] um used Base 44, which is kind of a live coding tool. So, that means that we uh And for the ones who are not uh familiar with live coding, it's basically uh you can enter prompts and create whole apps, web apps, or native apps um with tools. So, it's about software development in And here we used Base 44, um which is one of several um software uh applications um in that uh realm where um with which we created our own small system mapping tool, um which was capable of telling the story about a system in a visual way. And read it out loud. >> [clears throat] >> So, this [snorts] was um also just an experiment. So, we didn't work on this for months. Let's say it's really about a couple of days um prompting and trying things out. And this was uh one of the first results. And this topic is now uh about the blockade of of the Strait of Hormuz and the kind of effects as >> Think of the Strait of Hormuz as a tap. When blockade intensity rises, it tightens that tap and global energy supply drops. And when supply drops, global energy prices shoot up. That's just basic scarcity playing out in real time. Now, here's where it gets interesting. Higher energy prices don't just hit your petrol bill. They push up transportation and logistics costs because trucks, ships, and planes all run on fuel. And they drive up fertilizer scarcity and cost. >> So, you see So, the AI >> [clears throat] >> The interesting thing here, first of all, this tool was capable of creating the system map um on their own, similar to what you what we've just saw to uh from Systemic. Um so, it's also um an AI-based system map. And with uh one click, you basically then could have the system story and basically everything what we now see here is AI-generated. Um and yeah, the the the uh the the voice you just heard is the basic uh voice from I think it's a standard voice from the web browser or something. Um so, of course, this could be much more advanced with adding other um voice tools um and so on. But, uh the point here is that for us, it always was not easy to find a way how to explain system maps in an interactive way and this could be one interesting approach that we've have kind of a storytelling module or something like this going through variable by variable and explaining the feedback loops and so on. >> [gasps] >> So yeah, SS was just a test for us again quite interesting and we would definitely experiment with this more in the future. And the last example I have brought today with me is about again a bit about the perspectives which are in a system and we realized I don't know some years ago that it really helps if when we are communicating a system map that we have some human perspective in it which of course simplifies it again somehow so we have the complex system in the system map and we want to communicate it but somehow using human perspective which is then a bit simplified really helps to get into it. So it doesn't compile all the information in the system map but just gives a feeling to it and I think this feeling um yeah, we have a hunch and there's a really interesting things also to be developed in the future. And yeah, so again here just really quick examples which we prompted with Claude on the one hand so we have this created again and of personas and asked for which stakeholders are in the in the map that case again in the Strait of Hormuz example and so what are the effects and what are the perspectives of the people and then put this into Synthesia which is a A tool which creates AI avatars and and speak like some speak out loud. And here are two quite interesting slightly funny examples of that. >> The blockade forces us to reroute vessels, increasing transit times and operating costs across our network. Higher insurance premiums and security risks make it more expensive to move goods reliably. Our customers ultimately feel these disruptions through delays and higher transportation costs. >> So, this was you could imagine I don't know a CEO of a shipping company which has the challenges that the Strait of Hormuz is blockaded blocked and so on and so forth. You see here of course he looks a bit generic and a bit too happy for the situation I would say. But yeah. And for an example that's out. And the other one is from a farmer which yeah is affected by the fertilizer prices. >> I'm quite afraid of the effects of all of this on my farm. The prices for fertilizers already started to raise. >> And yeah, so these are just quite quick examples of how we could use video generation image generation to make to create more emotions. I think again here we have to treat it with caution if we are generating these abstract persons which are again it's a question about where does the data come from? Is it too stereotypic from the perspective is it kind of correct? So, I think there are a lot of questions coming with it. And not falling into the trap of using biased material or Yeah, stereotypical uh perspective. But nevertheless, um I think if we could use this as an introduction into the topic, then underlined with a lot of uh deeper information, detailed information, and the system map, then I think this combination could be quite interesting to follow uh that direction. And that's already it from my side with the roles. Um as I said, there are a lot of maybe other roles out there, and definitely a plenty of other tools you could use. Um this is just uh the tip of the iceberg, let's say, um uh what is what is out there, but these are the tools we experimented with. And to conclude in the end, so the benefits we found in different kind of roles is that the time can the effort can, of course, be reduced in creating a system map, uh analyzing data, and so on. And so, in that sense, it makes it more accessible to teams with limited research capacity. So, um I think that could be one of the really positive effects of using AI that more people can do system mapping because it uh but there's not that much uh resources and uh needed. Um because, for example, system mapping projects we are running, they are uh the duration is between, I don't know, 1 to 2 months to 6 months or something like this. So, these are really big um uh yeah, you need a lot of effort to go and create a system map, analyze all those data, and here this is, and so on. And if you involve all the stakeholders. So, there is kind of a shortcut, which is which is really interesting. Um then um obviously, it also gives us um different perspectives, like we already saw. So, maybe things we have not found independently. And so, especially if we are working with the human perspective on the data, and then maybe adding AI AI perspective as a add-on or as a kind of fact checking if there maybe we missed something or or like something like this. So, that really works well for us. Um so, the basis is our work, where we are really understanding the topic, but we then go and check it back with AI. And um yeah, the third benefit is that it less experienced practitioners maybe also enter these complex domains. So, um a bad system map is maybe better than a no system map. So, I don't know. You can challenge me on that. But uh if I don't [snorts] know, let's say in uh policy regulation topics or in government, where I surprisingly don't see that much system mapping happening, maybe these tools could help to um bring in more system thinking, even if it's maybe not uh then bulletproof to everything. But yeah, let's see. Even uh having the approach uh uh easier to uh to access might be a good thing. And yeah, in the end, I was also feeling that we create a system map is getting faster than we can have maybe more time on the analysis of it. >> [clears throat] >> So, these are the the benefits we found for now. But there are also plenty of risks, I think. We I I also talked about it uh highlighted some of them. Um These risks are also more or less general to the usage of AI, but of course you have hallucinations, you have untraceable research data, which I think is the most crucial point. So, I would not use web-based search in in a real project so that you don't know where it's coming from. In some cases, you can really track it back or get a list of the papers or the the data resources. Um but in the end sometimes if you're just using it the fast way, then we don't know where it's coming from. Maybe there are hidden biases in the in the data. You can actually I think we can be sure that there are hidden biases in the data because the whole um web is one of the part of the web which is based on English language and the Western perspective is quite more uh bigger than other perspectives of other countries and languages. So, there's definitely um a bias already in there. >> [clears throat] >> Um what we also saw and we're kind of afraid of is that the the personal learning gets lost so within the process. So, the the ones of you who were creating system maps and have gone through the process, there is actually it's quite an hard of a process sometimes, but there are also these moments of aha where you really understand something and say, "Okay, we have I haven't seen this connection and so on." And if you get the system map as a result by just pressing enter, then this learning in between is lost. So, it's kind of inside the model. So, I also wonder if this is a good thing. Um so, also here balancing between personal learning, really engaging with the data, and not only using the result I think um helps here. And also if we take AI as a shortcut, this also skips the conversations and discussions uh with stakeholders. So, also here I would suggest to use it as a starting point, maybe as conversation starters, but don't miss out on the conversations and discussions and all this in between where the learning and the reflection also within the the stakeholders is happening. So, ideally in our project, if you bring together uh stakeholders, for example, like in the Peru example, um that they are talking to each other and that they have a kind of a uh common understanding created and same language or language is kind of a bit more aligned in the end. So, all this uh also gets lost if you just uh yeah, uh cut our way through uh to this final system map and to the leverage points. Um yeah, so that's it for now. Um the benefits and risks we found, um again, also here we are quite curious about your reflection on that. Um and it as a final conclusion, so when used with care and awareness and intention uh including AI in our system mapping processes or practices can have great benefits, but it comes with the risk and downside that require thorough consideration. So, I think uh yeah, like I said, we can speed things up, um but we always have to take the time to check it back and bring the human into the loop of the project. And with that, this was my input for now, and I would and before we get into the breakouts shortly, hand over to Bowen. >> Brilliant. Thank you so much for that, family. And and yeah, I think it's a it's a really it's a really fascinating topic and and something which I particularly appreciated about, you know, reading your articles on this as well is um you know, that you're really, you know, obviously bringing the wealth of experience that you have on the systems mapping side and all of the you know, fun, challenging, nuanced, you know, um areas of that and really trying to think creatively and productively, but also I would say quite critically as well about you know, where AI can play a role. Um it was also interesting kind of looking at the comments in the Zoom as well uh about how we feel about each of these things. And it did seem that at least a couple of us in the audience didn't know how they or they didn't really like that this this idea of having AI avatars. So, this is interesting feedback, right? And but it obviously, as you said, depends on the use case and um you know, being aware of where the where where the information comes from, not treating it as a finished product, and you know, a bunch of other things as well. Um but yeah, of course, we're all you know, we're all in the same boat. We're all trying to figure out what's happening with the you know, this very very disruptive um technology. Um so, on that note, uh we're going to go into some uh small groups now, the breakout rooms. Um I think we're going to keep it to about five people per room. Is that right, Fabian? Approximately. >> Yeah, yeah, something like this. Um I'm just into it. Yeah, I'm just >> Yeah, I'll also I'll also read out the questions but before we get into that as well. Um but uh yeah, so we'll we'll we'll take around 15 minutes for this. And um I mean, you don't you don't have to follow these, but we did put out like a kind of a structure for for the discussion um on the Miro board, and uh if if you don't mind just zooming in a little bit, Fabian, into one of the one of the tiles, and then we'll maybe before we go into the rooms, I'll just read it out as well. And I'm also going to share a little link to the mirror board. Yes. So, the first one is, "How have you used AI systems mapping so far, if at all?" Um next, "Are we missing any AI roles or applications that you've discovered?" And by applications here, we mean approaches. Um so, like, you know, any roles or like approaches to you using AI in systems mapping, you know, or related areas that that you've discovered that would be really interesting to see, you know, yeah, what else are the people doing here. Um and then, the third question is is more around the actual tools that you'll be using. So, which AI tools have you been using for for for mapping? Um and also, any other open questions, thoughts, concerns, maybe, um that you have would be also really interesting to to to hear and and and uh uh discuss together as a group. And if also um if at least one person from each breakout room, if if one person can kind of volunteer themselves or um uh just share kind of the key takeaways or learnings or any interesting things back to the group. So, we'll take about 15 minutes after the breakout session as well, just to discuss together. Um that would be fantastic. Just so we can really harness the collective intelligence and um and [clears throat] yeah, see see see how we feel about these topics and what learnings we have. Um and so, yeah, I hope that that was clear. I'll also broadcast the Miro I'll I'll attempt to broadcast the Miro link in the breakout rooms as well, so you have that also there. Um and yeah, I think that that's that's pretty much all there is to the breakout rooms, unless I'm missing anything, Fabian. >> No, that's right. I just saw that is um four people are dropping out. And let me say uh reduce. So, I will let you use some uh breakouts shortly or that we have enough people in the room. Let me see. That's All right. I think now we have a good good little group. Um all right. Then I would say let's get started. Yeah, around 15 minutes, maybe a bit longer um for discussion um which I'm between the breakouts. And then we come back here for further reflection. Uh for those who maybe have to leave more early, um if you're curious about uh and want to dive deeper into the topic, we have uh more deeper training on that now developed on our side. You can have a look at it and also our normal uh introduction to system mapping. And for today, we have also a small discount for you. So, if you uh maybe think about and uh want to dive deeper into um these topic boxes, maybe I will just drop your contact dates here and we inform you and you will get a bit of uh a voucher code here on our side. So, but with that, I would say uh let's get started into the breakouts. And yeah, I will just close them then in around 15 to 20 minutes. You'll be back here. And just pick one of these templates. Yeah, great to see you all back here. Welcome back. Um I hope you all had some uh interesting and productive conversations. Um So, yeah, I I think I think we can probably get through each of the breakout rooms then. Um Is Yeah, curious like what came up in the sessions? Does anybody want to start? Any any interesting points? Um we can I can share the I can also share the mirror board or yeah, exactly. So we can have a look at what was on the on the stickies. Maybe we can start with the breakout room A, just go down to keep things simple. >> All right. I can take over as I speak as the presenter from the team. >> Volunteered by others to be >> Yeah, well, well, volunteering is >> Joking, joking. Sorry, please. >> Um No worries, all good. Uh so we talked about Well, um the the the thing that it you can create uh that um and a map uh so quickly, uh it changes the the application of it or or or brings it to an to another point where when you talk about whatever topic shall it be, I don't know, a leadership, uh knowledge, whatever it is in a in a in a company, you can also say, "Ah, and now we're talking about this specific topic, but let's have a look on it on a on a on a system level." And then you create a map and it leads the discussion to to to another to another level in a way. And that's that's uh something um well, we discussed and and and and some of us uh use it for. Uh second one um was about um well, you have a business idea and and you want to have uh um an an an idea of the impact uh on a on a system level if if this idea would be would be implement. And the third one we discussed was creating scenarios, so you can change in the depth of the prompt a bit and and it create the creates different scenarios. And you can play around with that quite quite quickly and and change certain parameters and and you get another another map and and learn uh quite quickly. Um when we talk about roles that may be missing, we had one which is the scenario builder. So, building on on what I just said. Um while for example, the six thinking hats from Bono or or devil's advocate or whatever. So, um depending on based on different perspectives or or or kind of types, um you can create different different scenarios. Another one we had was the assumption challenger. So, usually you come with certain assumptions about what the system looks like and what an impact on on on a system could be. And the assumption challenger would take these assumptions and and challenge them. By bringing in way way more knowledge, I I assume. Um and then there is one where I might have to ask my colleague what he meant because that was the last one before we got got cut out. Um So, the you're moving a bit faster, Fabian. So, I'll I'll go to to the board. Can you give a benchmark with ah So, well, I have this prompt where you can ask AI to create a benchmark of um other system maps that were created in the same field. So, that may other people have already asked for system maps and and and I don't know, compare the one you created with with others. And what's next? Um which ones we have already used? So, we have used the one from Bowen. Thanks a lot for that. So, the Miro plugin we use, and then Claude, and and ChatGPT, and I guess also other other AI tools. And the question we we we want to want to share with the audience was, how can we create transparency about what we, like what prompts we use, so what was our process, and what AI was doing in in the process. So, that's it from our team. >> Yeah, thank you. Yeah, that's really interesting. Also, thinking about yeah, to make it clear what the process was created with AI and what not, so that also in the documentation and in the communication of the system that we have in the end, it's clear how it was created, and so that we can really yeah, follow through where things come from. Um so, I think that's a yeah, it's a really good good reminder to also say it and and what was AI influenced, and so on. Yeah, it's really interesting. Okay. >> Yeah, there's maybe maybe just a couple of things I can I can just add to that cuz it relates to kind of the work that I was doing. Um it is it is a really important challenge, and I think that there's a lot that we can I'll just say very brief. The two things which I built into the Miro application, one is AI reasoning. Uh because yeah, exactly that, right? Like when you see a system that built by AI, why why did why what was the reasoning that that it put that there, right? Um cuz you obviously come to your own assumptions. And secondly, also, I mean this is this is the pro plan because, you know, there's a whole yeah, I mean costs to kind of add all these additional things, but you can also do research. If you do research, then then it comes with citations. But that said, I think there's a whole lot more we can probably do to bring transparency. I mean, the last thing I'll say is I mean Anthropic are actually working on reverse engineering the so-called brain of AI. I mean we have to use these words crudely and not to personify AI too much, but you know that they map these kind of almost neural networks that that kind of glow when certain answers are given, but you know that's above my pay grade, but yeah it's it's not a solved problem yet of like why did the AI you know what was the AI so-called thinking when it comes up with something, but yeah I think it's a wonderful point. So, just wanted to add that. Okay, maybe we can move on to breakout room three uh B song. >> Uh maybe I can take this over. Um Yeah, so I think we just have more general discussions here. And I think there were quite some interesting things in our conversation. Um for example, that there was a perspective that after kickstarting a project and I don't know doing some research and kickstarting the map that then the human should take over and all the analytical parts. So, I think that's quite interesting um perspective also thinking about when to use AI and when to use it not use it. So, I think that's yeah also quite flexible in that sense. Um maybe a bit connected is that if you have a small group creating a system map with just their knowledge in the room let's say, this could be really enhanced with system mapping especially in the beginning. Yeah, giving more perspectives, more insights like what you also showed and saw today in kind of inspirations. So, especially in this workshop models you could even have a small laptop being there. I don't know. Uh and ask the AI for input or inspiration during the workshop with the with the phone. Really uh interesting. Some kind of the uh bystander, maybe also critical thing which ask questions or represents the end user or the people who are affected by um by the interventions created. And maybe one thing I um found particularly interesting also was uh the competence of evaluating a system map um uh is also not given at any moment. So, not maybe so because the people need knowledge about the methodology and maybe experience in system mapping. So, when do you really need uh know that the system map is good or bad? And so, I think that's an interesting point. So, this competence, if you don't have that, maybe same as uh I don't know um evaluating fake news on social media. So, this competence um is really important to have in order to distinguish between a good system map or a bad system map or what kind of things which maybe are uh wrongly connected and so on. So, this is an open question of this competence If the if it this is not there, then it might be challenging to uh yeah, to find some errors also. Blind spot in the AI AI results. So, this is definitely interesting. Also, the question if AI itself can help or give some context and give some competence also. Yeah, when it gives some exa- results, that it also gives you maybe idea about how um the methodology in that way should work. So, quite interesting in that. Are there maybe any other perspectives from my group as I just talked? >> I think you have summarized our group B and I I think it's the same as the group C. So, uh So, it's a general summary of B and C, I think. >> Two in one. Wonderful. Uh great. Thanks for that, Fabian. So, no nothing to add from group C. Should we move on then to group D or I'll just pause for a second if if anybody wants to add anything else to group C? Okay. Uh yeah, my my wonderful group of group D. >> Feel free to unmute if you want to share something. No. >> Yeah, sorry, yeah, because I was No, no, I think it Fabian made a good summary of of the discussions that we had in in group C. I think it was group C. Or maybe we merged as one. >> A telepathy between two groups always good to see. Um great. Yeah, group D. Um maybe I'll just kick things off very brief briefly and and I'll and I'll I'll let the others kind of add anything, but I think one thing that was really interesting is there was a very like nuanced and critical view in terms of like, you know, what might be the best ways. We could see some areas that you could help, but that there were definitely some like caveats or things we need to be careful about and some, you know, just genuine questions as well. Um it's a very much a sense-making session. So, I I didn't really add a lot there, but uh yeah, that's that was kind of the framing from from what I sensed. Anybody want to share any any specific insights? >> Well, I said, thank you. >> Matty, anything you wanted to add cuz you had a really interesting perspective yeah on the nuances of of using AI. >> Yeah, my my my biggest concern was the leverage point aspect of using AI to find them. As as we all know, the most successful or the most important leverage points are within the way we think as humans and and the paradigms that we live in. And I'm not sure how AI is able to by itself make a judgment on where that leverage point is, right? Because you you you need to live it for you to be able to actually sense what a paradigm what living in a paradigm actually is or feels like. But I do I do see a a role where AI assist a a person sense it themselves within a system and actually finding where that leverage point is. That that was kind of like my my take. I I do not think that the the a judgment of an AI on where a leverage point is can be taken very seriously, but I do think that AI has a great role and capability of assisting um people to find them. Be a conversation, be a assumption, uh critiquing, uh etc. etc. etc. That that was kind of like my my main point. >> But what they what what what what I think is it's very interesting what you're saying now because um even you shouldn't be, let's say, uh restricted by the AI or or trying to see uh be critical about the let's say the paradigm, uh the human perspective. I think what you could do is actually uh just start like you always do with the humans. But actually see what AI could come up after the exercise you would do with the mapping and see if if you put them some arguments in into the the the the mapping and and see what he comes up with or the the AI comes up with and see if there's any how how how he mapped or she, which we don't know yet. Uh what what what what what the suggestions are of the map that came the AI came up with and see where the you're different. And that might highlight certain interesting aspects and then you could actually debate why why did we as a group, as humans, come up with certain areas that the AI didn't discover or didn't bring up? And then it can be very interesting discussion and why the artificial intelligence highlighted certain perspective that we didn't raise in our in our map and that can be very interesting actually about either awareness of of of some some uh some blind spots that we have or maybe some blind spot the AI has in in uh and and that can then be that that then as a human perspective. I don't know, but it can actually start an interesting debate. But then you have to go from, let's say, uh first the artisanal way of actually doing it in in a full human-oriented workshop, but then actually uh bringing them up and then offset them and and and identify the differences between the artificial generated one and the human generated one can be very interesting debate generating, I think. >> Yeah, definitely. Yeah, also [clears throat] I think the I'm sorry. Just mention that and that the the the um Yeah, that you first start with the humans and then bring it in so that it's not up influence in that sense. I think that's also an important one because I realized in my when I worked with it myself with my first created it with AI then I really torn to this kind of suggestion. So, I think that's a really important thing. Yeah, to start with the human perspective and then add it to it. So, I just wanted to say that as well. Mati, sorry. >> Yes, thank you. >> Okay. >> Right, I'm just aware of time. Um I think very briefly last last thing I'd like to just mention for our breakfast it was also interesting point and and could indicate something in terms of our domain as as humans moving forward given what AI can do is that as we know AI requires digital data. And that's basically its ground of of truth. And there are certain things that either are not codified which are important for understanding the complexity of a system's map. You know, such as oh, there was a meeting and actually the CEO hates this person and so therefore system change will be blocked. Whatever it might be. So, we know it, but it's not codified. But then arguably there are also some things that cannot be codified that maybe transcend the boundary of words. Maybe it's like a felt experience. It's an emotion. It's a cultural thing, you know, which which is you know, which AI arguably cannot directly ingest because it's not that medium of information. But that is you could argue a very human medium. Like when you're in a room, what's the vibe you get? How do you you know, what's the body language? What's the energy of a room? That's quite difficult to codify and but something that we feel as as as humans. It's just kind of natural part of our kind of psychology, right? So, you know, maybe that this becomes a a big part of increasingly a big part of what we do because AI cannot. So anyway, it's just an interesting idea which which came out from the session. Um But yeah, is there anything Fabian before we before we close and wrap up that you wanted to share? >> Um no, I think yeah, super interesting thoughts and perspectives. So like I said, everything is kind of in flow. So I think if we meet I don't know in half a year or a year, we have a new thing to keep perspective how this all all these tools can be used. Yeah, no, just to yeah, let you know, if you have interest in well, if you're interested in further learning about AI and system mapping, yeah, just have a look here on the on our training offers and also maybe on blog articles and so on. So we have written also a bit of background [clears throat] our our thoughts about this whole topic a little down. So if you're interested, just have a look and yeah, it would be great to see you around here in the system information network or in other occasions. I think yeah, it was really quite interesting exchange today. >> Brilliant. Thank you so much. I will share I've just shared a a link to the mirror board. So what we like to do especially for the tech hub because it's it's still pretty new. So I think this is our third event actually. Um And we want to yeah, continue exploring this intersection with systems thinking and and and technology. It really depends. It's a very community driven thing. So really appreciate your thoughts and and feedback. You know, you can give like a rating of you know from scale one to 10 what you thought of the session. And also if you have like, you know, burning topics or things that you're like, I really I'm trying to figure this thing out and I'm just not really seeing the right kind of events or content or research and and and I really, you know, want more in this area then then, you know, please put up a sticky because this is exactly, you know, what what helps fuel us. Right, when we're kind of thinking about what next events to do or what content or research we're providing. So that would be really really useful for us. Um and yeah, so thank you all. It's been a really fascinating discussion. Um you know, on the mirror board you've got the resources to like the article and and and the different tools and things that we mentioned. Um and yeah, on that note, thank you so much for joining and hope to see you again in our future events, sessions. And thanks so much Fabian for for the wonderful content and thought-provoking ideas.