---
id: "workshop:defining-problems-with-the-public"
type: "InnovateUS workshop"
title: "Defining Problems with the Public"
url: "https://innovate-us.org/defining-problems-with-the-public"
date: "2025-10-16"
duration: "60 minutes"
language: "English"
series: "Reboot Democracy: Designing Democratic Engagement for the AI Era"
instructors: ["Claudius Lieven"]
recording: "https://www.youtube.com/watch?v=TD75M3-uH64"
transcript: "automatic, from the recording"
source: "InnovateUS (innovate-us.org)"
licence: "CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/), as stated on innovate-us.org"
---

# Defining Problems with the Public

InnovateUS workshop · 16 October 2025 · 60 minutes

All too often, governments set out to solve problems without fully understanding how citizens experience them. At the same time, they collect vast amounts of information through consultations, workshops and social media. As one leading politician put it, ‘We are rich in data but poor in insights.’

This workshop with Claudius Lieven from the City of Hamburg and the Digital Participation System (DIPAS) shows how digital participation platforms can be used to collect massive amounts of citizen feedback and transform it into a clear picture of the problems through artificial intelligence, helping communities solve problems.

By clustering, summarising and mapping thousands of citizen comments, the DIPAS_analytics software developed by the City of Hamburg over the past three years enables governments to identify patterns in citizen contributions, distinguish concrete suggestions from moods and emotions, and base policy-making on the actual concerns of residents. The DIPAS software and method, which has already been implemented in nine German cities, shows how AI can turn citizen participation into a tool for defining problems more precisely – and thus developing better solutions.

## Learning goals

- **Explain why defining the problem matters** and how public input can sharpen problem framing in policymaking
- **Use AI-supported methods** (like those in DIPAS) to transform citizen feedback into structured insights and problem statements
- **Interpret clustered and mapped feedback** to distinguish between general sentiments and actionable proposals
- **Apply lessons from German cities** to design engagement processes that surface community-defined problems in their own contexts

## Speakers

- **Claudius Lieven**, Project Manager, Hamburg DIPAS

Part of the series [Reboot Democracy: Designing Democratic Engagement for the AI Era](https://innovate-us.org/workshop-series/democratic-engagement).

Recording: https://www.youtube.com/watch?v=TD75M3-uH64

## Transcript

Automatic transcript of the recording: names and terms can be misheard, and the times can be a few seconds off. Each time links to that moment of the video. The transcript alone: https://innovate-us.org/defining-problems-with-the-public.txt (plain text) and https://innovate-us.org/defining-problems-with-the-public.json (every segment with its start and end second).

**Agueda Quiroga**

[0:02](https://youtu.be/TD75M3-uH64?t=2) Okay, so now I will open the waiting room, and we will wait like three more minutes to start, and I will, okay, so I will, I will disappear, give me a moment to disappear. Welcome everyone, if you're just joining us, bear with us, we're going to give it a few minutes to let people in from the waiting room, and we'll get started in just a moment. As you're coming in, let me just encourage you to please mute yourself, and to pass the time while we let people in from the waiting room.

[1:04](https://youtu.be/TD75M3-uH64?t=64) We've put up a short three-question anonymous survey, so we'll just ask you to take that, and I'll make some announcements in just a moment, but we're just letting people in. Thank you. I think what we'll do in the interest of time, we have a lot of people joining us, but I'll go ahead with some administrative announcements, and that will leave us maximum time for our speaker and our conversation, so let me get us started as people come in from the waiting room. Again, if you're just joining us, please feel free to go ahead and rename yourself by right-clicking on your square, we'd love to know who's here, you can add anything you want, your affiliation, your organization, your agency, pronouns as you prefer.

[2:11](https://youtu.be/TD75M3-uH64?t=131) Please also make sure that you're on mute, and go ahead and take the short three-question anonymous survey, which you'll see in front of you, it just asks you three questions about what you're hoping to learn, and we simply use this to develop our future programming. Thank you to everybody for being with us today. This is the fourth in our series on AI and public engagement, our Reboot Democracy series. This is a collaboration between Innovate US, the GovLab, the Allen Lab for Democratic Renovation at Harvard, and Professor Danielle Allen, and we're very, very happy today to have with us Claudius Lieven. Claudius is a political scientist and urban development expert, and most importantly, for our purposes today, director of the DPUS program that is using AI in exciting and innovative ways to help citizens speak, and more importantly, cities listen to what they have to say.

[3:11](https://youtu.be/TD75M3-uH64?t=191) Claudius has been with the Hamburg Ministry of Urban Development and Housing since 2008, where he's led a lot of innovative programs that I hope we're going to hear more about today. And unusually today, let me just add one more administrative announcement, Claudius, before I pass the mic to you, is we are experimenting with AI-enabled real-time translation. So you can, we're going to post the instructions for that in the chat. You need to just choose English as the spoken language, as the original language, and then you can click on your preferred language if you would like it to be in German or in Spanish or in something else. Agueda, anything else I need to say on the translation before we get started? No, I already passed the information on how to turn on the closed captions in the chat. So if anyone has a question, please reach me through the chat and I will be happy to answer.

[4:12](https://youtu.be/TD75M3-uH64?t=252) Okay, I'm going to end the poll and so we can get our screens back. And Claudius, let me pass it to you. We're eager to hear from you about the work that you have been doing building open source analytics to help government agencies to listen. And we just want to welcome everybody and there will be lots of time for conversation and questions. Please, as we're going along, feel free to use the chat to type any comments, any discussion. We have lots of experts here in the room and people love to share with one another. You can answer one another's questions. If you have a question for Claudius, I would just ask that you use the Q&A button at the bottom of your screen. So if you don't see it, just click on more and you will see the Q&A button. That's the right place to type your question or to let us know that you have a question so that we can help to moderate and organize things as we go. Claudius, let me pass it to you. As more people come in, welcome and tell us about what's happening in the city of Hamburg and across cities in Germany.

**Claudius Lieven**

[5:22](https://youtu.be/TD75M3-uH64?t=322) Yes. Welcome, everybody, as well, from my side. And thank you very much, Beth, for inviting me to this very interesting course of events and for the opportunity to report a little about how we do digital citizen participation in Hamburg, Germany, and how we are beginning to use AI to help governments. To improve and make the participation more impactful. I would begin to present to share my presentation. Okay. Yes, we should be able to. Let's take me off spotlight and put your slides on. Thank you. There we go. There we go. Okay, let's begin at the start.

[6:25](https://youtu.be/TD75M3-uH64?t=385) Okay. Yes, I'm going to tell a little about how we move from digital citizen participation to the learning city. I'm Claudius Lieven, head of a staff unit for citizen participation in the Ministry of Urban Development and Housing in Hamburg. Hamburg is one of the 16 federal German states. I'll go from the where to the what and on to the how we do it. Firstly, where am I at the moment? I'm in northern Germany, central northern Europe, about 100 kilometers from the North Sea in between Amsterdam, Copenhagen, Warsaw, Vienna. And I'm a hamburger, but not such one. And we are a lot of hamburgers here. But this is how our city looks. And it is as well circular as the hamburgers.

[7:26](https://youtu.be/TD75M3-uH64?t=446) Well, that's the old shape of the defensive wall from the medieval ages. But it's still in the city morphology. And it is a very aquatic city. It's by the river Elbe, a big stream, and it's a port city. And that's why we are really fond of ships. They can go to the center of the city right into the harbor. It's 70 square kilometers of harbor in the heart of the city. But let's get a bit more serious. Hamburg is a city republic. And this is a bit of our DNA. It became 1510 a free city in the German Empire, and still we name ourselves the free and Hanseatic city of Hamburg. The Hanseatic was this trade league centuries ago, but it made us become a federal state, a city republic. And the state parliament is since centuries called the Bürgerschaft, which means just citizenry.

[8:34](https://youtu.be/TD75M3-uH64?t=514) And so it is a self-governing city. We had far-reaching autonomy over the centuries, aside of these unfortunate 12 years between 1933 and 1945. Ministerial duties in Hamburg are performed by currently 11 senators who each had a ministry, small ministry, of course, in comparison to bigger countries. And this free city tradition is as well inscribed in the political system. Of course, it's a parliamentary republic, and the city parliament is the most important body for political decisions. But we have as well the possibility to do referendums, or the citizens have the possibility to start a campaign which leads to a referendum.

[9:37](https://youtu.be/TD75M3-uH64?t=577) And we had just two referendums on the last Sunday. One was adopted by the majority, the other declined. But that's how it goes, and now we have two decisions, and the city government will enact the referendum that has won. But I am here to talk about another mode that we call the deliberative mode. And this is about public debate in the agora of the city, and our topic especially is the digitization of this public debate. Especially in urban development, because my unit is in the Ministry of Urban Development and Housing, and we advise and support authorities, district, public companies in participation procedures.

[10:38](https://youtu.be/TD75M3-uH64?t=638) We organize information and participation on citywide topics ourselves, and we develop new methods and instruments for citizen participation, and this is especially digital citizen participation. On the left side, you see our logo and our motto, and that is in German, Auf Dialoge Bauen, or in English, Building on Dialogue, because we are very purpose-driven. Our aim is to further develop the city and to make it possible that a lot of apartments, of housing, can be built. And that is a thing where citizens have a lot of need to discuss about the changes in their neighborhoods and in the urban environment. And that is why we developed the DPAS system. DPAS is an acronym for Digital Participation System, and I'll go to explain a little bit how we use it.

[11:41](https://youtu.be/TD75M3-uH64?t=701) DPAS consists, broadly spoken, out of five elements. The most important is the online participation tool, that is quite similar to other online participation tools like GoVocal or MapJanair or Consul. It is a bit more data-driven. I'll talk more about that later. And we do, as well, digital participation on-site, events using touch tables and smartphones and combinations of those. We are connected to our urban data platform, from where we receive all the maps, aerial photos, 3D models, and a lot of other urban data. And we have specialized evaluation tools, as well, a point I'll deepen later. And a navigator, which is a KPI dashboard to give us an overview, as well to give citizens and politicians an overview about what happens in the system and in the debates.

[12:48](https://youtu.be/TD75M3-uH64?t=768) Here you see a screenshot of the navigator. You can easily access it when you Google DPAS navigator. And we have carried out more than 130 processes in the last 10 years and are, this year, at about 16 processes. So it's a very continuous work and a debate that is going on. The online tool is map-based in the primary mode. So citizens give feedback to plans, we present them. And they can comment on the feedback, on the contributions one citizen has written, and they can rate it, like sums up, sums down. You can, as well, do some other things with it, like compare plans and have digital storytelling. But I'll focus here on the map-based participation, because most of the feedback comes through this channel.

[13:56](https://youtu.be/TD75M3-uH64?t=836) And so the question is solving problems with the public. And here I have given an overview about the topics of the processes of 2025. Because, there you see, we talk a lot with the citizens about concepts, about redesigns, ideas, and as well a little use participation. So it's about how we design the city of the future, how we are continually developing the city further, redeveloping sites, roads, public spaces, and all this. And this is what is being presented through the system to the citizens and where we gather feedback. And where we then have to analyze the feedback and derive meaningful information, recommendations out of that. Because in this deliberative pillar of our democratic arena, we do not take decisions.

[15:02](https://youtu.be/TD75M3-uH64?t=902) It's about recommendations, it's about debate, and the decisions are taken by the democratically elected bodies. But the meaningful debate provides a lot of insight and information about what the citizens want, what they need, maybe about their preoccupations as well, so that we learn where we have to adopt further precise plans or to explain more in detail what is coming. And this system, the DIPA system, is an open source tool, we have published it on various platforms, and meanwhile as well the civil society in Hamburg has adopted it. And this is a process, here you see on the map about 2000 replies, and it is about unsealing the soil. It means you can make suggestions where the pavement should be removed so that more water can enter into the ground.

[16:12](https://youtu.be/TD75M3-uH64?t=972) Why is this important? It is the idea of the sponge city. We live in the time of climate change and we have to deal with heavy rainfalls and with flooding within the city. And one way to prevent that is to unseal pavement, paved areas. The idea is coming from the Netherlands, there it is very popular. A group of citizens' initiatives has adopted the idea and implemented a DIPA system, and they call themselves local power. And it is now the first time we have a bottom-up initiative in a private-public co-creative process, in cooperation meanwhile with the Ministry of Environment. They have said, wow, this is a good idea. And it is not always a public space, it is as well private spaces. So the driveway to your garage doesn't need to be paved. It can as well be a soil, a surface that lets water enter in.

[17:16](https://youtu.be/TD75M3-uH64?t=1036) So the URL is here. And so we see that we have meanwhile not only the top-down, because if we offer participation it is top-down by the government, but it is as well now becoming a bottom-up. That is something we really like to see, and we encourage and hope that more of that is going to happen in the next years. Another thing is that DIPA is not only used in Hamburg. There are some other German cities who have as well adopted the system. This is a picture of Munich, so a process about the further development of the city centre of Munich. And in Munich, of course, it is not hamburgers, it is pretzels. So they changed a little bit the icons, that it is a bit more Bavarian. Just one example. Yes, so this is how we apply the digital participation, and we see the scaling.

[18:26](https://youtu.be/TD75M3-uH64?t=1106) Last year it was more than 11,000 contributions, and these contributions often tend to be like letters. So it is not just a short text message, but if you think on 11,000 letters, you can imagine that is a whole lot of information. And how did we achieve this, that the citizens use it so intensively? The first thing is the tool is very easy to use for the citizens. It does not take you much time to write your contribution, if you do not want to write 1,000 characters. And that helps the citizens to just provide us their view. An attentive service from my team, who supports especially within the administration, the district planning authorities and the public companies who use it. That is as well important to have.

[19:29](https://youtu.be/TD75M3-uH64?t=1169) The continuity, meanwhile, we can say that for Hamburg, DPAS is quite well known and learned within the administration, as well as in the public, within the citizens. And the most important thing, I think, is the trust that the citizens have. Feel no barrier to use this system. That is because we work very transparent. The source code, of course, is open source. We work a lot with open data and we provide everything we know and we can write down as open knowledge, so that everything is understandable and transparent. On the other hand, the users and the usage of DPAS is anonymous. We do not store any personal data.

[20:30](https://youtu.be/TD75M3-uH64?t=1230) That lowers the barrier of expressing oneself. But we have no problem with hate speech, I must say. The citizens really use it in a very deliberate way and provide information. Communication is important. Reporting, presenting, publishing, and working with the stakeholders, initiative, science, foundation, so that trust in this approach grows over time. That is, for us, taking the participation promise serious. We really want to give the citizens a meaningful say in the debate about the development of the city. Taking it serious means working with a lot of information. Just a short view about a little bit of text. It is only text, but there is a lot of information within it.

[21:36](https://youtu.be/TD75M3-uH64?t=1296) We have the possibility, since some years, to do statistical analysis. That gives us a very good short overview about quantities of reactions. The green dots mean a lot of people voted sums up to contribution. On the lower right side, you see a line of more reddish dots. That is a conflict line. There is one major road where we have problems with the traffic, bicycle lanes, and all that. That expresses directly in the data. You can have, from the statistics, a quick overview where the action is highest and where conflicts arise. Here is the conflict line. On the other hand, we cannot, with this method, read out the text that the citizens provide us.

[22:41](https://youtu.be/TD75M3-uH64?t=1361) Of course, 20% of the task solved by doing a statistical analysis, 80% is here in the textual data. We began some years ago with the development of an AI-based system which helps us to cut through all this, a lot of information. This has now become an orchestrated, open-source AI pipeline. We use primarily Lama 3.1 and Roberta, a BERT derivative which is specially trained for the German language. We have developed this pipeline. You can see it from top down. It begins with a named entity recognition service. That's for us to eliminate private data. Sometimes people write their names or telephone numbers, and we do not want to have that in our data.

[23:45](https://youtu.be/TD75M3-uH64?t=1425) Then it goes on with a key message extraction and so on. I won't go here in details. Later in another session, maybe we can have more of that. The core functionalities are the identification of the key messages within the posts, the assignment of these key messages to predefined categories like mobility, housing, green spaces, or more detailed ones. The aggregation of the messages into aspects. These are sub-clusters, and the aspects are aggregated into clusters. Those can be spatially represented as well as the aspects in the clusters. Of course, we still have statistical information out of the data sets of the contributions.

[24:49](https://youtu.be/TD75M3-uH64?t=1489) One example. We have just had in this summer a debate about the remodeling of the train station at the center of Altona. This is one of the Hamburg districts, a former independent city. There, the train station shall be relocated. At the moment, we asked the citizens for comment and feedback for the public spaces surrounding this building you see in the upper right. This is a train station from the early 80s. Citizens commented and wrote contributions, commented and rated those in a plentitude. If we look on the statistic at first, we see that the topic of urban design is that what has received the most response.

[25:52](https://youtu.be/TD75M3-uH64?t=1552) It has as well received a lot of ratings, several thousands, and a lot more than all the other categories which we see here. All in all, it's about 3,500 ratings and 400 contributions. We asked ourselves, okay, what is in here? Yes, this is what we are focusing on. This is a short screen video where I just entered into the AI analysis system to check what is happening in this category of urban design. These are all the categories. Now, I look on the one with the most contributions. Okay, this is urban design. We have more than 2,300 ratings positive here, 520 negative, and 57 aspects.

[27:00](https://youtu.be/TD75M3-uH64?t=1620) Okay, let's look into these categories. To get an overview, we ask for the assessment that citizens, I look for the assessment that citizens gave and look on the most commented contributions. The AI sums this up together with, okay, we have a problem with the identity and the centrality of this place. We fear the loss of historic building substance, and we need to better use the urban potential of this area. Well, so this was for us a moment of aha, because we explicitly asked another question. We asked, give us feedback to the public spaces around the building, but the citizens said, no, we don't want to talk about the building.

[28:02](https://youtu.be/TD75M3-uH64?t=1682) That is, we thought we can talk about the building later because of the German railway. But it was, in fact, they said, no, no, we have a problem with this building. And the learning effect now for us, the data helped us and the analysis helped us a lot to inform our senator and those who are steering all the redevelopment process. Now we have to shift our perception about how we can do this planning process. We need to address the question of the identity and the question how we deal here with, that was what was even before this building. That was a brick built neo-Gothic train station. A lot of people wanted it back. Of course, we will not build that back, but we will now in the urban design competition, put a special emphasis on the question of identity and historic connection so that we in the best way take up the citizens' preoccupations or not preoccupations, what they desire, what this place should do for the center of the district.

[29:24](https://youtu.be/TD75M3-uH64?t=1764) This was a very short example how the AI helps us for rapid extraction of key messages out of a participation process. And it allows us to have a transparent, value-neutral and traceable analysis. And this enables well-founded conclusions and recommendations to those who are steering the urban development in our city. This increases the impact of the. What we try and what we want to achieve. What we see in the same moment is that the use of AI in evaluating citizen contributions requires sensitive communication. Because if we just say we use AI, we need to explain some things. And this system is absolutely open source, it's transparent and we are going to publish everything of it.

[30:25](https://youtu.be/TD75M3-uH64?t=1825) And that is, we think, very important to maintain the trust of the citizens. On the strategic level, we are looking forward to that we have the possibility to uptake georeferenced local knowledge at scale. Because we have tens of thousands of contributions and we'll collect more. And we can make it a persistent knowledge layer in our urban digital twin. So not just to analyze a single process, but groups of processes, topics and to analyze the debate over time and how topics and conflicts develop. Yes, this, we hope, leads us to the possibility to become more or less a learning city. To have a feedback loop, a better feedback loop from what the citizens think and perceive how the urban development is carried out.

[31:38](https://youtu.be/TD75M3-uH64?t=1898) And so that we have a solid base to be closer to the demands and the needs of the citizens in how we are carrying out the urban development. So far for the first presentation and I'll stop the screen share and we are open for the discussion. Claudius, thank you so much. That was really wonderful. What I want to do is to start by just asking you some clarifying questions. There are many questions that have come in in the chat that are, let me call it clarifications. And encourage people in the meantime to put more reflective questions in the Q&A so that we know we want to talk. While I'm asking questions, we'll keep the recording going. As is our practice, once we open it up for conversation, we will turn off the recording so that people can feel free to ask anything and everything.

[32:43](https://youtu.be/TD75M3-uH64?t=1963) So a whole bunch of clarifying questions. The first one is, do all projects get mapped and used with DPAS or how do you select which projects? Should I give you one at a time or would you like all of them? Or three of them? Okay, three is a good number. So the first is about how you select a project or is it all projects? I'll give you three. Second clarifying question is, you said it's anonymous, but is registration required? And then there was a third question about sort of the length and phases of a project. Is there a standard cadence for how long a project lasts, sort of an engagement lasts? Does it have phases or does it vary from project to project? Let me stop there. And then we'll do one more round. And again, please put further questions in the Q&A so we can open it up. Yes. The question, if all projects are mapped, all projects that are carried out with DPAS are mapped.

[33:49](https://youtu.be/TD75M3-uH64?t=2029) Because it's a map-based systems, but it's not all participation projects of Hamburg. We are doing participation in urban development. And there are as well other participation processes run by the Department of Social Affairs or sometimes from the Education Ministry. They have as well participation processes, which are mostly not map-based. So most things in the urban space around mobility, around green spaces, around urban design, that is carried out with DPAS. And it is, that is important, the informal participation. We have as well a formal participation, which is obliged by law. That is working a little bit different as well, a digital prong. But that is not that transparent and not that easy to use.

[34:51](https://youtu.be/TD75M3-uH64?t=2091) There you have a registration necessity. And with DPAS you have no registration necessity. Our philosophy is to work really with a low barrier. The system is as well digitally barrier-free according to our standards in Germany, so that you can use it as well as a screen reader. It's not so easy with a map. Of course, maps are quite hard to make them barrier-free. But this is as well an aspect. And the anonymity, of course, has advantages and disadvantages. One advantage is we can share the data easily within the administration. Otherwise, if it be mixed with private data, that would be much more difficult. And it has this low barrier regarding presenting yourself. On the other hand, of course, we often would like to know more about our customers or who has said which age groups, from where do the people come and so on.

[35:58](https://youtu.be/TD75M3-uH64?t=2158) We can integrate questionnaires into the system. Where we as well do not directly collect personal data, but things like highest education, zip code, age group, gender, these things. So, to get a bit of socio-demographic overview about who has taken part in the participation. Yes, the processes. We, in general, advise to do not only digital participation, but to have a mix of on-site events, at least information event at the beginning, or some as well workshops, if possible. And then accompanying the online participation.

[37:00](https://youtu.be/TD75M3-uH64?t=2220) And within the planning process, usually the online participation is opened around four weeks. Sometimes shorter, sometimes a little longer. But in the planning process, when we want to get to a redesigned plan, we usually calculate it should take us half a year to a year. And so you cannot open the commenting phase for a year. You need to have like four weeks. Usually we see as well in the pattern that in the first days, the reaction is more intense and then it goes down. And after four weeks, it's very little that then comes. So, that is how it is embedded into the planning process. And one more clarifying question, which is how many staff does it take now versus previously? But I think there's a staffing question.

[38:01](https://youtu.be/TD75M3-uH64?t=2281) What does it take to do this practically? Yes. I have in my unit two colleagues who just do the attentive service. So, who are advising districts, public companies, how to carry it out. They carry it out themselves. They are the local administrators. They have as well to do the evaluation afterwards. And in the local authorities, it is the colleagues. There is no dedicated personnel for that. It's the planners, the urban planners usually who do it, often supported by specialized bureaus, planning agencies who support them. So, we cannot directly say, okay, we reduced the staff broadly, but we can say and we can see we increased the activity broadly.

[39:08](https://youtu.be/TD75M3-uH64?t=2348) So, we have a lot more participation processes than we had 10 years ago in this form. So, with the same personnel, we see it there can a lot more be done than before. So, we have so many questions waiting, and I have such a cold that I am going to open it up to others to ask their questions. Again, if you can mute yourselves so we don't hear you in the background, I'm putting the queue in the chat. but I'm going to invite those in.

* * *

Original page: https://innovate-us.org/defining-problems-with-the-public · Licence: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/), as stated on innovate-us.org
