How Germany’s most forested district uses participatory planning and AI to strengthen participation

/

Social healthcare graphic from the social conferences. Source Siegen-Wittgenstein

This is part of a series of stories on digital participation platforms produced with support from the Friedrich Naumann Foundation’s Global Innovation Hub, Taipei. To read others in the series, see these case studies from Northern Ireland and Colombia.

The District of Siegen-Wittgenstein, located in North Rhine-Westphalia, is known as the most forested district in Germany, with nearly 71% of its area covered by woodland. At least 270,000 people live in the district, and a lot of them live in small towns and villages separated by stretches of forest. For many residents, it’s a beautiful and fulfilling place to live and work, yet the geographic and demographic realities also pose unique governance challenges.

Rethinking tradition and participation

Resource distribution has often followed the Giesskannenprinzip (watering can principle) when funding came from the federal government, it was spread evenly across municipalities regardless of their specific needs and the different challenges faced by different communities. However, when Jason Wagner from the District of Siegen-Wittgenstein and his planning team stepped in two and a half years ago, they wanted to try a new approach. “We want to be as close as possible to people's problems. So we opened up participation to hear about their needs, which meant speaking directly with citizens and stakeholders, and implementing projects on the smallest possible scale,” he explained. Instead of treating every area equally, they sought to identify where resources were most urgently required and design projects that respond directly to citizens’ needs.

For Wagner and his team, moving beyond the Giesskannenprinzip meant rethinking participation and building bottom-up processes by combining the use of digital tools and local voices. So in 2024, the district organized ten “participatory social-planning conferences” in ten of its eleven municipalities. The purpose of these events was twofold: firstly, to identify local needs and problems, and secondly, to strengthen cooperation and networking between residents and the district administration. Each event gathered between 30 and 90 participants around three discussion tables, bringing together citizens, municipal employees, local politicians, care service providers, youth workers, and other community representatives. The conferences gave citizens the chance to articulate both the problems they face and share the resources they already have. 

“People often had little contact with the district administration, so being there physically, listening to them, was very important,” Wagner recalled.

Piloting a digital platform

After the 2023 iteration, the team saw a need for a digital platform: they needed to include a digital platform in future conferences to support their efforts. To support the 2024 conferences, the district piloted deliberAIde, an AI-powered platform for transcribing and analyzing discussions. Jason Wagner heard about the platform from a colleague and thought it was suitable for the district’s purpose and needs. 

“We heard about the tool internally and discussed its offerings because we needed a tool capable of transcribing the processes we were doing at that time, and AI was a big theme; we thought, this is exactly the product we need. We decided to contact deliberAIde. At that point, deliberAIde was still in its alpha version. We spoke with the deliberAIde team and tested it together.”

“So, it wasn’t a case of us saying, we need a digital participation tool, let’s search and find something. It was the other way around. We heard about it and realized it fit perfectly for the moment. We were a bit lucky. It hasn’t even been released yet, so we are really like the very first pilot partner. The district is a super early, pre-adopter,” Wagner recalled.  

Often, governments and organizations will have a budget for a digital participation tool and its use during the process, but the district had none. “We had no costs because we were testing it for ourselves. We had the benefits on our side, and I hope deliberAIde also benefited. We gave feedback through interviews where we shared what was working, what wasn’t, and in that way, it was a situation where both sides gained.”

The tool worked as follows: 

The deliberAIde platform uses speech-to-text and anonymization models to record and anonymously transcribe discussions. It uses Large Language Models and image generation models for summarization, visualization, analysis of discussions and the creation of comprehensive documentation materials and reports.

A laptop was placed on each table, with the group’s consent. “The process itself was simple. We placed the laptop in the middle of the group and asked everyone for consent, as required in Germany. We explained how the AI works like Netflix, but in reverse; it streams through, listens, and transcribes the discussion in real time. Afterwards, the audio is erased. That part was very important to clarify, because in Germany, data security is a sensitive issue. People needed to know that everything was anonymized; no names, addresses, or numbers are included in the transcripts.” In the version that the district used, participants did not interact directly with deliberAIde; the users were facilitators and moderators, who ran it on a laptop or tablet. For participants, the only interaction is giving consent for recording. 

According to Wagner, “We used deliberAIde for around 10 or 12 of the discussion tables for transcription as the conversations were lively, sometimes loud, sometimes with people really standing their ground. It was very interesting to get these direct insights from residents, because we are planning for 270,000 people, and this kind of information is very valuable when compared with the quantitative data we already have.”

deliberAIde simultaneously recorded and transcribed conversations while anonymizing all data. After discussions, transcripts were available immediately for analysis. Wagner appreciated the tool because once the discussion was over, they had immediate access to the already anonymized transcripts. 

Wagner also emphasized, “The platform also allowed us to cluster ideas, color-code them like post-its on a digital board, and create protocols directly from the transcripts. Since this was still a test phase, we ran the AI alongside our traditional methods, manual notes, colored cards, and pin boards. We could even ask the transcript later to highlight unanswered questions, topics that came up multiple times, and emerging connections between different discussions. However, it also means you have much more information to handle. But the AI helped by automatically clustering ideas and organizing them visually. That saved us time and gave us a richer picture than manual notes alone.”

 For the district, the tool offered three main functions, including: 

  • Comprehensive documentation: Every idea was captured as a digital sticky note.

  • Clustering: AI grouped similar ideas together for easier analysis. 

  • Tagging: users could create categories to sort and track discussion themes.

The clustering and tagging were particularly helpful for their conference as they found it easier to condense hundreds of ideas into themes, and thereby identify specific categories such as citizen concerns or responses to key questions. Wagner recalled, “Because a lot of times in long, long discussions, you get a hundred or 200 ideas, you can use the clustering functionality to cluster similar ideas together, and that will give you an overview of all the topics that have been discussed.”

As German is a high-resource language, there’s a lot of training data available to make the platform effective in capturing what the district needed. However, according to Wagner, “sometimes the AI didn’t catch every word correctly, but that was often due to the setting of large rooms, background noise, and many people speaking at once. This made it difficult for both the moderators and the AI to understand everything, for our purposes, though, that wasn’t a problem, as we didn’t need a word-for-word transcript. What mattered was capturing the discussion, the back-and-forth, and the main points. With German, the system worked well overall. Of course, dialects can be tricky; Bavarian German or some northern accents are hard even for native speakers to fully understand. In some villages, people spoke in such a heavy local dialect that even we struggled, so it’s understandable that the AI also had difficulties. But for standard German, there were no real issues.”

Benefits and outcomes

Even though they were piloting the tool, they found immediate benefits. For example, they found that during the discussions, the social planners could fully focus on moderating discussions and interacting with citizens, instead of moderating and taking notes simultaneously. [Social planners are specialists who analyze municipal data and social developments in order to plan and manage needs-based offerings, services, and infrastructure for improving living conditions. Typical tasks include social reporting, needs analyses, policy consulting, and the development of integrated concepts with the participation of local authorities and stakeholders.] 

The platform helped the district in comprehensively capturing ideas shared during discussions, identifying themes, and extracting other relevant insights. Previously, the team would have to manually sort through notes to create documentation after the discussion forums. With deliberAIde, the social planners could automatically create documentation and summaries within minutes, freeing resources and time to focus on next steps and implementation, instead of spending countless hours and days on documentation. 

Wagner explained that using the tool enabled them to generate valuable insights that went directly into policymaking. After the 2024 conferences, the team was able to submit two citizen-driven project proposals to local politicians for decision-making. The two proposals include:

  • Social support access: Residents highlighted the lack of a “single window” for help in critical situations. While many services exist, they are fragmented. In moments of crisis, citizens need one simple number or platform to access immediate professional help. Therefore one of the proposals is to improve access to social help via AI: The district plans to develop a multi-channel, low-barrier platform (web/app, hotline, messenger) that uses AI to clarify needs and quickly direct people to the right local services. It should complement, not replace, professional counselling, and its development will be scientifically supported by the University of Siegen.

  • Childcare and healthcare: Residents described long distances to doctors and limited childcare facilities, which matched district data but gained urgency through personal stories.

    • The second proposal is about third places, it includes a county framework to develop existing social infrastructure, especially small rural childcare centres, into open, intergenerational meeting points with clear criteria, governance and links to funding and social space planning.

They have also managed to identify other recurring themes and issues affecting the communities, and where the resources could be better directed, including mobility. With villages scattered among forests, reliable transportation remains a pressing issue.

The issues that came up included mobility since 71% of our region is forested, childcare shortages, and limited medical services. For example, in some municipalities, there are only two doctors, which makes access difficult in emergencies. Our quantitative data already showed gaps, such as limited childcare facilities, but through these conferences, we heard directly from families; who had difficulties finding a kindergarten place. These experiences and personal feedback added an important layer to the data, and by doing so, insights from these conferences are already feeding into concrete processes in Wittgenstein. They provided us both the big picture and the human stories behind the data.

Wagner also appreciated the positive impacts of the tool on citizens: “Most importantly, the process built trust. This came through strongly in the conferences. People told us, ‘For the first time, we feel like we are being listened to.’ That was a big benefit of using the tool we hadn’t originally anticipated when planning the conferences, but it emerged very clearly during the process. Residents could see their contributions appear directly in the final reports as what they said in the discussions was documented, visible, and traceable. That made residents feel that the planners and policymakers were truly listening and basing their work on citizens’ needs. And hearing that directly from citizens, especially those living in more remote, forested areas, was extremely meaningful.”

Reflections

For the district, the experiment proved that digital tools can strengthen democratic participation when used carefully and transparently. The AI did not replace human facilitation but enhanced it, freeing moderators to focus on dialogue quality while ensuring that every contribution was documented. Wagner summed up, “When people in the villages told us they finally felt heard, that made all the difference.”

 

Challenges and Learnings

Process design

  • Funding approach: The Giesskannenprinzip (“watering can principle”) spreads resources evenly, regardless of real, differing needs in municipalities. This limited targeted problem-solving. Shifting from equal funding to need-based planning helped resources reach communities with the most urgent challenges.

  • Hybrid participation works best: Combining face-to-face conferences with AI-supported digital tools enriched both citizen engagement and data capture.

  • Human stories complement quantitative data: Personal accounts from residents (e.g., families struggling to access childcare) added urgency and nuance to existing statistics.

  • Direct policy impact: Insights translated into actionable proposals, including projects addressing mobility, childcare, and access to healthcare.

AI benefits and adjustments

  • Privacy concerns: Data protection is highly sensitive in Germany, which requires transparency about how AI tools like deliberAIde handle and anonymize discussions.

  • Dialect and transcription accuracy: While AI worked well for standard German, strong regional dialects and noisy settings sometimes reduced transcription accuracy.

  • Information overload: Capturing every idea digitally created large volumes of data for the district, requiring new ways to filter, cluster, and prioritize insights.

  • AI facilitates inclusivity and transparency: In this case, automated transcription and clustering reduced facilitator bias and ensured citizen voices were documented and traceable in final reports. Citizens valued seeing their words reflected directly in reports, strengthening confidence that officials were genuinely listening.

  • Time and resource savings: AI-driven documentation allowed staff to focus on dialogue during events, while still generating reports within minutes afterwards.

Ready to explore further?