22 August 2023
AI tools made their debut on the participation scene years before the explosion of interest caused by the launch of ChatGPT. Yet as generative AI revolutionises one industry after the other, it’s clear that we can expect a big leap forward in how digital participation platforms incorporate the new functionalities into their tools. Democracy Technologies invited three expert practitioners in the field to tell us how AI is changing their digital participation tools, and to discuss the wider implications of the changes.
In particular, as AI grows more sophisticated and ubiquitous, the panel highlighted the need for the participation sector to lead the way in developing AI tools that adhere to rigorous ethical standards, and that keeps citizens’ needs at their centre. As Eva Mayer of CitizenLab put it, “we want responsible AI. We know that these models are not always ethically based. It’s a discussion that we need to continue having, in particular in our field.”
Current applications of AI in participation focus largely on making large volumes of data easier to digest. Participation project administrators are typically faced with inputs from thousands of citizens. Natural Language Processing (NLP) can help make these data sets manageable, giving them a structure that allows us to navigate them more rapidly and accurately.
DIPAS is the digital participation system of the City of Hamburg. It allows citizens to access detailed information about ongoing planning procedures and projects, as well as leaving feedback and exchanging with other users. The team first began exploring the use of NLP to help structure contributions in 2018. After experimenting with basic functionality including heading generators and clustering, they decided they needed a dedicated project to ensure.
“The whole project has become much larger than I originally anticipated,” said a representative of the City of Hamburg with the department of Urban Data Analytics, which develops the software for the Ministry of Urban Development and Housing. She noted that participation projects come with their own special requirements. The NLP needs to be able to handle specialist terminology, as well as local dialects and abbreviations. Additionally, extracts from user contributions need to stay connected to their original context to avoid misinterpretation. All while preserving both the anonymity and the autonomy of the contributors.
“The user of the analysis software should of course always be in control of what happens,” she added. “Every snippet from a contribution should always be linked to the original contribution so that we can grasp the entire context during the analysis.”
Integrating generative AI
CitzenLab has been using AI in its participation project for some years now. Their early AI model “Insights” was applied to projects in various cities around the world, including Vancouver, Canada, and Leuven, Belgium.
This system focused on the tagging and analysis of the thousands of inputs collected during the projects. But their new system, due to be launched later this year, introduces a new conversational engine based on generative AI. Alongside functions such as sentiment analysis and keyword analysis, users will be able to pose questions about the dataset to the AI, making it easier to identify trends and patterns.
“We want our new AI model to give governments an easy way to go through complex data,” explained Eva Mayer. “That means AI that supports true sense-making and reporting, creating additional transparency and pointing to the inputs that can help the local governments decide what to do next.”
Citizens Foundation in Iceland have also begun applying generative AI to help sort data. The non-profit organisation’s projects include Better Rekjavik, an online platform that combines agenda setting, participatory budgeting and policy crowdsourcing. Last year, the platform attracted 1,700 contributions. Citizens Foundation have begun using AI to help analyse the data set.
The foundation has also developed an experimental tool called “Policy Synth,” which points towards a possible way of using AI beyond making data more accessible. Beginning with an initial statement of a problem provided by the user, the programme uses AI to generate search terms in response to a and to then compile a set of subproblems and proposed solutions. A test case can be viewed here.
Citizens Foundation CEO Robert Bjarnason explained: “after we have this first set of solution solutions, we use AI with human feedback to do this sort of genetic algorithm where we evolve solutions and we change them, we randomly mutate them, we bring a new solution from the web and we create pros and cons and so on.”
AI is already well on the way to reshaping the everyday work of participation administrators. The panellists were keen to emphasise that the purpose of the technology is not to automate decision-making processes, but to improve the quality of communication between citizens and project administrators. They emphasised the role the democracy technologies sector will have to play in developing human-centric, responsible AI.
“The AI we have now is definitely nowhere near as smart as humans today, and definitely not as collectively,” commented Bjarnason. “But even if we did have AI that was smarter than us, we would definitely want to align it with human interests all the time.”
While generative AI performs increasingly well when giving answers to general questions, participation aims to benefit from the input of those who know their communities best – local residents. “We build on the local knowledge and ideas of real people; if they choose to use AI components to come up with new ideas to improve the planning process, that’s fine too!,” said the Urban Data Scientist from the DIPAS project.
AI-based participation tools are not about automating decision-making processes. Rather, they have the potential to allow governments to hear individual voices on a scale unthinkable without the technology. Some very big steps in this direction have already been taken within the industry. Watch this space.
If you want to be the first to know about future events, sign up for the Democracy Technologies Community.