Innovation Initiative Overview

Digital participation platforms are important tools to increase citizen integration and engagement and enhance government responsiveness. However, analyzing such high volumes of citizen participation on these platforms is extremely time-consuming and challenging for city officials. This technical difficulty can prevent them from learning valuable lessons. Therefore, creating a digital participation platform is not enough. It is also important to facilitate access to data analysis so that public employees can view the integrated information and make informed decisions.

The public sector is facing the same challenge on a large scale as our computing tech company. Deloitte has published a report on governments employing AI technologies, which concluded that natural language processing could help free up 1.2 billion hours of work and save up to $41.1 billion annually for governments worldwide. The UK government, considered the reference in terms of digital governments, cited in its 2020 strategy that a better understanding of citizen needs based on data and evidence is the highest priority for the next generation of governments. The UK's digital transformation consists of three key components: improved online citizen-facing services, improved efficiency to deliver citizen services across channels, and more effective digitally-enabled collaboration internally. BCG also reports that AI will boost the efficiency of democracies since governments will use data to create a detailed representation of the community's demands and amend and adapt public policies accordingly. These are all undoubtedly positive directions;  however, in reality, there is a huge gap not bridged by these directions, in addition to the challenges faced by the under-resourced and under-staffed public administrations.

CitizenLab seeks to bridge the knowledge gap that currently exists in the public sector. Most small to medium administrations understand the need for better operational processes and large-scale data analysis, but they lack the required expertise, means, and tools to generate custom solutions. Through this initiative, we aim to empower public employees and provide them with the necessary knowledge about machine learning augmented processes that will help them analyze citizen contributions, make better decisions, and collaborate more efficiently internally.

Concerning the initiative's technical details, CitizenLab has developed over the past year its own natural language processing techniques, allowing it to automatically classify and analyze thousands of contributions submitted by citizens on the participation platforms. The algorithms identify the main topics and classify them by demographic trait or geographic location.  AI can analyze ideas regardless of the used language, which means that it works for multi-lingual platforms. Furthermore, the platform administrators can access this information through smart dashboards which present real-time data. The topic modeling feature makes it easy to identify the priorities of citizens and make decisions accordingly. It helps public employees understand the needs of citizens. For example, cities may ask their citizens to express their opinions on environmental matters but the comments would only tackle topics like transportation and taxes. Input classification by demographic groups and geographic location can provide the administrators with a deeper insight into how priorities vary. For example, one neighborhood prioritizes better roads, while another needs more traffic stops.

We believe that governments and citizens benefit from this innovation. By automating the time-consuming task of data analysis, our platforms free up time for administrators to interact with citizens. It lets governments better understand the needs and priorities of citizens, which leads to making better-informed decisions. This open and transparent process encourages people to trust their governments, increases support for public policy decisions, and pushes them to participate in the decision-making process.   We have launched our technological initiative across all available participating platforms and it is widely used by our clients. It has made a visible impact on the way they process insights and made them more confident to use and share the findings of the platform. The additional time provided by the automated analysis and reporting has allowed them to interact and collaborate with citizens to implement various ideas.

The next steps are to increase the adoption of this technology and to ensure that all our clients are making the best use of the automated dashboards. In the long term, this technology can be implemented on a larger scale to include social media platforms, public forums, and other places dedicated to public discussions. The recent example of the Grand Débat in France has demonstrated the importance of this technology: Without relevant and trustworthy data analysis, these large-scale debates would not take place and the engagement of citizens would not be possible.

Innovation description

What makes your project innovative? 

Citizen participation platforms are almost always limited to collecting the contributions and opinions of citizens. They help governments gather input from citizens but do not offer them any support whatsoever in analyzing this input. The lack of support in this crucial task will lead to undetected insight which can be extracted from these contributions and citizen engagement will not generate the expected impact.

This is when our platform, which provides the necessary analysis for this input, comes into play. By using machine learning to analyze the collected ideas, we are able to provide an integrated service for governments. The analysis is completed within one single platform, making it easier to supervise the projects. This feature increases efficiency, decreases administrative costs linked to citizen engagement, and leads to better decisions.

Finally, we have combined our technical expertise and deep understanding of citizen engagement. By leveraging our experiences in the public sector, we have refined our algorithms to reach the most crucial information that cities need for their decision-making process.

Innovation description

Participating entities

Finally, we relied on the expertise of our team and public sector experts to determine our product's scope of impact and its benefits for the public sector. We collaborated with consultants in natural language processing. They helped us design an effective product which we were then confident to implement and share with governments. We have also contacted many cities and public employees to understand their needs and ensure that the product meets their requirements.

Users, stakeholders, and beneficiaries

CitizenLab targets two types of beneficiaries: governments and citizens. In the short term, administrations who use this platform mostly benefit from this service. Public employees have been able to save a lot of time by accessing information and collecting insights. In the long term, citizens are the ones who will feel the positive impact of this innovation. Thanks to their contribution, governments can make informed decisions and enhance their services.

Innovation impact

Results and impacts 

Since this product was launched in late 2018, we have witnessed cases where the automated analysis has left a true impact on local administrations and their relationship with citizens.

For example, the Belgian city of Kortrijk uses smart dashboards to easily process the contributions of 1,300 users on the platform. The latter has gathered and classified the ideas into main topics to reveal the result of the discussions. The city shared the results with its citizens, proving that this process is a true dialogue rather than a top-down initiative. The city of Temse also consulted with its citizens on transportation and located the ideas on a map of the city. This helped the administration determine where the main challenges were focused and where funds needed to be allocated. CitizenLab currently helps YouthForClimate to process 4,000 ideas posted on the participation platform and turn them into 16 public policy recommendations. On a large scale, topic modeling has helped identify the most important topics tackled by participants. 

Challenges and failures

Our initiative faces two main challenges: classification algorithms and the level of human adoption.

Classification algorithms gather, categorize, and summarize citizen input. Although the algorithm should be flexible and scalable, it must also adapt to the different administrations' workflow since used classifications differ by country and by region. Our classification algorithms should also support several languages on the same platform and establish semantic links between languages, which further complicates the technical aspect of this algorithm.

On the human side, a clear workflow must be planned and the technology must meet the real needs of the users to maximize adoption by the administrations. We have learned that we cannot sell a product before informing the users of its benefits. Moreover, the interaction between humans and machines is crucial. How can we interpret and "trust" the output of the machine? What role can outputs play in people's workflow?

Conditions for success

The quality of the contributions is the first condition for the success of this initiative. This technology relies on clear and detailed contributions from citizens. This means we need to ensure that citizens are guided to share the right types of contributions.

User adoption is the second condition. In order for the tool to be adopted, it needs to be trusted and easy to use. Public employees should understand its benefits and feel that they can rely on its results. We can achieve this by enhancing the user experience, explaining the adopted approach, and ensuring it is integrated with the existing tools and workflows.

Finally, large-scale regulatory evolutions can be a critical factor in the product's success. If citizen engagement was applied across states and regions, cities will be encouraged to invest in our platform. Consequently, a virtuous circle is formed:  the more cities use the product, the more algorithms can improve and the better the product becomes.

The possibility of replicating the experience

The increased appetite for citizen engagement in decision-making results in a need for automated data analysis. While many citizen participation platforms are being established all around the world, very few have integrated analysis capacities that can be leveraged. As we have noticed with the difficult analysis of the Grand Débat contributions in France, this is a widespread challenge that prevents citizen contributions from leaving a real impact on the decision-making process. This technology can be replicated on any other platform. There is also another benefit to using our product. Since we work with many cities, our algorithms have learned how to process multiple data sets which are more efficient than any other one-time local solution.

In the long term, this technology we are developing to analyze multi-lingual contributions can also be used to process various discussions on social media platforms or public forums. It can also help governments understand the needs of citizens on a larger scale and adapt policies to meet those needs.

Lessons learned

Throughout this process, we have had the chance to learn a lot about technology and the human factor behind the technology.

Regarding technology, natural language processing and machine learning are evolving at a great speed. Off-the-shelf commercial products can be very useful, but their benefits do not last.  We decided early on in the project to invest in our own technology and create a product that we can fully control. This has allowed us to adapt to changes and different markets, and discuss openly with cities how the technology works and its development stages. Thanks to our in-house team of experts, we have the freedom to experiment and improve the product continuously.

We also found that the project is worth all the time and money we invested in the initial research. The decisions made in the initial stages when building the algorithms have a decisive influence in shaping the technology later on. The same applies to languages and training models. It is easier to migrate to some languages than others. Therefore, make sure to choose the right language at the beginning. Moreover, the thresholds you establish early on in the project are also a crucial factor in the evolution of algorithms and their level of accuracy.

We have learned that no matter how good the technology is, what truly matters are the humans behind it. In order for the product to succeed, public employees must show they are interested and trust that it will generate reliable results. Employees are concerned with the results achieved by the tool, rather than the modern technology it was based on, and this is the principle we have adopted while communicating with our clients. We have exerted great efforts to make the platform as easy to use and as results-oriented as possible.

Therefore, while developing a public-oriented tool that, make sure it serves a clear and defined purpose. Administrations lack the time and resources, and public employees will only invest in tools that have proven their efficiency. Do not forget to test the tool regularly in the development stage to make sure it aligns with user needs. Finally, the market may need the product but people may not be aware of it and they might not believe in it at the beginning. Therefore, we should be prepared to educate the users and convince the public sector of its importance.

There is also a very important human factor while organizing the contributions and opinions on the platform. Machine-learning processes rely on clear and detailed contributions but that's not how most contributors write. Therefore, we test and edit our platform constantly to help users submit the input format that we need. Just like public employees, we must convince citizens of the platform's benefits. Therefore, we have helped cities highlight the platform's advantages and develop a clear message about the importance of engagement.

Before concluding, it has been a unique experience to see the sector develop alongside the tool. We had witnessed the growing demand for citizen participation platforms, and we are now noticing an increased interest in automated data analysis. Public employees have become more aware of matters related to data protection. The fact that we are being asked tough questions on the subject in preliminary meetings is a good sign. Consequently, we recommend anyone launching a similar product to maintain the highest security and ethical standards.

References: OECD OPSI. (2018). https://oecd-opsi.org/innovations/unlocking-the-potential-of-crowdsourcing-for-public-decision-making-with-artificial-intelligence/