Can natural language processing and machine learning technologies foster quicker political response to citizen views on digital democracy platforms? With the Nesta Collective Intelligence Grants Programme, this is the central question we will be researching over the coming months.

In many citizen participation projects, the main challenge isn’t to gather citizen input: it’s to analyse it. In recent years, social media and new technologies have made it easier than ever for governments to reach out to citizens and collect their input on any given topic. As the amount of collected information grows, it is becoming increasingly difficult for administrations lacking in time and technical skills to process this data. As a result, meaningful insights are getting lost in the process and many citizen engagement initiatives fail to reach their goal.

Ideas posted on Vancouver’s citizen participation platform

At CitizenLab, we believe that helping administration process and analyse citizen contributions is an essential part of citizen engagement projects. Easing up analysis would not only save governments both time and money: it would help them tap into the collective intelligence of their constituents and make better decisions. This, in turn, could increase trust in public decision-making and encourage dialogue between governments and their citizens.

Applying Natural Language Processing (NLP) to citizen participation

In order to make this a reality, we’ve developed our own Natural Language Processing (NLP) techniques. This technology, based on machine learning and artificial intelligence, analyses large amounts of unstructured citizen input (ideas, comments, votes) and quickly extracts the main insights.

Clustering ideas on Vancouver’s citizen participation platform.

Ideas are automatically classified, grouped together or geo-referenced. Administrators of the platform can see at a glance what topics citizens are discussing, how the topics differ across different demographic groups, and how conversations are located around the town or region. It could be that in a neighbourhood, older citizens are asking for better roads whilst their younger counterparts want more public transportation. With reliable data at their fingertips, policy-makers are better equipped to make decisions and to design policies that truly respond to their citizens’ needs.

Locating citizen ideas on a map

If you’re interested in natural language processing and machine learning applied to citizen participation, we’ve written more extensively about how we’ve developed this technology in this article.

Digital transformation? It’s the human, stupid!

We’re keenly aware of the challenges that come with developing radically new technologies for governments. Firstly, our technology has to be both scalable and adaptable to the idiosyncrasies of different administrations. It needs to reflect the fact that classification models, language used and context can vary quite widely from one country or even one city to the next.

More importantly, governments are rigid structures where it can be challenging and long to implement change. A 2017 survey of civil servants in the United States showed that only 30% had “a high or very high interest” in implementing AI. More importantly, less than 16% had already done so. It isn’t enough to give cities a platform with NLP and machine-learning abilities: we need to deeply understand how civil servants work in order to offer a tool that seamlessly integrates with their existing workflows.

CitizenLab and Nesta’s Collective Intelligence Grant Programme

In line with this challenge, we are very excited to share that we have been awarded a grant for Nesta’s Collective Intelligence Grants Programme. This £20.000 grant is awarded to projects combining human and machine intelligence to solve social problems.

The Collective Intelligence grants programme is administered by Nesta, the innovation foundation that brings bold ideas to life to change the world for good.

We will use this opportunity to develop our natural language processing technology, focusing on the human aspect of the process. To our mind, this is where the real challenge lies: developing cutting-edge technology won’t help citizen engagement unless it is truly embraced by the civil servants who lie at the heart of the process.

In the coming year, we will be looking at how civil servants currently collect, analyse and use citizen-generated input, both manually and using our tool. We have chosen to have a comparative approach and will be following 6 cities of different sizes and geographical locations. By having an in-depth look at existing workflows and methods, we are aiming to refine our tool and develop a technology that is truly suited to civil servants’ needs and integrated with their processes and knowledge.

With a higher adoption rate of AI amongst civil servants, we hope to help governments increase their efficiency and process input more rapidly. With more reliable data available, governments will then be able to make better decisions and develop policies that truly fit their citizen’s needs, therefore reinforcing trust and legitimacy. This is the beginning of an exciting journey!

We will regularly report via our newsletter on the findings of the collective intelligence experiment, so make sure to subscribe if you haven’t yet.

More on this topic

There are currently no comments.