Algorithms for democracy? – Identifying pitfalls    

An increasing proportion of people feel removed from the parliamentary system, with very little chance of participating in and influencing political decision making. Conveying the voice of citizens in decision-making requires a diversity of forms for participation. We need new ways to reach and involve people.

Sitra, together with 12 Finnish municipalities, experimented with robot phone calls combined with AI-assisted analysis to identify citizens’ concerns, ideas, needs and provide insights into the development of the municipality. We critically studied the whole process and created an analysis framework to help Sitra understand the social and democratic opportunities and risks associated with the process and its tools.

Robot phone calls and AI-enabled new ways of participation

“We believe that long-term decision making in society is improved by bringing together different opinions and ensuring cross-generational input. New forms of participation are needed to break the spiral of differentiation of participation. This is where modern technology can help”, says Joonas Leppänen, Leading Expert in Sitra’s New Forms of Participation program.

Participatory calls (“Osallisuuspuhelut”) were robot phone calls using novel technologies to establish new channels of citizen engagement to support traditional processes of communal democracy. Sitra ran an experiment with 12 municipalities across Finland, collecting ideas and insights from citizens with an open-ended question posed via automated calls to as many citizens as possible.

The citizens’ responses were then transcribed into text, anonymized, and analyzed using AI, resulting in themes and topics for municipal decision-makers to consider.

“We wanted to turn the democratic process around in two ways: First, the municipality, not the citizen, was the active party in this experiment. The citizens just needed to speak their mind. Second, within the pilot projects, the citizens had the possibility to set the agenda for political discussion – decide what should be discussed. The overarching aim was to add to the ways of participations and create more fluid ways assisted by technology”, says Joonas Leppänen from Sitra. 

Reviewing the process to ensure alignment with values of democracy

When taking modern technology in use, it is critical to understand the context, limitations, risks and biases affecting the desired outcomes. To ensure that the robot phone calls were made according to the principles of democracy and to learn more about the process and what should be considered in the various stages, Sitra asked us to help them in exaining the entire process.

The aim of our work was

1. to understand how the objectives of the democracy and participatory experiment were realized
2. to build transparency to and understanding of what really happens in the process
3. to produce a preliminary model and principles for similar audits in the future.

“First, we identified the values that should drive the process. What kind of values of democracy are at stake, what are the values we want to follow? What kind of things do we want and don’t want to see happen?”, says Antti Rannisto, Insight Lead from Solita.

“Then, our multidisciplinary team of experts created an analysis framework reflecting these values. Using the framework, we carried out a critical evaluation of the process by examining the materials, tools, actions, and outputs of each phase, and by interviewing people involved in the process as well as citizens who participated in the experiment. We took a careful look at the process itself, the role of technology in it, and human work involved. Throughout the process, we were assessing how well the experiment was aligned with the defined values”, continues Anna Metsäranta, Head of Sustainable AI from Solita. 

Good learnings to improve the process

In the assessment process we found both potential and realized risks and errors that would be likely to scale if moving forward. Some of our key findings point to the vast amounts of human labor, very traditional expertise and holistic oversight needed in the process of utilizing AI. This is often hidden from the picture when AI is falsely portrayed as automation of everything, especially when it comes to key human traits such as interpretation, in-depth understanding and controlled creativity.

We also highlighted the challenges of understanding citizens through such a complex chain of information production, involving multiple human and machine agents in the process, passing bits of information from one to another.

“It is critical that we can identify the weak points in this process. They can significantly distort the final outcomes and interpretations. Using technology to promote democracy we must ensure that all stages are meaningfully transparent and aligned with the kind of democratic principles and values we hold important. Our extensive critical evaluation revealed some very important challenges and sparked ideas for designing better alternatives in the future”, says Antti Rannisto.

“Robotic calls, when used correctly, can add to the portfolio of democratic participation. They aren’t a replacement to existing channels but can have a place as part of the bigger picture. The evaluation was very useful and provided lots of ideas for us. We now have a much better understanding of the social and democratic opportunities and risks related to the process and the tools”, says Joonas Leppänen.