One of the most common and least political of all the ways citizens attempt to wield power over government is by giving feedback on public services. Whether complaining about a teacher to a school or reporting a broken streetlight to a local government, “complaining” services play an important role in citizen engagement. While often unglamorous in nature, these moments can be the gateways to deeper engagement later, giving citizens their first experience in demanding better results from public authorities (Bode 2017).
Combinations of various current and emerging technologies will change this. Traditionally, for illicitly dumped trash to be reported, a person would have to spot the trash and contact the authorities to report it. In the future, cameras and other sensors on passing municipal vehicles, like police cars and school buses, will pick up two- and three-dimensional data traces of the trash. The data will be analyzed by machine learning systems specially designed to recognize out of place objects. These systems will then generate and prioritize clean-up tasks. Finally, clean-up crews will be dispatched to clean it up (where governments can afford this service).
This is only just one way in which semi- or fully automated data systems will spot problems before citizens get around to it. Dirty or undrinkable water will be detected by sensors plugged into the water networks. Broken public infrastructure (such as street signs) will be detected most likely from the camera footage taken by municipal vehicles. Noise pollution from houses or factories will be identified by microphones already embedded in a huge variety of devices. Even teacher or police underperformance might be detected automatically and remotely through different kinds of data analysis.
The more data you have, the less participation and voluntary input from citizens will the government need.”
The net effect of an automated data system will be to give local governments adequate resources to detect and solve more problems before local citizens have to report them. The number of people who make complaints or report problems will be reduced. It will also be more convenient for individual citizens, saving them time and effort. This convenience, however, may come at a price.
Hollie Russon-Gilman, lecturer in technology and public participation at Columbia University, assesses the potential situation as worrisome. “If people don’t see what prompts the government to respond, they will fail to understand a basic dynamic of participation, because the nexus between the citizen’s voice and the government’s response tends to disappear.”
This change seems likely to accelerate as governments start to leverage troves of data to engage in predictive responsiveness — using data analysis to prevent problems before they emerge. For instance, local governments in Asia and Latin America have been experimenting with tools that combine different sources of data, such as from weather monitoring and Twitter, to predict dengue outbreaks (Marques-Toledo et al. 2017). In a similar vein, Kansas City in the United States is experimenting with AI solutions to predict where potholes will occur, and the City of Chicago has been systematically using predictive analytics for food inspections and to combat rodents. All these point in the same direction for citizen engagement.
In the words of MIT Professor Cesar Hidalgo, “The more data you have, the less participation and voluntary input from citizens will the governments need.”
If automation could eliminate the need for citizens to report municipal problems, could it also eliminate citizens need to partake in deeper forms of participation?
Reporting a vandalized bus stop and voting in a general election may not seem to have much in common. However, they both are ways in which citizens express to governments their desires for action through official channels that are built and maintained by those same governments.
Some thinkers have started to argue that if automation can do away with trivial citizen feedback like pothole reporting, then it may also be able to do away with more weighty kinds of citizen feedback. For instance, Hidalgo has recently suggested an experimental system using AI-powered representatives, commonly called digital twins, to increase people’s ability to take part directly in legislative decisions (Hidalgo 2018). Drawing from data on the user’s preferences and behaviors, the system would predict how the user would vote on a bill being discussed in a given congress or parliament. 16 Most people would probably react with horror at the thought that machines might simply vote for them. But foundational research is already being done that could enable this dystopian vision. Several projects have been carried out to predict individual voting and policy preferences drawing from digital behavioral data, with some success (Kristensen et al. 2017). For instance, researchers have been able to determine political preferences with growing levels of accuracy based on Facebook likes. Even seemingly insignificant online actions, such as liking “Harley Davidson” or “Hello Kitty,” can tell a lot about an individual’s wants and political leanings (Kristensen et al. 2017). As individual data trails get longer and more detailed, and as machine learning techniques get steadily better, it seems likely that the ability of computers to predict people’s political beliefs based on their online activities are only going to get better.
It is not hard to imagine governments observing this, and then making the case that they want to analyze the social media data of citizens to be more responsive and do what the people truly want rather than deciphering the blunt, vague signal that a vote gives. It is even possible to imagine scenarios in which social scientists start to provide evidence that the desires and intentions of citizens detected by computers are more legitimate, more granular, and in a sense more “true” than traditional mechanisms such as votes, petitions, or polls.
Few people appear to be enthused by this scenario. Ben Berkowitz, founder and CEO of SeeClickFix, conveys a sense of dismay in the face of a radical automation scenario. “There is something really sad about it. The experience of a human that can change something, with automation, that goes away. I don’t know when the moment comes when we realize we have gone too far — and we have lost the capacity to provide that moment of empowerment.”
16: The model put forward in this case is experimental and with nonbinding effects on the actual lawmaking process. While this may still seem fanciful, AI-politicians are starting to emerge, if primarily as publicity stunts. In the recent 2018 Russian elections “Alisa,” an AI-powered virtual assistant developed by tech firm Yandex, ran for president. Employing slogans like “the presidents who knows you best” and “the political system of the future,” more than 80,000 people voted on Alisa’s website to nominate her for the presidency. In last year’s mayoral election of Tama, a city in Metropolis Tokyo, a robot named Michihito Matsuda received more than 4,000 votes, with the campaign based on the promise that AI would change Tama. New Zealand’s virtual politician SAM expects to run for the country’s next general election in 2020. According to its official website, SAM is powered by citizen “views, values, and opinions, not just data.” For more information about these examples see https://www.themoscowtimes.com/2017/12/07/artificial-intelligence-robot-alisa- nominated-for-russian-president-a59845; https://www.softcarecs.com/artificial-intelligence-robot- alisa-is-nominated-for-russian-president/; https://u.today/robot-secures-4000-votes-in-mayoral- election; and http://www.politiciansam.nz/. ↩