Author: Margaret Hellner
Imagine a city proposing to end bus service at 8 p.m. A hospital cleaner submits a comment explaining that her shift ends at 11 p.m. and the bus is her only affordable way home. Officials use artificial intelligence (AI) to summarize residents’ responses, and her objection appears under “passenger inconvenience.” The council receives an orderly report, but the problem of getting home after a late shift disappears into a vague category. AI has made consultation easier to process while making the consequences harder to see..
Now suppose riders meet, compare work schedules, and propose retaining departures at 11:15 p.m. and midnight. The cleaner discovers that restaurant workers face the same problem, and they agree to attend the council meeting together. These relationships illustrate Putnam’s social capital: networks and trust that help people cooperate (Blaug and Schwarzmantel 2016, 377). This supports Cohen and Arato’s argument that democratic discussion must allow people to explain their circumstances and consider others’ perspectives (Blaug and Schwarzmantel 2016, 372). The resulting proposal comes from working through a shared problem, not simply generating language about it. This raises the question of who gains power when that process is outsourced. In this example, calling the problem “inconvenience” rather than “access to employment” changes what officials get to defend. If residents cannot challenge that description before a vote, the city controls both the decision and the account of public opinion used to justify it. A correct count of objections would still leave the council insulated from the reasons behind them. Cohen and Arato’s emphasis on actual dialogue helps explain why collecting comments is insufficient when citizens cannot contest how their concerns are represented (Blaug and Schwarzmantel 2016, 372).
Castells’s account of Occupy Wall Street, a 2011 protest movement against economic inequality, shows what is missing from this arrangement. Participants connected online and occupied physical spaces; the wider network of movements also used local assemblies for debate and decision-making (Blaug and Schwarzmantel 2016, 434–35). Their political activity involved people confronting questions together. Like the bus riders organizing around their schedules, participants could build the cooperative relationships Putnam describes. Substituting an AI account of their views would remove opportunities to argue over what the group actually wanted. That danger also matters across borders. Castells places Tunisia’s uprising against dictatorship and Occupy Wall Street within a wider wave of protest, while distinguishing resistance to dictatorship from protests against governments protecting financial elites (Blaug and Schwarzmantel 2016, 434). If an AI-generated solidarity statement described both simply as demands for “better economic opportunities,” opposition to authoritarian rule could disappear. Putnam’s movement from “I” to “we” (Blaug and Schwarzmantel 2016, 377) becomes politically troubling when that “we” erases differences between participants. A shared identity should emerge through discussion about whose demands a movement will support.
Democracy requires officials to answer the cleaner’s actual objection, just as international solidarity requires activists to hear one another’s demands. An AI-generated account should not stand in for either conversation. For Model UN delegates, representing communities means listening to their actual demands. When efficiency means citizens become easier to categorize and harder to hear, it serves institutional convenience while weakening their ability to challenge power.
Bibliography
Blaug, Ricardo, and John Schwarzmantel, eds. 2016. Democracy: A Reader. 2nd ed. Edinburgh: Edinburgh University Press.