In 1787, General George Washington was reluctant to intervene in discussions about the constitution of the newly formed United States. He knew that his voice carried tremendous influence and he was reluctant to sway deliberations. But a single issue was so important that he weighed in: the ratio of members of the House of Representatives to the number of citizens they represented.
Authors of the Constitution proposed one representative per 40,000 citizens. Washington rose to make the one substantive comment during deliberations, noting that the “smallness of the proportion of Representatives had been considered by many members of the Convention an insufficient security for the rights & interests of the people” and suggesting that 30,000 to one would increase the chances that the new Americans could have their opinions heard by their representatives. His suggestion passed unanimously, and the erasure of 40,000 and replacement with 30,000 is visible on an original copy of the US constitution.
The fundamental challenge of representative democracy is listening at scale.
Since democracy’s Athenian roots, the job of a representative has involved listening to the concerns of their constituents and then making decisions by deliberating with their peers. Deliberation allows representatives to compromise and settle on workable solutions, while listening ensures that decisions are aligned with the priorities of those represented.
But listening to more than a handful of people is hard. Athenians solved the problem by having hundreds of representatives to listen to only a few thousand citizens. (Most people who lived in Athens were not citizens—less than a fifth of 4th-century Athenians, or 40,000, were male citizens who’d completed the required military training to be eligible to vote.) The number of representatives—500 in the executive boule, 6,000 in the popular assembly or ecclesia—was limited by the technology of the time: the amphitheatre, which allowed one representative to address hundreds but not tens of thousands of others.
A few countries have maintained Athenian ratios of representation—in Iceland there are roughly 10,000 citizens per parliamentarian, offering the possibility of personal representation. But such a thing is no longer possible in the United States, where the average representative speaks for more than 770,000 citizens. The UK is between the two, with roughly one MP per 92,000 citizens or 68,000 parliamentary voters.
How does one MP listen to almost 100,000 voices? The cynic might argue that many representatives don’t, and even the most optimistic would acknowledge a significant selection bias in the citizens who have time, knowledge and motivation to write a letter to a representative or call a constituency office with a concern.
There’s a secret history of democracy told through the technologies of “generative listening”, a term Eric Gordon, a professor of civic media at Emerson College, uses to explain the challenges of listening to constituents and responding constructively. In the 1600s, petitions emerged in England as a way for a group to demonstrate the force of their opinion through assembling thousands of signatures in support of a civic demand. One notable petition in 17th-century England collected the signatures of women who demanded the banning of “that drying, enfeebling liquor”: coffee. Historians believe many signatories were wives worried their husbands’ presence in newly opened coffee shops would lead them to become politically engaged, potentially threatening social and familial stability.
Despite their venerability, petitions were the technology the Obama administration turned to in 2008 as a way of listening to constituent concerns. The White House promised a response to any petition that received 10,000 signatures. They quickly discovered that it is far easier in the age of the internet to assemble signatures online than it was in an era of paper petitions.
Many petitions went viral, notably one demanding the US invest in building the Death Star from the Star Wars films. The Obama administration dutifully complied with the promise to respond, explaining that they would never waste taxpayer money on planet-destroying technology so poorly designed it could be destroyed with a single proton torpedo shot into an exhaust vent. But they also substantially changed the number of signatures needed to require a response and the project as a whole was quietly abandoned during the Trump administration.
Polling, the much-maligned art of interpreting public will through random sampling and surveys, was not originally intended to predict the outcome of political races but to measure opinion at scale. George Gallup, perhaps history’s most celebrated pollster, honed his techniques by surveying readers on what they wished to see in their newspapers. He measured customer preferences around toothpaste and automobiles for years before daring to apply statistically adjusted survey methods to questions of civic importance like opinions on foreign policy.
Part of polling’s appeal is the promise of scientific rigour that journalism—a form of civic listening explicitly protected in many democracies—often falls short of. Journalists are meant to identify matters of public issue and importance and help representatives and citizens alike hear the voices of affected citizens by amplifying their stories. But it’s easy for the stories shared by journalists to reflect the political leanings of reporter or editor, or simply feature the voices of those a reporter already knew to listen to.
Polls allow us only to hear answers to the questions we choose to ask
Polling proposes a scientific fix—if we know the population includes so many citizens of a particular ethnicity, occupation, education level, political preference or gender, we can accurately listen at scale by surveying a sample of citizens whose demographics are the same as that of the population as the whole. And if we cannot find enough welders who are recent immigrants, we give more weight to the responses of those we are able to survey.
Yet polls have proven brittle in recent years, underestimating support for Brexit in the UK and discounting Trump’s appeal in the past three US elections. It’s possible that old techniques for recruiting participants are systematically failing to reach citizens who are sceptical of institutions, a category that presumably includes pollsters. Or perhaps we are asking too much of polls, demanding they predict the future instead of allowing us to understand how respondents reply to the questions we choose to put forth. Polls allow us only to hear answers to the questions we choose to ask.
The popular embrace of social media as a participatory public sphere for civic dialogue opens another path for listening at scale: social listening. Bluefin, a firm started by MIT Media Lab professor Deb Roy, promised advertisers it could interpret consumer opinion from their posts on Twitter. The technology, which used machine learning to turn millions of individual statements into opinions about brands and products, was sufficiently exciting to Twitter that it paid millions of dollars for the business.
Roy finds himself, like George Gallup before him, moving from consumer preferences to civic listening. His research group at MIT organises community meetings, which it records and then transcribes with AI in order to analyse the opinions expressed. Roy’s systems combine the oldest of civic listening technologies—the public meeting—with some of the newest. While it’s unclear whether the perspectives amplified will change governance, the use of AI to enable listening at scale is worth a close and careful look.
The return of Donald Trump to the White House can be understood as part of a global wave of chauvinistic nationalism, a shift towards authoritarianism in a scary world, a scream of populist rage at social and governmental systems that many feel are failing them. But it can also be understood as a failure of our systems for civic listening.
Around 90 per cent of counties in the US—a subdivision of a state roughly analagous to a district or borough in the UK—shifted to the right in the 2024 elections, giving more votes to Trump in 2024 than they did in 2020. For the millions of Americans—myself included—who were surprised by this, we have to ask whether the systems that allow us to hear our fellow citizens are failing and how we might repair and rebuild them.
