Your 10,000 Twitter Mentions Are Worthless

Your 10,000 Twitter Mentions Are Worthless

Picture a city council member with a vote scheduled for next week. The bill is locally contentious. Two piles of input land on the desk.

The first pile is large: ten thousand mentions, comments, and replies online. The post went semi-viral. The second pile is small: eighty responses from people who have been verified as actually living in the district the council member represents.

Almost everyone's instinct is that the first pile matters more. It's a hundred times bigger. It's the public reacting. The eighty feel like a rounding error next to it.

The instinct is wrong, and understanding exactly why is the difference between a feedback channel that works and one that quietly poisons every decision that runs through it.

What the big number actually measures

The ten thousand mentions measure one thing: how far the post traveled. That's it. Reach is not opinion, and it is certainly not constituent opinion.

Strip the pile apart and it falls to pieces. Most of the ten thousand don't live in the district — they don't live in the city, and a meaningful share don't live in the country. Some aren't people at all. Of the ones who are real and local, they skew hard toward the most activated: the angriest, the most ideological, the ones already plugged into an organized campaign that routed them there. Considered agreement doesn't generate engagement, so it's invisible in the count. Outrage does, so it's overrepresented. The number isn't a measurement of what the district thinks. It's a measurement of what spreads.

Treating it as constituent signal is a category error. It answers the question "what went viral," and the official is asking "what do my voters want." Those are different questions, and the big number happens to be a precise answer to the wrong one.

We already know how to do this everywhere else

Consider how you read a restaurant's reviews. A place with 4.7 stars from sixty verified local diners is more trustworthy than the same place at 2.1 stars after nine thousand people who never ate there brigaded the listing over an unrelated controversy. You don't need this explained. You instinctively weight by whether the reviewer actually has standing — did they eat the food — and you discount volume that doesn't.

We apply this reasoning to lunch without a second thought, then abandon it completely the moment the stakes become democratic. We let an official's sense of "what people think" be set by whatever cleared the algorithm, with no check on whether any of those people are the ones the official answers to. It's a standard we'd find absurd for a burrito and accept for a budget.

Weighting is not silencing

The fix is not to throw out the ten thousand. It's to stop pretending they're one undifferentiated mass and separate the input by standing.

A serious system shows an official three numbers, not one. In-district: the people you actually represent. Out-of-district: everyone else engaging with your work. And the total. None of these is deleted; they're disaggregated, because they answer different questions.

The out-of-district number isn't garbage, either — it's just a different signal. A flood of national attention on a local bill tells the official their work has salience beyond the district, which is real information, sometimes valuable information. The error is never that outsiders weighed in. The error is letting "the internet is loud about this" masquerade as "my constituents want this." Once the two are pulled apart, both become useful. Fused into a single number, both become noise.

The quiet cost of getting it wrong

Here's why this isn't an academic distinction. An official who responds to raw volume is, in practice, responding to whoever is loudest and best organized. And loudest-and-best-organized is not a neutral sample of the public — it's structurally tilted toward the already-powerful: advocacy operations with mailing lists, national campaigns that can manufacture a surge, anyone with the apparatus to flood a channel on command.

The working parent with a direct stake in the school-zoning vote and no campaign behind them produces exactly one signal: their own. In a system that counts volume, that signal is worth nothing against an organized push. In a system that weights by constituency, that parent counts as much as anyone who actually lives on their block — which is to say, correctly. Weighting isn't a tax on participation. It's the thing that lets an ordinary constituent count at all against people whose entire advantage is manufactured volume.

The reframe

The civic-input problem is almost always described as scarcity: officials don't hear enough from citizens. That's backwards. Officials are drowning in input. The inbox is full, the mentions are endless, the town hall has its regulars. There is no shortage of volume.

What's missing is provenance. None of that volume comes weighted by who's actually entitled to weigh in, so an official can't tell a constituent from a stranger, an authentic concern from a manufactured surge, a hundred real neighbors from ten thousand accounts that found the post. When you can't tell those apart, all of it collapses into the same undifferentiated roar, and the rational response to a roar is to tune it out — which is exactly what most officials, sensibly, have learned to do.

Signal was never going to come from more volume. It comes from volume you can trust the source of. Ten thousand strangers and eighty neighbors are not the same data. We've just built every channel we have on the assumption that they are.