Skip to content
Citizen Infrastructure

How we picture disagreement

We are used to seeing public opinion as two bars, red and blue, pushed as far apart as the chart can make them. Lisa Schirch and Beth Goldberg argue that this visual form is not a neutral readout of a divided society — it is a design choice that manufactures the division, by erasing everything the two sides share and every shade within each. It feeds what researchers call the perception gap: the well-documented finding that partisans wildly overestimate how extreme the other side is. People guess that roughly 55% of the other party holds extreme views; the real figure is closer to 30%.

Opinion landscapes

Section titled “Opinion landscapes”

Now that AI can read thousands of open-ended answers at once, there is an alternative to the two-bar chart: an opinion landscape — an interactive map of the full distribution of what people actually said, showing the clusters, the spread, and, crucially, the overlap. As the authors put it: “A bar chart showing a 30% gap between the parties tells one story. But a distribution chart showing the 70% of the area where those groups overlap tells a story that emphasizes their commonality and is perhaps more truthful.” The claim is not that the opinions change — it’s that the picture of them does, and that a fuller picture is both less polarising and more accurate.

The evidence that form matters

Section titled “The evidence that form matters”

The case rests on a run of studies where changing only the visual shifted perception: colouring an electoral map purple rather than red-and-blue reduced perceived division; party-labelled polls pull respondents toward their “team”; repeated exposure to the range of opposing views lowered extremity; and a 25-treatment megastudy found that correcting the perception gap reduced partisan animosity. The clearest working example is the We the People deliberation: rendered as an opinion landscape rather than two opposing camps, three in four of a nationally representative group reported understanding opposing views better after under an hour online.

Why it belongs to citizen infrastructure

Section titled “Why it belongs to citizen infrastructure”

This is the visual side of bridging and uncommon ground: where those find the statements that win support across clusters, this finds the picture that lets people see it. It pairs naturally with AI sensemaking, which produces exactly the distributional data an opinion landscape needs. The takeaway for anyone building a participation platform or reporting on public opinion is blunt: visualisation is never neutral, so choose the form that shows the whole distribution — including the overlap — rather than the one that maximises contrast.

Sources

Section titled “Sources”