Alberto Cairo on mastering the art of reading a chart
Oct 25, 2018 12:01 PM
“Good charts enable good conversations,” Alberto Cairo says. “Bad charts – or charts that are misinterpreted – hinder those conversations.”
It’s been said – and said again – that “every picture tells a story.” That “a picture is worth a thousand words,” that “seeing is believing.” But today at JMP Discovery Summit, Alberto Cairo – consultant, designer and Professor of Visual Journalism at the University of Miami – asks us to think again about those familiar adages.
Pictures, he says, can expand our perceptions, yes; but they can just as often hide more than they reveal. When viewers make inferences – or overlook missing information – they can easily come to an incorrect understanding of what the data really shows. Statistical thought experiments like Simpson’s paradox and the ecological inference fallacy evince this gap between appearance and reality.
“Good charts enable good conversations,” Cairo says. “Bad charts – or charts that are misinterpreted – hinder those conversations.” And in a world where data visualization is becoming increasingly ubiquitous as a way of communicating ideas, we all have a responsibility to become better informed readers.
Yes, that’s right: we have the responsibility, he says. The onus falls to consumers of visual media to interrogate and contextualize our sources – just as we might evaluate the potential biases of a piece of written journalism. Readers shouldn’t trust a chart to present data in a way that tells the full story. Nor should they trust themselves to interpret a chart correctly upon cursory glance alone.
But fair warnings aside, charts are important. They’re a part of our culture. They’re a powerful tool and even, perhaps, an art form. And that’s why Cairo is arming us with a checklist of questions to ask ourselves when separating the fact from the fallacy. Below, Alberto Cairo’s failsafe four-step checklist for next-level chart interpretation.
Is the data represented accurately?
Does the graphic include a sufficient amount of data?
Is uncertainty relevant? If so, is it revealed in a way that can be easily interpreted?