Skip to content

Human rights data used the wrong way can be misleading

While data is important for human rights advocacy, the risks of misleading people are also very real and advocates must insist on rigor.

Published:

In a world of evidence-based policy and data-driven decision-making, it’s time for the human rights advocacy world to more fully embrace new methods. Of course, the construction, use, and reliance on quantitative indicators in human rights settings are all rife with danger. But the promise of using data to understand rights problems, their causes and solutions, is too great to pass up. The key—as I discussed with Todd Landman in a recent episode of The Rights Track—is critical engagement.

In this context, the Data Visualization for Human Rights project, based at the Center for Human Rights and Global Justice (CHRGJ), grew out of a desire to understand how human rights work could more effectively harness the power of visualizing data and data stories to advance rights fulfilment. Through a series of randomized user experiments, we have learned that data visualization is a powerful persuasive tool that needs to be carefully tailored to its audience and context. In one study, we found that readers who had a strongly negative opinion about a rights issue were less likely to find charts and graphs persuasive and more likely to be persuaded by rights data presented in a table. Those readers who came to the topic without strong opinions were, on the other hand, more persuaded by data displayed through graphs and charts.

These findings suggest that human rights advocates would be wise to analyze their various audiences and tailor their presentation of data accordingly. In another randomized user experiment we conducted, we found that it is strikingly easy to mislead readers when using deceptive visualization techniques. By inverting the y-axis, for instance, we could successfully reverse the message a reader perceives even when the data was reported accurately. Similarly, common techniques like starting the y-axis at a value other than zero can mislead the reader into seeing much greater values than the data support.