64th ISI World Statistics Congress - Ottawa, Canada

64th ISI World Statistics Congress - Ottawa, Canada

Debunking misleading graphs – what method works best?


Sanne J.W. Willems


  • D
    Dr. W. Wijnker
  • P
    Prof.dr. I. Smeets
  • D
    Dr. P. Burger


64th ISI World Statistics Congress - Ottawa, Canada

Format: CPS Abstract

Keywords: communication, factcheck, graphs, misinformation, visualisation


The increasing use of graphs on social media enables quick understanding of complex information. But it also facilitates the spread of misinformation when graphs are designed to be misleading. How can we debunk these misleading graphs?

Graphs are very useful to communicate concisely about complex issues to a broad public. However, although they facilitate intuitive reading of data, trends, and predictions, there is a great drawback to graphs: the ease to mislead. Many misleading graphs pop-up on social media feeds and users take little time to read them. These hasty readers often draw wrong conclusions.

Violating design conventions is one way of creating misleading graphs. For example, omitting the baseline of the vertical axis of a bar chart results in an overestimation of differences between the bars, and is a common trick to exaggerate differences between groups.

Research on how to effectively debunk text-based misinformation has already resulted in some practical guidelines for fact-checkers. However, little is known about debunking misleading graphs. Therefore, we aim to fill this gap by studying different debunking strategies and their short- and long-term effects.

In our two-survey experimental study, we investigated and compared the effectiveness of four correction methods as debunking strategies to correct bar charts with manipulated vertical axes. The correction methods focus on different phases of graph reading, either correcting the misleading initial perception or stimulating accurate reading. Additionally, we investigated whether the correction effects last for at least a week and explore whether there are any differences between people with various levels of graph literacy and education.

In this presentation we will show the set-up, results, and conclusions of this comparison of debunking strategies. This study is part of a larger research project aimed at providing guidelines for factcheckers, science communicators, and (data) journalists on how to effectively combat misleading graphs.