Cindy Xiong
Cindy Xiong is the first author of 'The Curse of Knowledge in Visual Data Communication', published in IEEE Transactions on Visualization and Computer Graphics in 2020 with Lisanne van Weelden and Steven Franconeri. In four experiments with small student samples, people who knew a pattern in a graph predicted that naive viewers would see it too, even when told to ignore what they knew. The study extends the curse of knowledge from instructions and teaching to charts.
Who they are
The vault records Xiong as a visualisation and cognition researcher. The copy of the paper read for this site was an accepted manuscript hosted by a Northwestern University visual cognition lab, where Franconeri's group works. Her later positions and wider publications are not reviewed here. The page rests on one paper, whose volume, issue and pages were confirmed at Crossref: volume 26, issue 10, pages 3051 to 3062.
Key papers
Key ideas
Knowing the story changes what you think others see
Viewers can extract many patterns from the same chart, so two people may take different messages from it. In the experiments, participants first read one of several backstories about political events affecting polling data. The backstory made one pattern in the graph especially salient to them. They then viewed the graph and predicted what a naive viewer, who had not read the backstory, would find most salient.
Despite explicit instructions to ignore their own knowledge, participants predicted that naive viewers would find the primed features salient. They also rated the features they predicted for others as more salient to themselves. The authors frame this as the curse of knowledge: experts, or anyone who knows the story, struggle to imagine a novice's state of mind, which helps explain why presenters fail to connect with audiences.
The experiments
The four experiments were within-subject designs. Experiment 1a had 18 students and used a line graph with visual annotations. Experiment 1b had 29 students and a line graph without annotations. Experiment 1c had 21 students. Experiment 2 had 17 students and a bar graph stacked by political stance. In each, participants predicted up to five features a naive viewer would find salient. The effect was weaker without annotations, and some participants did not show the bias at all.
Caveats
The paper notes that the analyses in Experiment 1a were post hoc, that all participants were college students, and that the number of participants in each experiment was small. The findings are therefore best read as consistent evidence from small samples. They do not give an effect size for how far knowing the story distorts a presenter's prediction in practice. See Curse of Knowledge.
Sources
- Xiong, C., van Weelden, L., & Franconeri, S. (2020). 'The curse of knowledge in visual data communication.' IEEE Transactions on Visualization and Computer Graphics, 26(10), 3051-3062. doi:10.1109/TVCG.2019.2917689