Same Data, Diverging Perspectives: The Power of Visualizations to Elicit Competing Interpretations

TL;DR

This work investigates how visualization design shapes pattern saliency and decision making, arguing that interpretations of data can be multi-stable akin to ambiguous figures. Through four experiments, the authors demonstrate that chart type (bar vs table vs line) and storytelling techniques (annotation vs highlighting and recoloring) bias what patterns readers notice and the conclusions they reach, with annotation showing the strongest causal influence. They identify intra-personal consistencies in pattern extraction across visualizations and reveal that visual saliency interacts with individual differences to guide interpretation. The findings have practical implications for visualization design and ethics, suggesting the need for careful, potentially guideline-driven storytelling to avoid bias while still communicating data effectively.

Abstract

People routinely rely on data to make decisions, but the process can be riddled with biases. We show that patterns in data might be noticed first or more strongly, depending on how the data is visually represented or what the viewer finds salient. We also demonstrate that viewer interpretation of data is similar to that of 'ambiguous figures' such that two people looking at the same data can come to different decisions. In our studies, participants read visualizations depicting competitions between two entities, where one has a historical lead (A) but the other has been gaining momentum (B) and predicted a winner, across two chart types and three annotation approaches. They either saw the historical lead as salient and predicted that A would win, or saw the increasing momentum as salient and predicted B to win. These results suggest that decisions can be influenced by both how data are presented and what patterns people find visually salient.

Figures (9)

• Figure 1: Visualizations contain many patterns one could focus on. Visual annotation can guide people's attention to patterns they might otherwise miss. Left: Annotation showing that the green lines are mirror images of each other. Right: Annotation showing that the bottom two lines cross each other twice. Adapted from Xiong et al., see xiong2019curse.
• Figure 2: Set-up for Experiment 1. Participants first read the table or the bar chart, predicted a winner and provided justification. Next, they matched their justification with one of four options we offered.
• Figure 3: Results from Experiment 1a. Left: distribution of slider responses on election outcome prediction for bar and table. The color encodes the specific green or blue supporting features participants chose as their reasoning for their predictions. Right: prediction results for binary forced choice task from bar and table conditions. Error bars are calculated by estimating standard errors for proportions and then multiplied to a total sample size.
• Figure 4: Left: Distribution of slider response on election outcome prediction. The color encodes the specific green or blue supporting features they found salient, see Figure \ref{['fig:exp1']}. No participant indicated the pattern encoded by dark green as visually salient. Right: Summary of binary decisions, showing a congruence between salient patterns with predictions of which party would win. Error bars represent standard errors calculated by estimating standard errors for proportions and then multiplied to a total sample size.
• Figure 5: Procedure for the line chart portion of Experiment 2.