Average Estimates in Line Graphs Are Biased Toward Areas of Higher Variability

Area of Science:

• Cognitive Psychology
• Human-Computer Interaction
• Data Visualization

Background:

• Line graphs are common tools for data representation.
• Existing research highlights various visual perception biases.
• A novel bias, variability overweighting, in line graph interpretation was previously undocumented.

Purpose of the Study:

• To investigate and document the variability overweighting bias in line graphs.
• To explore methods for mitigating this bias in data visualization.
• To understand the potential cognitive mechanisms underlying the bias.

Main Methods:

• Conducted two preregistered experiments with a total of 560 participants.
• Participants estimated average values from line graphs.
• Compared bias magnitude between line graph encoding and dot encoding of the same data series.

Main Results:

• A consistent variability overweighting bias was observed in line graph interpretation.
• Estimates of average values were significantly biased toward regions of higher line variability.
• Switching to a dot encoding significantly reduced the observed bias.

Conclusions:

• Variability overweighting is a significant bias affecting accurate average value estimation from line graphs.
• Dot encoding offers a potential solution to mitigate this bias.
• Findings have implications for designing more effective and less misleading data visualizations.