Use bar encoding instead of line encoding when aggregation is fixed

When the same grouped summaries or paired-value rows must stay intact, switch from a line encoding to a bar encoding to weaken causal interpretation. For example, keep the same 2-, 8-, or 16-bin summaries but draw bars instead of connected lines, or use bars rather than paired line traces on nominal index displays.

Connected lines make a non-temporal relationship look more like a continuous trend, which can strengthen causal interpretation.

Mechanism: With the same data manipulation held constant, line segments emphasize continuity across values, while bars weaken that continuous causal-seeming trace. The change works only when the aggregation level does not also change.

Evidence: In Experiment 2, bar encodings received lower causation ratings than line encodings when both used the same binning levels. In Experiment 3, non-aggregated paired-value displays also showed lower causation ratings for bars than for lines (Xiong et al., 2020).

Notes: The paper also notes that two-bar bar charts appear to be a special case that can still invite strong causal interpretations.

• User Goal: Reduce causal interpretation without changing which values are summarized.
• Task: Compare two encodings of the same observed relationship.
• Data: Non-temporal paired quantitative values, either binned into groups or aligned by a nominal index.
• Chart Setting: A line encoding is being considered for the same values that could also be shown as bars.
• Audience: Readers judging whether one variable could be causing the other.
• Success Criterion: Lower causal agreement while preserving the same aggregation level.

Break it when: The bar version and line version do not keep the same aggregation level. Why: The paper found aggregation level had a larger effect and can overwhelm the mark-type difference.

Sacrifice: You give up the continuous visual trace of the line. Risk: A two-bar bar chart can still invite a strong causal reading. Mitigation: If you choose bars, avoid collapsing the data into only two groups.

Mistake: Replace a bar encoding with a connected line while leaving the same grouped summaries in place. Why it fails: With aggregation held constant, line encodings were judged more causal than bar encodings.

Failure Sign: The display connects non-temporal summarized values with line segments. Quick Check: Redraw the same values as bars and compare the two versions side by side. Stronger Test: Ask viewers to rate a causal statement for both versions and keep the lower-causality version.

• Redraw the connected summaries as bars while keeping the same values and grouping.
• Keep aggregation constant when comparing the line and bar versions.
• If the current bar chart has only two groups, increase the number of groups before finalizing.
• Re-test the revised chart with the same causal-interpretation prompt.