When Charts Exploit Cognitive Biases: A Conversation with Jens Eriksson

Data visualization is not neutral. Every chart encodes decisions that interact with the viewer's cognitive biases in predictable and often exploitable ways.

Anika van der Berg · March 26, 2026

A Neglected Intersection

I have spent most of my career studying how cognitive biases shape consumer decisions — on pricing pages, in checkout flows, and through marketing copy. But there is a domain where cognitive biases operate with particular force and surprisingly little scrutiny: data visualization.

This realization crystallized for me last autumn, when I came across the work of Jens Eriksson, a Stockholm-based information designer and former analytics lead at Spotify. Eriksson writes with unusual clarity about the ethics and mechanics of data visualization, and his essays on the Lie Factor and the perceptual limitations of pie charts struck me as doing something that most visualization writing does not: connecting design choices to their psychological consequences.

I reached out to Eriksson, and we exchanged a series of emails over several weeks. What follows is a synthesis of that conversation, organized around the cognitive biases most relevant to chart design.

Anchoring in Axis Design

The anchoring effect — the tendency for initial information to disproportionately influence subsequent judgments — is one of the most robust findings in behavioral science. Tversky and Kahneman demonstrated it in 1974, and it has been replicated hundreds of times across diverse domains.¹

In data visualization, anchoring operates through axis design. When a bar chart's y-axis begins at zero, the viewer anchors on the proportional relationship between bars. When the axis is truncated — starting at, say, $95 million instead of zero — the visual difference between bars is amplified, and the viewer's perception of the magnitude of change is distorted.

Eriksson has written about this through Tufte's concept of the "Lie Factor," which quantifies the ratio between the visual effect and the actual data effect. A Lie Factor of 1.0 means the chart is truthful. A truncated axis can easily produce a Lie Factor of 3.0 or higher — meaning the chart exaggerates the effect by a factor of three.

What interested me about Eriksson's framing is that he treats this not as a technical problem but as a moral one. "A chart with a Lie Factor of 3.0 is the visual equivalent of a misquotation," he writes. From a behavioral science perspective, I would add that it is also an exploitation of the anchoring bias: the truncated axis sets an anchor that distorts all subsequent magnitude judgments.

The Framing Effect in Chart Type Selection

Prospect theory tells us that people evaluate outcomes relative to a reference point, and that losses loom larger than equivalent gains. This framing effect extends directly to visualization choices.

Consider two ways to present the same data: a company's market share declined from 34% to 31% over a year. Presented as a line chart trending downward, this activates loss framing — the viewer sees decline, threat, erosion. Presented as a bar chart showing "31% market share," with no temporal dimension, the same data activates a more neutral evaluation. The information is identical. The psychological impact is not.

Eriksson made a related observation in our correspondence: "The choice of chart type is the first editorial decision. A pie chart showing 31% emphasizes the part-of-whole relationship. A line chart showing the same number emphasizes change over time. Neither is wrong, but they are not the same argument." This aligns precisely with what the framing literature would predict. The chart type is the frame, and the frame shapes the judgment.

Area Perception and the Ratio Bias

There is a cognitive phenomenon known as the ratio bias — the tendency for people to be influenced by the absolute size of a numerator and denominator rather than their ratio. In one demonstration, participants judged "9 out of 100" as a higher probability than "1 out of 10," even though the former (9%) is lower than the latter (10%).²

A related perceptual distortion occurs with area-based encodings in charts. Bubble charts and pictogram charts encode quantity as area, but humans perceive area non-linearly. As Eriksson notes, a circle with twice the radius has four times the area. A bubble that appears "twice as large" to the viewer actually represents four times the value. This systematic misperception means that area-based charts routinely overstate the differences between large and small values.

Eriksson's recommendation — to prefer position-based encodings (bars, dots) over area-based encodings (bubbles, pictograms) — is supported by William Cleveland's research on graphical perception. But the behavioral science adds a layer: the misperception is not random. It is systematically biased toward overestimating large differences, which means area charts will reliably exaggerate inequality, growth, and dominance.

Color and Attentional Bias

Attentional bias — the tendency for perception to be affected by recurring thoughts and salient stimuli — has obvious implications for color use in charts. Eriksson has written that color in data visualization should serve exactly four functions: distinguishing categories, representing quantities, highlighting specific data points, and encoding a data variable. "Decoration is not one of them," he writes.

From a cognitive perspective, the problem with decorative color is that it creates false salience. A bright red segment in an otherwise muted chart will attract disproportionate attention regardless of its informational importance. This is the Von Restorff effect in action — the well-established finding that items that stand out from their context are more likely to be remembered and attended to.³

In marketing dashboards, this is routinely exploited. The metric the presenter wants you to focus on is highlighted in a saturated color. The metrics that tell a less favorable story are rendered in muted greys. The data is all present, technically, but the attentional architecture of the chart has been engineered to direct the viewer's cognitive resources toward a predetermined conclusion.

Implications for Practice

A Note on Collaboration

My exchange with Eriksson reinforced a conviction I have held for some time: the fields of data visualization and behavioral science have much to learn from each other. Visualization researchers understand perceptual accuracy with a precision that most behavioral scientists lack. Behavioral scientists understand the cognitive mechanisms — the biases, heuristics, and judgment errors — that determine how viewers actually process visual information. The intersection is where the most important work remains to be done.

Eriksson's writing is available at 004-jens-eriksson.vercel.app, and I recommend it to anyone who creates or consumes data visualizations — which, at this point, is essentially everyone.