An independent study reference written by Dr Phuc V. Nguyen. It is not official subject material — for assessment requirements always follow your subject outline and vUWS.

Visual Communicationintermediate

The ethics of charts

Q: A national map shades each state by the raw number of workplace injuries, and New South Wales is darkest. What is the most defensible reading?

The map mostly reflects that New South Wales has the largest workforce, so a rate per hundred thousand workers is needed to compare safety. A count map is largely a population map, because larger populations generate more of almost everything. Comparing safety requires dividing by the exposed population, for instance injuries per hundred thousand workers. The counts being accurate does not save the chart, since the reader is being invited to read intensity from a display of size.

Q: Two analysts produce identically misleading charts. One did it deliberately, the other accepted a software default without checking. How should the profession treat them?

Both charts are unacceptable, because the duty runs to the reader and the effect is the same, though the appropriate response to each analyst differs. The duty runs to the reader, who is misled either way, so both charts have to be corrected and neither is excused by the tool. Intent still matters to how an organisation responds, since carelessness calls for review practice and training while deliberate distortion is a conduct matter. What intent does not do is make the published chart acceptable.

A chart makes a claim, and whoever publishes it is answerable for what a reasonable reader will take away, not only for whether each printed number is correct. Three obligations follow. Represent magnitudes faithfully, so the geometry matches the data. Show the denominator, and show uncertainty on estimates and forecasts, because counts without a base and estimates without a range invite conclusions the data cannot carry. Exact administrative counts need no interval. Disclose the choices, meaning the source, the filters, the exclusions and the time window. Intent is not the test. A chart built carelessly from a default and a chart built to mislead do the same damage to the reader.

Why it matters

Ask what an ordinary reader would say the chart means after five seconds, then check whether the data supports that sentence. If it does not, the chart is misleading, however carefully the footnotes are worded. This is close to the standard used for advertising claims. What matters is the impression created, not the technical defensibility of each separate component.

Before you read on — recall

A national map shades each state by the raw number of workplace injuries, and New South Wales is darkest. What is the most defensible reading?

Formulas

Counts and rates answer different questions

\text{rate}_{i} = \dfrac{c_{i}}{n_{i}}

A map or bar chart of raw counts

c_i

largely reproduces the size of each population

n_i

. If Sydney records more complaints than Hobart, that is mostly because more people live in Sydney. Dividing by the exposed population turns the chart into a comparison of behaviour rather than a comparison of size. Publishing counts where the reader will read them as intensity is one of the most common honest-looking distortions in business and public reporting.

Worked examples

Scenario

An operations report headlines a chart claiming complaints are at a record low. The underlying extract excludes complaints raised through the mobile application, which was the fastest-growing channel that year.

Solution

Every plotted number may be correct, and the chart is still false as read. The exclusion is invisible, so the reader takes the falling line as a claim about total complaints. Two things fix it. State the scope on the chart itself, so the title reads as complaints by phone and email only, and add the mobile channel as a second series so the total is recoverable. If the mobile data genuinely is not comparable, say so on the chart rather than in an appendix nobody opens.

Scenario

A model produces a forecast of next quarter revenue, and the chart shows a single line continuing smoothly past the last actual point.

Solution

A single line implies a precision the model does not have. Readers treat the projected segment as they treat the historical one, and the historical one is a record while the projection is an estimate. The honest form marks where actuals end and projection begins, and draws an interval around the projection showing the range the model considers plausible. If that interval is wide enough to include both growth and decline, that is exactly the fact the decision-maker needs, and the single line hides it.

Common mistakes

✗If every number on the chart is accurate, the chart is honest. Accuracy of the components and honesty of the whole are different tests. Scale choices, exclusions, missing denominators and omitted uncertainty can all produce a false reading from correct figures.
✗Misleading charts are the product of bad intent. Most come from defaults, deadlines and reused templates. Because the effect on the reader is identical, the professional obligation is to check the impression the chart creates, not to examine your own motives.
✗Adding a disclaimer resolves the problem. A footnote is read by a small minority, and it arrives after the reader has already formed an impression from the geometry. Corrections belong in the chart, in the title and in the axes, where the reader meets them first.
✗Ethics in analytics is only about privacy and personal data. Handling data lawfully is one part of it. Representing findings so that a reasonable reader draws a supported conclusion is a separate obligation, and it applies to charts built entirely from public, non-personal data.

Revision bullets

•The test is the impression a reasonable reader takes, not component-level accuracy
•Faithful geometry, visible denominator, uncertainty shown on estimates and forecasts, disclosed choices
•Counts reproduce population size, rates compare behaviour
•Mark where actuals end and projection begins, and show the interval
•Fixes belong in the title and axes, not in a footnote
•Careless and deliberate distortion harm the reader identically

Quick check

A national map shades each state by the raw number of workplace injuries, and New South Wales is darkest. What is the most defensible reading?

Two analysts produce identically misleading charts. One did it deliberately, the other accepted a software default without checking. How should the profession treat them?

Connected topics

Data Storytelling Visual Mistakes Truncated Axes Why Ethics Matters Ethics by Design

More in Visual Communication

Purpose of Visuals Visual Variables Choosing a Chart Dashboards

Sources

Cairo (2019)
Cairo, A. How Charts Lie: Getting Smarter about Visual Information. W. W. Norton, 2019.
Extended treatment of how correct numbers produce false readings, with the reader impression as the test.
American Statistical Association (2022)
American Statistical Association. Ethical Guidelines for Statistical Practice. Revised edition, 2022.
Professional obligations covering honest presentation of results and disclosure of methods and limitations.
Pandey et al. (2015)
Pandey, A. V., Rall, K., Satterthwaite, M. L., Nov, O., & Bertini, E. "How Deceptive are Deceptive Visualizations? An Empirical Analysis of Common Distortion Techniques." Proceedings of CHI 2015, 1469-1478.
Evidence that specific distortion techniques change what readers conclude, independently of the maker intent.

How to cite this page

Dr. Phil's Quant Lab. (2026). The ethics of charts. Derivatives Atlas. https://phucnguyenvan.com/concept/ba-chart-ethics