How Can Graphics And/or Statistics Be Used To Misrepresent Data?

How to prevent misinformation in data visualization?

Charts, graphics, maps, and diagrams don't lie. People who pattern it do.

Scientific discipline Valentine from xkcd - Randall Munroe

Information visualization methods business organisation the pattern of graphical representation to summarize the information in analytical processes.

From this offset statement, information technology is important for companies to understand the risk that both disinformation and misinformation could represent.

And while disinformation is divers as a "imitation information deliberately and frequently covertly spread in lodge to influence public stance or obscure the truth", misinformation sets itself autonomously. And then yes, charts tin can mislead. And, nosotros will all concur that sometimes, the realisation of an aesthetic visualization without whatsoever hazard of misinformation is a true art. Still, we need to keep in heed that charts, graphics, maps, and diagrams are only the icons of analytical results.

How misinformation happens?

This is non a surprise. A graphic can trigger false insights through the utilize of collective pregnant that are often biased toward cocky-confirmation. Either by exposure, oversimplification or worldview, misinformation through data visualization is based on private and various components.

I've personally a preference for the Travis Syndrome by which a person tends to overestimate the significance of the present. Just, they're many others:

Mere-exposure effect: issue by which people develop a preference for things considering they become familiar with them.
Familiarity backfire result: consequence past which familiarity increases the chances of people accepting data as truthful.
Overkill backlash event: issue by which less information and data that is easier to process is more likely to be accepted as truthful.
Confirmation bias: tendency to search for, interpret, focus on and recall data in a manner that confirms one'due south preexisting beliefs.
Worldview backlash effect: outcome past which facts and counter-arguments to a person's strongly held behavior just serve to reinforce those behavior.

This sample of biases, which tin can exist used to produce misleading charts, raise the importance to take into consideration that many of these biases take impacts on business organisation and economic decisions too as human behaviour in general. The total list of cognitive biases is accessible here.

Designed by John Manoogian III — Categorized by Buster Benson

What techniques can be used to forbid biased view in data visualization?

Follow the 4 pillars of visualization

A visualization is considered successful when you're able to say yeah to the following questions :

Is the purpose well-defined? As the purpose dictate the deliverable, information technology needs to exist clearly defined.
Does the content support the purpose? Considering that what's included is as important every bit what'southward excluded, just the relevant content needs to be included. As a reminder, the content, which is a human relationship between the reality and the data must always be de-politicized and a-political.
Does the structure reveal the content? An advisable structure needs to be used to represent clearly the human relationship between the reality and the graphic.
Does the visualization facilitate consumption? The visualization, which is the link between the graphic and the data, must exist useful by highlighting what matters and removing the rest.

By using the 4 pillars, the creation of graphics allows to be the closest to the reality and the data. More a graphic will exist far from these conditions, and more it can leads to ambiguation.

Learn the Gestalt Principles

The term Gestalt means "course" or "shape". In psychology, a gestalt is something that has particular qualities when considered every bit a whole merely are not obvious when considered into separate parts.

Psychologists Max Wertheimer, Wolfgang Kholer, Kurt Koffka and later joined by Kurt Lewin, studied perception and concluded that perceivers should be considered as active rather than passive. In fact, they suggested that perceivers practise not just collect information as is. They actively process and restructure information in order to understand information technology. This is the perceptual procedure and sure factors impact on this perceptual process.

Co-ordinate to Gestalt psychology:

We tend to meet things with the encephalon.
The mind understands external stimuli equally whole rather than the sum of their parts
We tend to social club our feel on a regularly, orderly, symmetric and simple basis .

Be aware of all the tricks

Building contrary stories: introduce irrelevant data to encourage u.s.a. to assign a greater value to the larger part. (Cudmore, 2014)
Broken/False scales: changing the Y range tells a different story, (Jones 2006 — Peltier, 2011)

Close up about the Y-axis — Phonation

Scarlet-picking: hibernate information inside other data to softening impact.
False causality: post-obit the gestalt principle, when 2 or more lines appear together in a chart, and they look similar to each other, we have the tendency to assume they are related. (Cudmore, 2014)

"Evil" politicians: charting tricks often used by politicians to exagerate slopes, recession, increase. (Underwood, 2013)
Fancy looking: used to distract yous while yous should focus on the information.
Gerrymandering: manipulation of geographical boundaries to grouping data in order to alter the result. (cf. Gestalt: Police of Proximity)
Hiding differences through manipulation: either past using different scales for one flick (Brown, 2013), cumulative data, fourth dimension intervals (Huff, 1993), absolute values or using log scales to look less dramatic. (Underwood, 2013)
Omitting data: mostly used with pie/donut charts past excluding missing records before being presented every bit a whole.
"Optical illusion" using 3D: information technology is difficult to distinguish between what object is the background and which is the foreground, hard to translate graphics and where the colour of an object is perceived differently given the context of the surrounding colors.
Overfitting: create a model that is overly tailored to the information you lot take and not representative of the general tendency.
Perception of area: difficulty to evaluate the proportion of a result past not indicating labels. (Whitelaw-Jones, 2013)
Problem of the hateful: when there are outliers to one direction or other, the hateful is misleading. Median or mode are amend in this case.(Vembunarayanan, 2014)
Sampling bias: drawing conclusions from a sample that is not representative of the population. (Vembunarayanan, 2014)
Simpson's paradox (contrary to the "cherry-picking") : when a trend appears in different subsets of data but disappears or contrary when the groups are combined.
Size equally a simulated indicator: tends to mislead by the inadequate use of chimera sized.
Splitting data into many charts : various observations from each category are e'er better in one bar chart. (Kaushik, 2014)

And if it'due south non enough, you can still use the "Lie Factor"

Less noticed only no less of import, Edward Tufte came upward with a formula called the "Lie Factor" to calculate the level of accuracy of a graphic. The closer the Lie Factor is to one.0, the more authentic the graphic is.

Information is power and visualization is use of power. More than the audience will be prepared to deal with data-driven visuals and more they volition be able to identify and avoid errors and lies. Unfortunately, we will never vanish misleading charts, no matter how nosotros bespeak out and denounce them. And then, if you consider that a visualization is not accurate, articulate, informative, educational, a-political, de-policitized, inclusive, genuine, supportive and non-judgemental, don't let yourself be fooled. Fight the noise past asking questions. After all, data visualization should never be considered as some truth only only used to show obviousness, facts and support arguments.

Want to go more How-to Guides? There are plenty more on datakoch :)