Responsible visualization

What is a Visualisation?

A visualisation is defined as a visual presentation of data (http://dictionary.reference.com/browse/visualisation). That covers a lot of things, from maps (visual representations of where things are, typically with shapes, labels, icons and colours all meaning something) to data-based drawings (e.g. XKCD data-based comics), graphs (line graphs, bar charts, pie charts etc), infographics and dashboards.

Types of Visualisation

There are many types of visualisation, in as many forms as the human imagination can devise, but you’ll see some types a lot more than others. Some of these are:

• Line graphs
• Bar charts
• Column charts
• Maps
• Choropleths
• Sankey diagrams
• Network visualisations

The periodic table of visualisation methods is a good way to find names for new visualization types. Datavizualisation.ch curates a list of visualisation tools that you can use to explore more types of visualisation.

Case Study: Human Rights Funding Research

1. How do you show the findings?

2. How would you should who is funding where?

First, do no harm

We create visualisations because they usually have more persuasive power and are more accessible to more people than columns of numbers. With that power comes responsibility, especially when those visualisations are used to make decisions that affect vulnerable people.

There's an RDF blogpost series and event on responsible visualisation, which will cover these issues in depth. Until then, here are some things to consider when you create visualisations.

Do you have to redact data?

• Aggregate data at different levels.

Make sure the data actually represents the comparison in the true form

• Aggregate and quantify using statistics.

Be true to the data.

• Stay away from assumptions
• Infographics are propositions
• If your data is updated, either ensure that your visualisations are updated too, or make it clear that they won't be.

How do people interact with your visualisation?

What is the story you are trying to tell

• Clear, concise story-telling strategy.

Interaction

• How do people interact with your data.
• Is it overwhelming?

Understand how people see visualisations

Keep text and visuals together

• There can be a disconnect about the text and the visual if these are created by two different individuals. Annotations are important: if you're using text, make it part of the visual in a way that's not separable even while someone is cut-and-pasting or remixing your visualisation.

Think about the medium

People will look at your visualisation on different media: laptop/pc, phone, print etc.

• Colours don't always work well across media: for instance, if your visualisation is likely to end up in print, make sure its colours look good in print (both colour and greyscale).

Know your audience

• Many people are colorblind: consider this.
• Right-to-left cultures also have right-to-left visualisations. Design for this, or ignore it, but make sure you make a decision if you're outputting visualisations in e.g. Arabic.

Choose the right colors

• Random colours are meaningless without a legend. Use scales or greyscale, and if you use colour to mean something, always include a legend in the visualisation. NB The human eye can see more shades of grey than colour.

Annotation

• Labels
• Titles
• Tweets
• Description

Use pie charts wisely

Pie charts are good for quick prototypes, but the human eye isn't good at discriminating between similar angles. If you can use something other than a pie chart, do it. If you have to have a pie chart, then please:

1. Don't do 3D, pretty please!

2. Don't show more than 3 data points.

3. Don't show small angles - your users won't be able to tell which ones are larger.

Reading List

• Mushon Zer-Aviv, Disinformation Visualisation: How to Lie with Dataviz
• Sajjad Anwar, You can't always/just throw a map at a problem
• Pandey et al, The Persuasive Power of Data Visualisation
• Graham Odds, How to design better data visualisations