Nobody likes uncertainty. We, as human beings, like control. We like to be sure we know what’s happening next. Often, my clients ask me to provide them with “bulletproof” results. They want their data analysis to be 100% correct and absolutely certain.
And that’s just not possible.
Data analysis and statistics are not tools that can provide you with certainty. All they can do is help you understand how certain you should be. And that’s why it’s important to know how to communicate uncertainty in data.
Why Can’t We Be Sure?
Think of it this way. When you visit a doctor because you’re unwell, they prescribe a course of treatment based on the information they have. There is no guarantee the medication will make you better. If you work in a school, you could test a new teaching method aimed at raising students’ test scores. But even if your trial was 100% successful, it wouldn’t guarantee you could raise every student’s score in the future.
We use data analysis in these situations. We can say that, based on existing data, five times out of six the doctor’s prescription will make you better. Or that the new teaching method improves test scores for the typical student by six to eight points.
That information is good to have when deciding whether to take your medicine or implement a new instructional program. But it’s not a guarantee. And that’s why we need to communicate uncertainty in our data.
If you’re telling stories with data (or writing narratives with data or communicating the results of data collection and analysis) anywhere outside of a peer-reviewed statistical journal, you’re going to have to communicate uncertainty in data.
Uncertainty Isn’t Sexy
Data analysis generally aims to help us understand what’s happening around us. Data storytelling, however, is done to help the audience understand the results. Sometimes, our desire to share our results with a larger audience — to have them be interested and take action — can make telling our story trickier.
The careful communication of probability is not always in alignment with effectively catching the roving attention of the human mind. In short, uncertainty isn’t sexy.
Data storytelling can often gloss over nuance. And it’s easy to understand why.
- Reporters and editors don’t always have the training to accurately interpret the studies they are covering
- Headlines that read “weak, unreplicated study finds tenuous link between certain vegetables and cancer risk” just don’t appeal like ones that say “foods that fight cancer!”
Uncertainty exists within any result. If you’re going to tell stories with complex or nuanced statistics, you’ll need to work on how to communicate uncertainty in data.
The Problem with Probability
There isn’t a hard and fast set of rules on how to communicate uncertainty in data. (At least, not one I’m aware of.) But that doesn’t mean you don’t have to try. Here are a few great examples from some data storytelling experts:
- Nate Silver of FiveThirtyEight wrote a great piece about uncertainty, probability, and how to communicate both in a series he started just after the 2016 US presidential election. As he says in this final piece of an eleven-part series, “The media’s demand for certainty — and its lack of statistical rigor — is a bad match for our complex world.”
- Alberto Cairo also recently published a very thorough and thoughtful piece on communicating uncertainty in data and hurricanes. He points out that a lack of graphical literacy can make it difficult for many audiences to differentiate between probability and fact.
- Want to be sure you really understand what uncertainty is? The University of British Columbia explains it in a simple, effective way using their Grammar Squirrel mascot.
- NPR also did a series on understanding uncertainty and risk. They do an excellent job of modelling how to communicate uncertainty in data.
Communicate Uncertainty in Data
There are three main types of uncertainty you may find yourself trying to communicate to the public. Your data is likely helping your audience to:
- Look for a signal (like when to evacuate in a hurricane)
- Choose between fixed options (like which medicine to take)
- Develop a sense of what is possible (like how to create a program to reduce poverty)
For each of these scenarios, the audience must understand the degree of certainty associated with the data. At very least, this should include a summary of the variability, sources of uncertainty, and the level of certainty.
The examples I listed above should give you a starting point for how to communicate uncertainty in data — but you’re not on your own. If you need help telling your data story in an honest, engaging way, let us help. Get in touch with the team at Datassist today.