Of all the phrases of the pandemic era, ‘flattening the curve’ might be the most memorable. The concept conveyed to the public how lockdowns were not attempts to make the virus disappear, but to stretch the spread of infections out over time in a way that would avoid a catastrophic surge of demand on the NHS.
It was, says Dr Bodo Winter, Associate Professor in Cognitive Linguistics at the University of Birmingham, an intriguing example of how data can be turned into a visual format that effectively communicates to non-expert audiences.
Unfortunately, Winter argues, there are plenty of examples of data experts failing to speak to their audience effectively, with sometimes devastating results. One historical example was the engineers of the Challenger mission, the doomed 1986 space shuttle disaster. Engineers knew that cold temperatures could be dangerous for the spacecraft and the forecast for the next day was projected to be very low temperatures. They warned NASA executives against the launch but they were ultimately ignored.
Winter wonders whether a well-designed graph, showing the relationship between temperature and damage on previous test runs, would have forced a different decision. “Data visualisation experts have retrospectively created graphs that drive home this message much better than what was shown at the evening of the launch. That's an example where not only have millions of dollars been wasted, but it was a national, emotionally harrowing disaster with everybody on board dying, because the people with the data and knowledge could not convince the ones making the decision.”
The problem is subtle, argues Winter. Data experts can be saddled with the ‘curse of knowledge’. They do not think about the knowledge deficits of their audience and how that might limit communication. “Decision makers generally have lower data literacy, and data analysts have very high data literacy, making it easy for this communication to go wrong because the people with the data often find it very hard to put themselves into the shoes of the people who are making the decisions,” says Winter.
A real world example of the ‘knowledge curse’ is the Red River flood from 1997 in North Dakota and Minnesota. Weather forecasts projected flood levels of between 47.5 feet and 49 feet, a number communicated to the public and local governments, but they “did not know what to do with that number. They built dykes at 52 feet, which seemed like a reasonable thing to do, then the flood came in at 54 feet and everything got swamped”.
For Winter, audiences were not fully informed about the uncertainty built into the model, leading them to think the flood would come in somewhere between the two measurements. “Using language and saying things like ‘approximately’ or ‘about’, would have helped to make people appreciate the uncertainty”.
Making numbers meaningful
Winter believes that, with the right training and support, data scientists can become more effective communicators. His interdisciplinary linguistics project, Making Numbers Meaningful, is exploring the role of data visualisation, language and gesture in shaping the communication of numerical information to decision-makers.
The project is seeking to expose the often hidden or under-appreciated dimensions of numerical communication, such as gesture. For example, Winter and his team reviewed the gestural dimension of communication related to ‘tiny numbers’. Analysing over 500 television news clips, they found that 80% of the time that somebody uses the phrase ‘tiny numbers’, they will gesture the size of the quantity by pinching their fingers together. “What is beautiful about that is it shows you that these gestures are ubiquitous when people talk about numerical information, but there are surprisingly few studies about it”.
Gestures can influence how numbers are perceived. “We often take language to a higher standard of accountability so if somebody says something outrageous we immediately complain, but if somebody performs a gesture that is suggesting in a more subconscious or more indirect way, that can actually sway decision making or opinions, but it often goes unnoticed”. Dr Winter believes physical gestures could influence how people interpret the significance of a number, like the number of people arrested in a protest.
The team is also analysing linguistic trends in large text databases, as part of the field of corpus linguistics. In particular, they are investigating the possibility of quantifying the extent to which a text uses precise or vague language related to numbers. Expressions like ‘several’, ‘many’, ‘few’, ‘high number’ or ‘low number’, are more open to judgement as opposed to mentioning exact numerals. “We want to look at the extent to which people talk about numerical facts vaguely or precisely and then ultimately, you could give me a text, and I could pin a number, a precision metric from how much vague language in relation to exact language is there”.
Upskilling the data science community
Winter says that data scientists are aware of the communications challenge they face, and they are not often trained in this area, while companies rarely invest in it. One survey showed that more than 25% of all data scientists have recently experienced a situation where they failed to explain key concepts of their analysis to decision makers.
Data scientists think of themselves as numbers people, and “they like the numbers, the data and coding, and they don't necessarily recognise the importance of communication and what can go wrong,” says Winter, who is considering developing a massive online course on communication for data scientists (MOOC). “The idea is to train them and make them aware that it's important to invest in communication. This applies not only to individuals, but also to institutions.”
Discover more stories about our work and insights from our leading researchers.