Good research is reliant on good data sets. An obvious statement, perhaps, but in our line of work it’s an important sentiment to emphasise. Not only is it impossible to draw useful conclusions from bad data sets, but at worst, it can be harmful and even dangerous to use data which is skewed. It’s therefore critical that we always strive to create high quality data sets that are as closely representative of the groups we're studying as possible. Of course, there will always be some degree of error in using a sample (or smaller sub-group of people) to represent a full group or population, but with the right techniques we can reduce these errors so that they have minimal effect.
Survey research into autonomous and driverless vehicles in particular can sometimes suffer from skewed data because it can be perceived as a niche topic. Often, the demographic that tends to respond is white, younger or middle-aged men who take an interest in the subject. Meanwhile, older people - particularly older women - tend to be missing from data sets, which means that we get a skewed view of the public's perception of the topic.
At DG Cities, we have been actively working to address the issue of skewed data sets, looking for ways to facilitate a more diverse response to our autonomous vehicle research. This has been the case for our work on Project Endeavour, a multi-partner project aiming to accelerate and scale the deployment of autonomous vehicle (AV) services in our urban areas. We have used social media advertising on Facebook to attract different demographics of people to complete our Project Endeavour surveys, which have zeroed in on the topic of driverless cars. We know that this subject is attractive to some more than others, so we've worked hard to incorporate different survey techniques to drive up participation from traditionally under-represented groups. We trialled several ideas and arrived at a number of successful approaches, including:
Using web-ads that include "safety" messaging (e.g. Do you think driverless cars will be safe?) to appeal to people's interests in their safety and the safety of others. We found that these messages were particularly effective at encouraging older people to participate.
Framing messages in the daily lives of people we're trying to reach. For example, we used messages that referenced commuting, which was effective at improving uptake among mid-30 men and women.
Using images that illustrated different driverless car use cases. For example, we used an image of a female rider, and a different colour palette. This appeared to attract younger women to complete the survey.
These techniques and others mean that as the Project Endeavour survey has rolled out, we've been able to reach a diverse community by gender and age, with far more women completing the survey than we anticipated. We have also been able to ensure that we are broadly representative of the age range of the UK population, an achievement we are proud of considering the niche nature of the topic. As we continue this critical work in reaching more diverse audiences, our next challenge is to encourage participation of ethnic minority people in the study, as the majority of the first wave of participants identified as white. We will soon be running the survey in local regions in the UK where Project Endeavour trials are taking place, and in these regions we hope to see how regional focused messaging can drive up survey engagement.
We believe that everyone should have their say on the future of autonomous vehicles. If you would like your voice heard, why not fill in our Project Endeavour survey here.