IASE 2020 Roundtable Conference
New Skills in the Changing World
of Statistics Education
6 – 10 July 2020, Nanjing, China


Discussant Summary Presentations

Keynote talk

Education, democratizing data, and software: Targeting the intersection

Chris Wild

Department of Statistics, University of Auckland

When I talk about democratising data here, I'm not talking about better public access to data or about critical thinking around data and data-based reports (both of which are critically important). I am talking about increasing the ability for people to do things with data themselves once they've got their hands on it – facilitating a broader spectrum of people being able to do more with data.

Largely, in analytics fields, we operate bottom-up education. Students work from lower-level conceptions and skills that can eventually be brought together to facilitate higher-level overviews. This emulates clambering over rocks in the foothills on the way to a mountain top. You might get occasional glimpses of the top but you will appreciate nothing of the view from the summit until you are a long way up. But you'll be a good climber. A top-down approach, on the other hand, seeks to emulate an eagle or drone's-eye-view – surveying expansive vistas from on high and zooming in, from time to time, to take a closer look at features that seem to merit closer inspection.

Bottom-up is a reasonable strategy for educating people who will be able to be able to build new things for themselves, and addressing the depths of an area, because that requires expertise with low-level tools. It is incredibly bad, however, for motivation ("Trust me. This will be good for you."). It is incredibly slow/inefficient for forming broad conceptions of what is possible in the data world, or for conferring a broad range of practical capabilities quickly. And it is never going to work for democratising data because non-specialists will never put in the prerequisite groundwork.

Even students on the way to becoming specialists need rapidly gained breadth of awareness as well as depth in sub-areas. Strategies that are optimised for depth are bad for breadth and vice-versa. But that shouldn't phase us because there's a strategy we all know about. It's called parallel processing!

In this talk we will blend discussion of principles and views of some data visualisation and analysis software we have been building that try to implement them.