Research needs to focus on basic computing skills
By Bruce Beckles
Next article: European interoperability goes global
Previous article: What we need to know about climate change
Suppose it was decided to improve people’s scientific knowledge by improving access to sophisticated scientific equipment: electron microscopes, particle accelerators, that sort of thing. With much fanfare, an exciting new programme of investment begins. Unfortunately, the complicated equipment, combined with the new and innovative ways of accessing it, turns out to be less intuitive and easy to use than one might have hoped. In response, and somewhat belatedly, resources are invested into providing training. Is any of this sounding familiar?
Over the last few years, myself and some colleagues at Reading and Oxford have been doing something unusual (at least amongst IT service and infrastructure providers in UK academia). We ventured out and interviewed ordinary researchers and the staff who provide them with ICT. We deliberately targeted not the early adopters of e-Research, but rather those people who were just doing ordinary research – that is research without the e.
When we presented our results at the 2009 UK e-Science All Hands Meeting, we showed that the ICT-related problems that researchers faced were much more basic than those that the existing e-infrastructure seeks to solve. Our results show that researchers face two big problems. The first is data backup, archiving and long-term curation. Crucially, this applies to both small amounts of data and large data sets. The second problem, upon which I will concentrate in this article, is a lack of knowledge and skills in scientific programming and scientific computing more generally. (None of the researchers cited a lack of access to powerful computing resources as a significant issue.)
Increasingly, it appears that to perform any kind of quantitative research, it is necessary to have knowledge not only of the relevant research methods, but also of the IT skills needed to put that knowledge into practice. This often comes down to programming of some sort, or of using the built-in commands of a sophisticated software package, like MATLAB or Stata, in a programmatic way.
For some reason, training provision for IT skills in our universities seems to be in relatively short supply. In many of our universities, the necessary computing skills are not taught at undergraduate or masters level, meaning that many postgraduates are not equipped for their research. Once a postgraduate begins his or her research, they find that IT skills are considered to be basic skills, which they are then expected to acquire quickly. Most try to teach themselves, with mixed results. Unfortunately, this means that low-quality scientific code is not unusual.
A good point, you may say, I've often thought that my colleagues' code was awful, but what does that have to do with e-Research – or e-anything for that matter? It's quite simple: researchers can't effectively engage with their existing ICT provision, never mind the existing e-infrastructure, without the basic computing skills to do so, and that often means programming skills. Until we address this need, our investment in e-infrastructure is necessarily going to produce very limited returns. And returning to my starting point: instead of investing in expensive national e-infrastructure, we should fund teaching and training in programming and scientific computing for our academics.
Let's not dwell on the past, but instead concentrate on how we can improve the future. The challenge now is to ensure that future generations of researchers have the IT skills and knowledge that have become necessary for basic research.