Doing things with data

Research data are receiving lots and lots and lots of attention and rightly so. Not only will data outputs likely become the main form of research output in fairly short order - and already are doing so in many fields - but they present a host of new issues with which the research community has to grapple. What we might describe as the mechanics - that is, standards, annotation systems, curatorial practices, formats, interoperability, preservation and suchlike - are one side of the coin. Though they encompass some complex issues, these are at least fairly easily described and ordered. On the other side are the at-the-bench or at-the-desk practices that become embedded in the overall behaviour patterns of researchers. Research behaviours can sometimes be pretty consistent across disciplines, but they are not in relation to data, at least not yet. 

Research funders, however, are acting swiftly to help establish one part of this behaviour as the norm - data sharing. Numerous funders, national and international research councils as well as private charities and sponsors, are already requiring the open dissemination of datasets once a research project is complete. This complements the growing number of similar requirements that research articles are made available for sharing through Open Access once ready for publication. Interestingly, the practice in many communities is that the data-sharing happens long before project completion, and sometimes in real-time as the data are generated. 

We looked in some detail at data practices in eight different disciplinary areas and the findings were published in a report for the Research Information Network earlier this year. Amongst other findings were two things that relate to much of the current discussion about data. 

One is about reward. Our current system of career reward for researchers remains based on journal articles or books as the primary rewardable output from their efforts. It was always thus, though those of us ancient enough to have used the outputs of the 50s, 60s and 70s in the natural sciences can recollect that the system then seemed a good deal more focused on summative quality and much less focused on numbers of outputs: it was common to find an eminent scientist of the 60s writing up a lengthy article every 2, 3 or 5 years that described his or her work over a considerable period, and which told a scientific story amidst a great deal of context. Today's salami-slicing practices don't do the same job at all. 

But back to data, and how data dissemination can be rewarded. No-one has the answer yet, but the issue is being talked about a lot. If researchers are to produce and share datasets, they should see some sort of real reward for this. Those who already share told us that they do so for two reasons. First, in their discipline it is seen as A Good Thing and they want the 'warm, fuzzy feeling' of being a good guy in their peer community. Second, if they make their data available to all then in many disciplines they are likely to be included as an author on any research articles arising from the re-use of those data by others. This is how things are working out in practice at the moment. That doesn't mean that such practices will carry effectively across other disciplines, nor that they will persist optimally even in disciplines where they are common. A better system for assessing and rewarding data outputs themselves, and the dissemination behaviour of the data creators, is needed. The bodies that can influence that most positively and properly are the research funders and universities, by developing the means to explicitly reward data dissemination in a manner analogous to the way they have always rewarded the publication of articles.

The second issue is about data re-use. I'm going back the mechanics now, rather than researcher behaviour on this. Making datasets available to anyone who might wish to use them is a complex task requiring much skill. There is the issue of where to make the data available and this topic is exercising the institutional repository community quite heavily. University repositories can store many types of dataset, but making certain types available for re-use is an additional tweak that may be beyond what we should require from each research-based institution. In the UK, our system of national data centres assumes much of this burden and provides the required level of expertise in handling and preserving myriad data formats that relieves universities of trying to reproduce it at multiple local locations.  

There are other routes, however, and at a workshop during the Open Access and Research Conference last week, Paula Callan, the institutional repository guru based at Queensland University of Technology, had some interesting points to make on this. One was made in a response when she was asked how to handle a complex digital object, such as a multimedia object, and make sure it could be accessed and re-used by others. Paula's answer was that at QUT they are exploring an arrangement whereby a complex object would be recorded in the repository but the repository metadata record would link to a website where the creator would be able to supplement it with a better interactive experience of the object. I understand that development work on this is underway at QUT at the moment.

This brings me to a final point about research data and that is the issue of what skills will be needed within the research community to deal with data in the future. It is a topic being addressed very seriously in many parts of the world now, for it is realised that some very highly-developed skills indeed will be required. The JISC has recently published the report of a study it commissioned us to carry out on data management skills in the UK. The report makes a series of recommendations on how scientists and librarians can be skilled-up to meet the new demands that a data-intensive research world will produce.  New career paths are opening up and new opportunities for personal development for those interested in digital data management.