That’s one small step for research, one giant leap for data 

CompSci Cat on the growing importance of data stewardship

Recently I had the very great pleasure of attending Future Labs Live Basel, to present a talk on FAIR data and Metadata. It was a wonderful two days full of interesting talks, lively panel discussions, and lots and lots of networking!  This is the third time I’ve attended this event, and every year I come away buzzing with new ideas, and there are always some key repeating themes that stand out. This year, the resounding thing that stood out to me, is how much we need data stewards.   

We have spent years as a community complaining about data not being shared or made available in a useable way, and have dedicated a significant amount of time in developing frameworks and principles such as FAIR (findable, accessible, interoperable, re-useable – see my first column); developed endless controlled vocabularies and standards for metadata, and this topic has spawned entire conference series. But what we haven’t done, is develop the people.   

As a company, if you wanted a beautiful shiny piece of software written, and you didn’t have the necessary skills to do it yourself, you would hire a developer, or as an academic you would make sure that you costed one into your funding application, or that the proposed team included one. This is because we realise and appreciate the value of a software engineer, and within the research community we have established the role of research software engineer, because we desperately need people with that skillset, and we are willing to dedicate time and money to that endeavour. We have also spent time establishing the role of data scientist, to analyse and understand our data.   

So we are putting this effort into our code, and into interpreting the data, but not in how to manage, structure and share the data, which just feels ridiculous! I would say that its pretty widely accepted that we should share our data, we should make it adhere to FAIR principles, we should produce beautiful machine-readable structured metadata that uses common controlled vocabularies so that we can embed the relevant context in it. The funding councils (quite rightly) have imposed the requirement for data management plans alongside research proposals; there are mandates that research funded by public money must make the research data available, and more journals are clamping down on ensuring that data is published alongside their papers.   

And that’s great, it’s great that we are aware of the problem and it’s great that work is being done to encourage sharing our data, but it’s not as simple as just imposing that it must be done. If the researchers doing the work don’t have the required knowledge, skills, or time to achieve this, then you can impose all you like, it still won’t get done properly. We need to start treating the data management processes with as much respect as we do the research, or the software development, and we need to start appreciating the level of expertise and time required to actually do this properly, and make sure that we actually account for this. In a nutshell, we need data stewards.   

For anyone unfamiliar with the term, data stewardship is essentially the role of managing the data of an organisation or project, and covers a whole range of tasks from ensuring the quality of the data, making sure its accessible and useable, and managing its metadata. It requires a very particular set of skills, and is an entire job within itself, and we need dedicated staff in these positions. Much like you aren’t likely to suddenly develop stellar coding skills overnight without substantial study, having a comprehensive understanding of good research data management, knowing about FAIR, metadata, standards for data and metadata including ontologies, having knowledge of repositories, data retention policies, data licenses, and so much more also doesn’t just happen. You can’t expect every researcher to have the years of experience that a proper data steward will have, and it’s just asking for trouble assuming that they will be able to muddle through.   

For a task that is afforded so much importance when its not done well, we aren’t currently putting any provisions in to do it well! And this doesn’t mean every project needs their own personal data steward, some may well do, but if we instantiate data stewards at our institutions then they can be brought in to advise and assist across different projects. As its equally important to ensure that they can work WITH the researchers, as that combination of data expertise and domain knowledge is absolutely key to making our data FAIR. The researchers can provide the valuable context to the data, and the data stewards can make sure this is captured in the metadata.   

So, I will get off my soapbox now, but I want to leave you with a final thought. If you want a highly complex task to be performed well, then it requires time, effort, dedication, and an in-depth knowledge of the task at hand; and when approaching any task as a team, we should be playing to our strengths. In a team of runners, you wouldn’t expect the same person to do the half-marathon and 1000 m sprint, because they are completely different distances and involve different training. We need to STOP asking our researchers, to do both their own job and that of a data steward and START building up a set of people who can do this work, and through this I truly believe that this will enable us to take a huge step forward in producing FAIR data.