It has been an extremely rewarding two days at the Symposium on Big Data and Human Development that Eduardo Lopez and I organised. We had a full room of people from academia, government, and the development sector – all speaking about how we might better use big data in the contexts of development.

There are many threads that we’ll try to tie up over the next few weeks (an edited book, some workshop reports, perhaps another conference next year, etc.). But in the meantime, it might be useful if I reproduce the notes that I used to sum up the event here. Those of you who attended, please do comment if you see that I omitted anything. Those of you who didn’t, please feel free to use this as a prompt to get involved.

***

This has been a much-needed conversation at a moment in which we’re awash with hype about ‘big data’.

We’ve learnt a lot about some of the potentials of big data: We’ve got new sorts of early warning signals. And – as we move from data to information to knowledge – we seem to be getting better at figuring out what to look for when it comes to disease tracking, or predicting things like student failure rates or corruption.

The fact that so much data comes from mobile phones has also created a specific opportunity to look at human mobility. And the relative democratisation of connectivity has important implications for deliberation and public participation at scales that have never before been possible.

But, with all that in mind, I want to pick up with areas that I think we still need to find ways to resolve as we all move forwards at this intersection of topics:

First, one theme that keeps coming up is that of data presences and absences really mattering. We have great data about some places, processes, people. But there are still big gaps – and, going forwards, we’ll really to address this head-on. If we’re using data to deploy scarce resources or deliver essential services, but there are blank spots on our map – then what strategies should we be employing to deal not just with our known unknowns, but also our unknown unknowns? Some of this might entail really getting good about asking questions about outliers in our models: Where are they, who are they, when are they?

Second, another important theme is not just data presences and absences – but even within the presences, there is the question of open versus closed data. So, for instance – many of us – me included – tend to use Twitter data to ask and answer a range of questions. And we do this because it is easily available and free and relatively straightforward to use.

But we should be careful that we don’t get into the sort of situation in which the tail wags the dog rather than the dog wags the tail – as my colleague Ralph Schroeder puts it. What sorts of questions are we prevented from asking because of a lack of open, available data sources? What sorts of questions or topics are we perhaps focusing too much energy on? And what sorts of questions do our data lend or not lend themselves to?

Third, and relatedly, we’re faced with some tension between issues of privacy, ownership, and control. How do we balance the desire to have more open data with best practices that prevent data leakage and still afford citizens with some control over their own data shadows?

There was an interesting discussion in the session that I organised with Richard Heeks at the DSA conference earlier this week about what we might learn from the literature on resource management – if we treat data as a resource.

And more broadly, are we happy with the current political-economy of development data? What current rights of access, control, and use should be rethought and challenged?

Fourth, how do we ourselves operate with maximum transparency – especially when we’re not just dealing with descriptive analytics, but predictive analytics, and even prescriptive analytics? If our research, and the data we use, impacts on real people in real ways – are we happy with the current scientific models of dissemination that we use – or do we need any sort of alternate strategies that better engage with the communities that are the users – or subjects – of development?

Fifth, what can, or should, we learn across contexts? Or specifically, what should we rethink and relearn in different places or contexts? What sorts of things aren’t transferrable? This is maybe where the repeated call throughout this conference for all of us to be thinking and collaborating in a multidisciplinary way comes in useful.