Tim's Blog

[Summary: notes for a talk on open data and International Organisations]

In just over a weeks time I’l be heading for Geneva to take part in Diplo Foundation’s ‘E-Participation Day: towards a more open UN?’ event. In the past I’ve worked with Diplo on remote participation, using the web to support live online participation in face-to-face meetings such as the Internet Governance Forum. This time I’ll be talking open data – exploring the ways in which changing regimes around data stand to impact International Organisations. This blog post was written for the Diplo blog as an introduction to some of the themes I might explore. 

The event will, of course, have remote participation – so you can register to join in-person or online for free here.

E-participation and remote hubs have the potential to open up dialogue and decision making. But after the conferences have been closed, and the declarations made, it is data that increasingly shapes the outcome of international processes. Whether it’s the numbers counted up to check on progress towards the millennium development goals, GDP percentage pledges on aid spending, or climate change targets, the outcomes of international co-operation frequently depend on the development and maintenance of datasets.

The adage that ‘you can’t manage what you can’t measure’ has relevance both for International Organisations and for citizens. The better the flows of data International Organisations can secure access to, the greater their theoretical capacity for co-ordination of complex systems. And the greater the flows of information from the internal workings of International Organisations that citizens, states and pressures groups can access, the greater their theoretical capacity to both scrutinise decisions and to get involved in decision making and implementation. I say theoretical capacity, because the picture is rarely that straightforward in practice. Yet, that complexity aside for a moment, over the last few years the idea has been gaining ground that, in some states has led to not only a greater flow of data, but has driven a veritable flood – with hundreds and thousands of government datasets placed online for anyone to access and re-use. That idea is open data.

Open Data is a simple concept. Organisations holding datasets should place them online, in machine-readable formats, and under licenses that let anyone re-use them. Advocates explain that this brings a myriad of benefits. For example, rather than finance data being locked up in internal finance systems, only available to auditors, open data on budgets and spending can be published on the web for anyone to download and explore in their spreadsheet software, or to let third parties generate visualisations that show citizens where their money is being spent, and to help independent analysts look across datasets for possible inefficiency, fraud or corruption. Or instead of the location of schools or health centres being kept on internal systems, the data can be published to allow innovators to present it to citizens in new and more accessible ways. And in crisis situations, instead of co-ordinators spending days collecting data from agencies in the field and re-keying the data into central databases, if all the organisations involved were to publish open data in common formats, there is the possibility of it being aggregated together, building up a clearer picture of what is going on. One of the highest profile existing open data initiatives in the development field is the International Aid Transparency Initiative (IATI) which now has standardised open data from 100s or donors, providing the foundation for a timely view of who is doing what in aid.

Open data ideas have been spreading rapidly across the world, with many states establishing national Open Government Data (OGD) initiatives, and International Organisations from The World Bank, to UN DESA, the OECD and the Open Government Partnership all developing conversations and projects around open data. When the G8 meet next week in Northern-Ireland they are expected to launch an ‘Open Data Charter’ setting out principles for high quality open data, and committing states to publish certain datasets. Right now it remains to be seen whether open data will feature anywhere else in the  in the G8 action plans, although there is clearly space for open data ideas and practices to be deployed in securing greater tax transparency, or supporting the ongoing monitoring of other commitments. In the case of the post-2105 process, a number of organisations have been advocating for an access to information focus, seeking to ensure citizens have access to open data that they can use to monitor government actions and hold governments to account on delivering on commitments.

However – as Robinson and Yu have highlighted – there can be an ambiguity of open government data: more open data does not necessarily mean more open organisations. The call for ‘raw data now’ has led to much open data emerging simply as an outbound communication, without routes for engagement or feedback, and no change in existing organisational practices. Rather than being treated as a reform that can enable greater organisational collaboration and co-ordination, many open datasets have just been ‘dumped’ on the web. In the same way that remote participation is often a bolt-on to meetings, without the deeper changes in process needed to make for equal participation for remote delegates, at best much open data only offers actors outside of institutions a partial window onto their operations, and at worst, the data itself remains opaque: stripped of context and meaning. Getting open data right for both transparency, and for transforming international collaboration needs more than just technology. 

As I explored with Jovan Kurbalija of Diplo in a recent webinar, there are big challenges ahead if open data is to work as an asset for development: from balancing tensions between standardisation and local flexibility, developing true multi-stakeholder governance of important data flows, and getting the incentives for collaboration right. However, now is the time to be engaging with these challenges – within a window of energy and optimism, and before network effects lock in paradoxically ‘closed’ systems of open data. I hope the dialogue at the Geneva E-Participation day will offer a small chance to broaden open data understanding and conversations in a way that can contribute to such engagement.

There’s lots of interest building right now around how open data might be a powerful tool for transparency and accountability in the extractive industries sector. Decisions over where extraction should take place have a massive impact on communities and the environment, yet often decision making is opaque, with wealthy private interests driving exploitation of resources in ways that run counter the public interest. Whilst revenues from oil, gas and mineral resources have the potential to be a powerful tool for development, with a proportion channeled into public funds, massive quantities of revenue frequently ‘go missing’, lost in corruption, and fuelling elements of a resource curse.

For the last ten years the Extractive Industries Transparency Initiative has been working to get companies to commit to ‘publish what they pay‘ to government, and for government to disclose receipts of finance, working to identifying missing money through a document-based audit process. Campaigning coalitions, watchdogs and global initiatives have focussed on increasing the transparency of the sector. Now, with a recognition that we need to link together information on different resources flows for development at all levels, potentially through the use of structured open data, and with an anticipated “data tsunami” of new information on extractives financials anticipated from the Dodd-Frank act in the US, and similar regulation in Europe, groups working on extractives transparency have been looking at what open data might mean for future work in this area.

Right now, DFID are taking that exploration forward through a series of hack days with Rewired State under the ‘follow the data’ banner, with the first in London last weekend, and one coming up next week in Lagos, Nigeria. The idea of the events is to develop rapid prototypes of tools that might support extractives transparency, putting developers and datasets together over 24 hours to see what emerges. I was one of the judging panel at this weekends event, where the three developer teams that formed looked respectively at: making datasets on energy production and prices more accessible for re-use through an API; visualising the relationship between extractives revenues and various development indicators; and designing an interface for ‘nuggets’ of insight discovered through hack-days to be published and shared with useful (but minimal) meta-data.

In their way, these three projects highlight a range of the challenges ahead for the extractives sector in building capacity to track resource flows through open data:

• Making data accessible – The APIfy project sought to take a number of available datasets and aggregate them together in a database, before exposing a number of API endpoints that made machine-readable standardised data available on countries, companies and commodities. By translating the data access challenge from one or routing around in disparate datasets, to one of calling a standard API for key kinds of ‘objects’, the project demonstrated the need developers often have for clear platforms to build upon. However, as I’ve discovered in developing tools for the International Aid Transparency Initiative, building platforms to aggregate together data often turns out to be a non-trivial project: technically (it doesn’t take long to get to millions of data items when you are dealing with financial transactions), economically (as databases serving millions of records to even a small number of users need to be maintained and funded), socially (developers want to be able to trust the APIs they build against to be stable, and outreach and documentation are needed to support developers to engage with an API), and in terms of information architecture (as design choices over a dataset or API can have a powerful affect on downstream re-users).
• Connecting datasets – none of the applications from the London hack-day were actually able to follow resource flows through the available data. Although visions of a coherent datasphere, in which the challenge is just making the connection between a transaction in one dataset, and a transaction in another, to see where money is flowing, are appealing – traceability in practice turns out to be a lot harder. To use the IATI example again, across the 100,000+ aid activities published so far less than 1% include traceability efforts to show how one transaction relates to another, and even here the relationships exist in the data because of conscious efforts by publishers to link transaction and activity identifiers. In following the money there will be many cases where people have an incentive not to make these linkages explicit. One of the issues raised by developers over the hack-day was the scattered nature of data, and the gaps across it. Yet – when it comes to financial transaction tracking, we’re likely to often be dealing with partial data, full of gaps, and it won’t be easy to tell at first glance when a mis-match between incoming and outgoing finances is a case of missing data or corruption. Right now, a lot of developers attack open data problems with tools optimised for complete and accurate data, yet we need to be developing tools, methods and visualisation approaches that deal with partial and uncertain data. This is developed in the next point.
• Correlation, causation and investigation – The Compare the Map project developed on the hack day uses “scraped data from GapMinder and EITI to create graphical tools” that allow a user to eye-ball possible correlations between extractives data and development statistics. But of course, correlation is not causation – and the kinds of analysis that dig deeper into possible relationships are difficult to work through on a hack day. Indeed, many of the relationships mash-ups of this form can show have been written about in papers that control for many more variables, dealing carefully with statistically challenging issues of missing data and imperfectly matched datasets. Rather than simple comparison visualisations that show two datasets side by side, it may be more interesting to look for all the possible statistically significant correlations in a datasets with common reference points, and then to look at how human users could be supported in exploring, and giving feedback on, which of those might be meaningful, and which may or may not already be researched. Where research does show a correlation to exist, then using open data to present a visual narrative to users about this can have a place, though here the theory of change is very different – not about identifying connections – but about communicating them in interactive and engaging ways to those who may be able to act upon them.
• Sharing and collaborating – The third project at the London hack-day was ‘Fact Cache‘ – a simple concept for sharing nuggets of information discovered in hack-day explorations. Often as developers work through datasets they may come across discoveries of interest, yet these are often left aside in the rush to create a prototype app or platform. Fact Cache focussed on making these shareable. However, when it was presented discussions also explored how it could make these nuggets of information into social objects, open to discussion and sharing. This idea of making open data findings more usable as social objects was also an aspect of the UN Global Pulse hunchworks project. That project is currently on hold (it would be interesting to know why…), but the idea of supporting collaboration around open data through online tools, rather than seeing apps that present data, or initial analysis as the end point, is certainly one to explore more in building capacity for open data to be used in holding actors to account.
• Developing theories of change – as the judges met to talk about the projects, one of the key themes we looked at was whether each project had a clear theory of change. In some sense taken together they represent the complex chain of steps involved in an open data theory of change, from making data more accessible to developers, creating tools and platforms that let end users explore data, andthen allowing findings from data to be communicated and to shape discourses and action. Few datasets or tools are likely to be change-making on their own – but rather can play a key role in shifting the balance of power in existing networks or organisations, activists, companies and governments. Understanding the different theories of change for open data is one of the key themes in the ongoing Open Data in Developing Countries research, where we take existing governance arrangements as a starting point in understanding how open data will bring about impacts.

In a complex world, access to data, and the capacity to use it effectively, are likely to be essential parts of building more accountable governance across a wide range of areas, including in the extractives industry. Although there are many challenges ahead if we are to secure the maximum benefits from open data for transparent and accountable governance, it’s exciting and encouraging to see so many passionate people putting their minds early to tackling them, and building a community ready to innovate and bring about change.

Note: The usage of ‘follow the data’ in this DFID project is distinct from the usage in the work I’m currently doing to explore ‘follow the data’ research methods. In the former, the focus is really on following financial and resource flows through connecting up datasets; in the latter the focus is on tracing the way in which data artefacts have been generated, deployed, transferred and used in order to understand patterns of open data use and impact.

[Summary: join an open discussion on the potential impacts of open data on poverty reduction]

Over the next two weeks, along with Tariq Kochar, Nitya V. Raman and Nathan Eagle, I’m taking part in an online panel hosted by the World Bank’s Striking Poverty platform to discuss the potential impacts of open data on poverty alleviation.

So far we’ve been asked to provide some starting statements on how we see open data and poverty might relate, and now there’s an open discussion where visitors to the site are invited to share their questions and reflections on the topic.

Here’s what I have down as my opening remarks:

Development is complex. No individual or group can process all the information needed to make sense of aid flows, trade patterns, government budgets, community resources and environmental factors (amongst other things) that affect development in a locality. That’s where data comes in: open datasets can be connected, combined and analysed to support debate, decision making and governance.

Projects like the International Aid Transparency Initiative (IATI) have sought to create the technical standards and political commitments for effective data sharing. IATI is putting together one corner of the poverty reduction jigsaw, with detailed and timely forward-looking information on aid. IATI open data can be used by governments to forecast spending, and by citizens to hold donors to account. This is the promise of open data: publish once, use many times and for many purposes.

But data does not use itself. Nor does it transcend political and practical realities. As the papers in a recent Journal of Community Informatics special issue highlight show, open data brings both promise and perils. Mobilising open data for social change requires focus and effort.

We’re only at the start of understanding open data impacts. In the upcoming Exploring the Emerging Impacts of Open Data in Developing Countries (ODDC), the Web Foundation and partners will be looking at how open data affects governance in different countries and contexts across the world. Rather than look at open data in the abstract, the project will explore cases such as open data for budget monitoring in Brazil, or open data for poverty reduction in Uganda. This way it will build up a picture of the strategies that can be used to make a difference with data; it will analyse the role that technologies and intermediaries play in mobilising data; and it will also explore unintended consequences of open data.

I hope in this discussion we can similarly focus on particular places where open data has potential, and on the considerations needed to ensure the supply and use of open data has the best chance possible of improving lives worldwide.

What do you think? You can join the discussion for the next two weeks over on the Striking Poverty site…