At the end of this month, the Open Data & Privacy project (funded by OSF) formally closes. Over the past year (during which we implemented the project), Javier Ruiz and I (as the core team) have attempted to gingerly navigate this complex environment of trying to understand if and how  open (data) and privacy can find a balance. I would like to believe that we have done this with some amount of success. Of course there has always been the support of a really knowledgeable expert community to tap into.  And I would like to duly acknowledge the input of in particular, Malavika Jayaram, Mark Lizar, Reuben Binns, Antti Poikola and Walter van Holst. Their amazing individual contributions are dotted in various posts on this website.

It has been a great period of learning and also growing in understanding of the issues we are dealing with from the varied contexts and sectors. I have attempted to document all the learnings from the project over the past year in periodic blog posts on this platform. This post however attempts to summarise (and highlight) what are the big (but less specific) aha’s for me based on my own critical reflection on the design and execution of the project’s key activities.

• Community is extremely important

Community is always valuable for any open data initiative and it’s not any different for the Open Data & Privacy project. However, I can say that because of the peculiarities of this project- the complexity of the issues, and the varied contextual nuances, community is extremely important! Well for one, as the project name suggests, having expertise in both open data and privacy is essential, but finding individuals who possess a deep understanding of the issues in both fields with significant depth is not easy. In essence, delivering on the projects core promises (which includes key knowledge outputs) requires having individuals on board that can engage with any number of these areas.  An interdisciplinary Working Group composed of various expertise is therefore vital. On the flip side however, this (having such a mix of people) also presents a challenge to the group in that there are quite strong individual ideas coming across often that find very little common ground with the others. This certainly has its disadvantages as it can stall activities that need some form of consensus to proceed. I discuss this in the next point.

• You don’t always have consensus, but that’s okay

Being a project that is intended to be community-driven, having buy-in is essential.  However, buy-in often requires some form of agreement on the issues, and for this community, that is not so easily achieved. For instance, at the very basic level, an agreement on what the key terminologies being used in the space mean is desirable. However, as experienced with the never-ending exchanges on many project platforms (principally the mailing list) there are no common interpretations of most of the terms that would be acceptable to each and every member of the group. Additionally, some of the more problematic definitions are still evolving somewhat. Even more complex are the issues (and the proposed solutions where applicable) and how they are manifested in different contexts, and for which there was little agreement about what is in fact the case. The effectiveness (or otherwise) of anonymisation is a case in point. It became fairly obvious (to us) much later on in the course of the project that it is entirely possible to work around these disagreements. For example, with regards to the terminologies, by having a living document (on wiki) which can be edited by the community, this issue of evolving definitions/interpretations is addressed.

• Finding a ‘one-size fits all’ solution is often not the answer
  

This point relates closely to the previous two.  In our attempt to get a set of community-driven principles developed to guide data publishers, we found out quite quickly that the requirements  of the various sectors (disciplines) are quite different and a ‘one size fits all’ solution is not only undesirable but also not feasible. To illustrate, the requirements of  government data publishers are are quite different (as some disclosure laws can apply)  from their counterparts in the private sector. Additionally, certain types of datasets, for example health data and location data are generally more sensitive and need to be uniquely treated. Data protection laws (and their interpretations and applications) also differ in various localities across the world.  Once again, finding a way around this was necessary. The solution was to design the principles to address a specific sector and context.  For example, the engine room has done some great work leading the development of a handbook (on responsible data handling) for the development sector. Open Knowledge’s Open Data Handbook will also contain a set of privacy principles more geared towards public data.

• Changing direction comes with the territory

As a concluding point, I can note that when working in an environment as complex and as unpredictable as the one presented by open data and privacy  considerations, it is inevitable that one gets pulled in directions different from what one intended on the outset. This is particularly so in relation to focus areas, and the specific approaches adopted towards achieving the project’s objectives. For example, in the case of the former, it was easy to see that though the project was intended to address privacy issues surrounding  open data specifically, there was a need to also address some other ones because the issues (such as data sharing, and internet security) are indeed interrelated.  Being agile and broadening our scope to support work in these areas was therefore necessary.

I would like to highlight that this project was intended to explore the environment of open data and privacy- to understand the issues and the actors, what is required to address the identified issues, and what is feasible given the resources (human, financial) that are available. I do believe therefore that the lessons learnt during the course of implementation thus far (some of which I have shared above) will greatly shape our design of future interventions in a more positive and efficient way.