¶ 1 Leave a comment on paragraph 1 0 The concept of networks as a key organising principle for all of my thinking emerged slowly. I can’t trace the exact point at which the idea of a network moved from a being a technical infrastructure to becoming a central concept. Early on in my writing I was intrigued by our ability to look at a lab notebook as a network of processes, materials, and data. A group of us attempted (unsucessfully) to obtain grant money to build a network of early adopters of Open Science. With Deepak Singh, I had a reputation as a while for being critical of attempts to transplant the success of social networks (twitter, friendfeed, facebook) into the research space.
¶ 2 Leave a comment on paragraph 2 0 The idea that a real understanding of networks as an abstract concept was central almost certainly came primarily from John Wilbanks. Appropriately the pieces of that idea came from many people including Michael Nielsen, Deepak Singh, Shirley Wu, Jon Udell and others but I believe I can trace a moment of crystallisation in a post I wrote summarising a Barcelona Science Commons meeting in 2008. John in his closing comments said “We are the architects of Open”, a sentiment echoed in a roughly contemporary slidedeck1, retrieved from http://www.slideshare.net/wilbanks/health-commons-intro] in which he states “we’re the network engineers for science”.
¶ 3 Leave a comment on paragraph 3 0 This idea, that by understanding networks in general, we can seek to architect solutions to problems was the seed that led to many things. At first this was a focus on interoperability in licensing of content and data. Over time this expanded to an understanding of interoperability more generally. It was John who introduced me to Metcalfe’s law, which shows that the value of a network expands more than linearly as a function of the number of connections. If the question is how to maximise the connectivity of a network, to reduce the friction of information transfer, then we can treat this as a design question. The object becomes to understand how to design and grow networks generally and to use this to engineer systems that would grow and attract researchers to use and contribute to them.
¶ 4 Leave a comment on paragraph 4 1 “Network Enabled Research” is one of my earliest attempts to synthesise these ideas. The focus here is fundamentally on licensing, but the core idea that licensing is just one piece of the puzzle is there at the core. It’s worth noting that this piece was written in response to Heather Morrison’s views that the Creative Commons Attribution license was not a good fit for Open Access. Her view was obviously at odds with my own, and remains so today, but it was important as the provocation that lead me to attempt this synthesis.
¶ 5 Leave a comment on paragraph 5 0 That synthesis came quite late – the piece dates from 2012. Moving back in time to two years earlier, “Why the Web of Data Needs to be Social” shows a more specific piece of that thinking developing earlier. It focusses on how traditional social networks are not a good model for data sharing – a theme that is in turn prefigured by the earliest piece, another two years older. The original version of “Five Rules for a Scholarly Social Network” was written in response to a question from Ian Mulvany, then lead developer for the bookmarking service Connotea, now Technical Director at eLife. It was also cited later as an influence on the design of Mendeley, the academic bookmarking service that was ultimately purchased by Elsevier. The idea that understanding the network configuration can help design successful systems runs through both pieces but is not yet fully worked out. Of course the naive creation of social media sites for researchers that seek to transplant the success of Twitter and Facebook continues to this day.
¶ 6 Leave a comment on paragraph 6 0 Five Rules finishes with one of my earliest articulations of the idea of navigating through a network of resources, with the ability to “pivot” from one view or axis to another. This wasn’t new at the time and a similar view was actually part of an early version of Mendeley. Yet this idea remains largely unrealised in practice. Ultimately this failure to realise the potential that a new view of a system can offer is also a common thread throughout this whole book.
¶ 7 Leave a comment on paragraph 7 0 If we return from this earliest piece to move forward in time, “Five Rules” while negative in its way is built on the idea that there is a huge potential for social tools in the research process. “Why the Web of Data Needs to be Social” is also aspirational in the possibilities it sketches out but it is in “Best Practice in Science and Coding” that we see the most straightforward vision of technical solutionism. Here the solutions that could improve experimental research are already available; version control, continuous integration, test driven design. They even have clear parallels that could make them understandable to experimentalists, but some years on even the effort to engage researchers who do computation with good practice has been a slow process, and the consequent crisis, whether real or imagined, in reproducibility of research from computation, to social psychology via the biomolecular and biomedical research lab continues.
¶ 8 Leave a comment on paragraph 8 0 These failures can be ascribed to a number of causes, many of them actually the same problems described in “Five Rules”: a lack of clarity as to what the improvement is, or what it is for, a lack of delivery of any actual improvement in real settings but most commonly failing to adhere to the first principle, working within the existing workflow.
¶ 9 Leave a comment on paragraph 9 0 That idea of sitting within an existing workflow, and improving on it, may seem at odds with (the usually over simplified presentation of) Christensen’s models of disruptive innovation where the disruption is supposed to occur initially through adoption is a space adjacent to common use of generally inferior products. The thread that brings the two together is the network effects. New services or approaches or ways of thinking are competing against network effects that bolster the incumbent. The second principle in Five Rules, show immediate benefits that don’t require network effects is the key. Fitting a change within one small part of the existing workflow and doing that piece well, while not disturbing the traditional tools elsewhere in the process is not just valuable but crucial. The existing process already benefits from network effects and overcoming that is essentially impossible from a standing start. Clever choices about which part of the process to “disrupt” so as to work first time, but then build network effects over time is the key. Mendeley did this well. Figshare did it well. Not by building social networks to start with, but by building a base on which they would thrive.
¶ 10 Leave a comment on paragraph 10 0 Disruption, or at least change if it is to be successful needs to address existing needs, and it needs to do so from a disadvantage. This theme will return in the later section on Assessment and Metrics. Can we hack existing systems of research assessment to drive adoption? Can we find points of leverage where change is possible? The final piece in this section also starts to lay the groundwork for this. Originally written as an editorial for Open Access week in 2013 it turns its attention away from the old argument of whether “open networks are good” to a more practical question – how can we tell where open projects are most likely to be successful? While not offering a direct answer it aims to sketch out an analytical approach that could start to address the issue. Across these set of pieces we’ve moved from aspirational techno-solutionism to a more subtle question, how could we tell when this approach will work?
¶ 11 Leave a comment on paragraph 11 0 Throughout all the pieces in this section the question of architecture is the common thread. The question of technical design is the main focus, developing in a range of contexts towards a synthesis that sees researchers, articles, data and information as nodes on a network that we can seek to understand. If we can understand the network we can understand how to make it work better. It almost doesn’t matter what those nodes are, indeed a focus on what the nodes are in one context (people in social networks) leads to a misunderstanding of how the network needs to be built in another (for data sharing).
¶ 12 Leave a comment on paragraph 12 0 This technical approach is productive, but there is an irony that runs throughout all these pieces starting in 2008. The underlying assumption of Five Rules is that a sufficiently large and engaged community cannot be built directly. Throughout all of this section the technical benefits and approaches are presented without really considering how to support the communities that will build and care for these networks. At the end of this section we have a view of how the system can be understood, how it can be studied, even how its dynamics might be analysed to identify where the easiest wins might be found. But in this abstraction we’ve lost some of the key questions. Who are we building this for? Who benefits? And why? Those were questions that I only came to, and we will only come to, later.