Interview with Michael Nielsen, author of Reinventing Discovery: The New Era of Networked Science

Welcome to the latest installment in my very occasional series of interviews with people in the scitech world. This time around the subject is Michael Nielsen, author of the recently published Reinventing Discovery: The New Era of Networked Science and prolific speaker on the Open Science lecture circuit. A recent example of his public speaking is his TEDxWaterloo talk on Open Science.

You can follow his blog here and read his recent Wall Street Journal article, The New Einsteins Will Be Scientists Who Share.

I'd like to thank Michael for his provocative and insightful responses. Enjoy!

Q0. Hi Michael, would you mind telling us a little about yourself and how you ended up writing and speaking about open science?

My original training is as a theoretical physicist --- I worked on quantum computing and related topics full time for about 13 years, and part time for a few years prior to that.

But at the same time as I was working on quantum computing, I was also following closely all the amazing things happening online -- things like the development of Google, Wikipedia, open source software, and so on. And as I watched it came to seem to me that these tools have begun (though far from concluded!) a revolution in the way we construct knowledge.

For a long time I expected that tools like this would also revolutionize how science is done. And we've certainly seen some exciting developments along those lines. But overall scientists have been very conservative in how they've adopted new online tools, in large part because of cultural barriers in science, barriers that mean scientists don't get a whole lot of credit for sharing knowledge in new ways.

I found this conservatism frustrating, and wanted to work to help change the culture of science. So in 2007 I decided to leave my tenured position as an academic to work full time on open science.

Q1. Your new book is Reinventing Discovery: The New Era of Networked Science. Briefly, what is it about and what is the intended audience?

It's about the potential of the network to transform the way scientific discoveries are made. I think the day-to-day process of science will dramatically shift over the next few decades, speeding up the rate at which discoveries are made, and making possible whole new ways of attacking problems. But that will only happen if the culture of science becomes much more open -- to reach its potential networked science must also be open science. And so the book is also a manifesto for open science.

Q2. A significant percentage of the people doing science out there are academics. It's easy to see how open science integrates into the research part of their jobs but how about into their teaching and service requirements?

There are many things academics can do to integrate open science into teaching and service. Here's just a few ideas:

1. Contribute to projects like Wikipedia and Citizendium, perhaps by giving students projects to improve articles in particular areas.
2. Academics can potentially combine research, teaching and outreach through projects such as Zooniverse, which is becoming a general purpose platform for connecting scientists to the general public, so the public can make real contributions to scientific research projects. Zooniverse are probably best known for Galaxy Zoo, a very successful project to crowdsource galaxy classifications, but they also run many other citizen science projects.
3. Academics can upload some of their teaching materials online, where they can be used by others. Aside from the intrinsic worthiness of doing this, it can certainly help improve teaching. YouTube, for example, gives detailed analytics -- you actually get a graph showing how much attention people pay to different parts of your video. From painful personal experience I can say that sometimes that graph plummets, as people leave your video in droves. Usually that's a great diagnostic that you're messing something up in your explanation, and need to improve.

Once you start looking into these and other similar possibilities, you realize that there are a multitude of ways to incorporate open science into the classroom and into service. Many of these ways are free or inexpensive, with the main limit being imagination.

Q3. It seems to me that the key to changing the way science is done is changing the incentive structures for working scientists. What could a new incentive structure look like that would encourage more openness? Are there some practical steps that can move things forward?

This is a question that an entire book could easily be written about. With that said, here's a few things that can be done:

1. Individual scientists can make a point of citing non-traditional research contributions, like open data sets, code, and videos. Eventually we'll see journals that make it possible to publish data, code and video as first-class research objects in their own right, with the same status as conventional paper publication. Some efforts in this direction include GigaScience, Open Research Computation and the Journal of Visualised Experiments. Citations to those contributions will then show up in conventional measures of academic productivity --- things like citation count --- and so give people an incentive to contribute in new ways.
2. People can build tools to measure the impact of non-traditional research contributions. The SPIRES service helped drive the adoption of preprint culture in physics, by providing a way of measuring the impact of preprints. There's no reason similar services shouldn't be set up for contributions to blogs, wikis, question and answer sites like MathOverflow, and so on. Indeed, MathOverflow already has a tool like this built in --- a measure of reputation for users. And there's other ideas exploring this space, like altmetrics and total impact. Do these replace conventional measures, like total number of citations? No, of course not. But people are often surprisingly aware of such reputation measures, and they will gradually enter the mainstream, show up on people's CVs, and so on.
3. People who work at grant agencies or in senior positions in academia can help legitimize new forms of contribution. Simply inviting scientists to submit non-traditional evidence of impact would be a good start.

These are all small but significant steps, and it's through such steps that a change to a more open scientific culture will gradually come about.

Q4. Or perhaps the key is to get them young: how do we need to change the training and mentoring of scientists get to encourage them to be more open?

I don't think there's anything terribly complicated required here. Just getting students involved in open science projects is a big help. People like Steve Koch have mentored students like Andy Maloney, who've done much of their work in the open. Those students then go off and carry those techniques elsewhere, slowly changing the overall culture of science.

Q5. Perhaps the classic example you use in your talks is the Polymath project -- an experiment in massively collaborative mathematics. Do you see a future for this type of project and do you think the model is generalizable beyond mathematics?

Yes, I see a big future for this kind of project, although I think that Polymath and similar projects will morph into other forms. The original Polymath Project was done using off-the-shelf tools -- Wordpress and Mediawiki -- that definitely aren't designed for massively collaborative mathematics. And so I think that we can develop much better tools, and also better social norms, that will make it possible to go much, much further.

To some extent this is already happening with the question and answer site Mathoverflow, which has attracted a strong and growing community of mathematicians. It's not uncommon to see a challenging technical question posted to Mathoverflow and then answered within minutes or hours.

As to whether this model is generalizable beyond mathematics, it certainly is, although with some qualifications. It depends on where the bottlenecks are in doing research. If the major bottlenecks are (say) construction of an experimental device, or taking samples, then obviously the net only helps a little. But if the bottlenecks are data analysis, or something more conceptual -- and I don't just mean theory, the bottlenecks in doing experiments are often conceptual -- then there is the potential for a networked approach to really help. What gets me excited is the fact that we're still in the very early days of this; there's a lot of room for people with imagination to go much, much further.