Creating and disseminating lasting knowledge

Martin Vetterli © Manu Friederich/SNF

By Martin Vetterli

(From "Horizons" no. 104, March 2015)Scientific knowledge belongs to all humanity. Whether something was discovered by the Sumerians, the Arabs or the Greeks is of no importance any more: today it’s general knowledge. By the same token, scientific knowledge should be freely accessible to everyone, whether it’s to be used by scientists themselves or popularised among the public. Exchanging knowledge is a prerequisite for progress, and holding back information for strategic or commercial grounds has rarely had any positive impact on society.

But scientific knowledge first has to be generated. The litmus test for good science has always been its reproducibility. Results that one scientist has found reliable must be reproducible by others. In maths, where this process is at its clearest but where nothing is simple, a published proof must be detailed enough for it to be checked by someone else. In other fields, such as medicine or biology, detailed information on the experimental set-up can afford credibility to the results and conclusions. But the experiments must still be detailed in such a manner that others can reproduce them.

In order to ensure reproducibility, data has to be exchanged freely so that its quality can be tested and its scientific claims confirmed or rejected. This aspect of research was greatly assisted by the introduction of the printing press in Europe by Gutenberg and, later, by the creation of scientific journals and the peer-review process. It’s obvious that the invention of the World Wide Web some 25 years ago will have a similarly broad impact on research in future. This is the challenge of reproducibility in the digital age.

Here’s an example: in the year 2002, Grigori Perelman solved a mathematical problem and won the highest award in maths. However, he chose a rather unusual way of publishing his proof: instead of giving his manuscript to a leading maths journal, he put his work on the public preprint server The result was that it gained immediate attention and was checked straightaway. The whole scientific community acknowledged that Perelman had solved this 100-year-old problem. Even if his approach seems surprising at first glance, nevertheless this case presents all the characteristics of the classical scientific process: a known problem is investigated and its solution published so that other experts can check it. The subsequent consensus confirms that the problem has been solved. All that’s unusual here is the missing link in the normal communication process, namely the printed journal, along with the lack of a traditional peer-review process by a few internal experts – though this was replaced by an online review and an open debate on the Internet that was arguably even stricter.

In this sense, the challenge of reproducible science in the digital age is revealed as an opportunity. There are already sufficient online platforms for us to drive forward the process of creating, checking and disseminating new knowledge. And I’m convinced that current trends such as open access and open data are only the beginning of a massive shift towards completely digitised, open science. As the Beatles nearly said: “All you need is sharing!”

Martin Vetterli is President of the National Research Council and a computer scientist at EPFL.