Reality Bending Lab › News › Open science vs. sustainability: a debate at the BNA Members' Meeting
Open science vs. sustainability: a debate at the BNA Members' Meeting
The British Neuroscience Association held its Members' Meeting online on 24–25 April 2024, and alongside the research talks, training sessions and panels it ran its Green Neuroscience Debate. I took part as the open science champion, opposite Charlotte Rae (University of Sussex) as the sustainability champion, with Hannah Hope, Open Research Lead at Wellcome, holding the funder's dual perspective between us. The motion, from the BNA's green neuroscience campaign, was this: a cultural revolution in sustainable neuroscience is needed more than a cultural revolution in open neuroscience now.
Charlotte organised the debate around a really interesting point of friction: open data is not free. Neuroimaging is heavy: a single study can run to hundreds of gigabytes before anyone has analysed anything, and publishing it means keeping it on a server that is powered, cooled, replicated and backed up, indefinitely, for a number of future downloads nobody can predict. The greenest byte is the one never written, and "share everything" is a norm we tend to adopt without ever costing it.
One of the alternatives on the table was a reasonable one: share derivatives. Compressed, preprocessed, cleaned, reduced, the data at the stage most reusers actually want it, orders of magnitude smaller than what came off the scanner, and cheap enough to host that nobody has to think about it again.
My position was that raw data is the cornerstone, and that almost everything we value about open science sits downstream of it. Reproducing a published pipeline needs that pipeline's input, not its output. Preprocessing is a long chain of contestable choices and a derivative bakes every one of them in permanently: you cannot un-preprocess a file to check it. And the datasets that have paid off best are the ones reanalysed years later with methods nobody had when they were collected, which only ever works from raw.
That does not make the environmental cost imaginary; it makes it the wrong place to start. Storage is a one-off write against an open-ended number of reuses, and a dataset that saves another lab from collecting the same data again has already paid for itself: which is an argument for open data on sustainability grounds, not against it. There are larger and far less load-bearing line items in what neuroscience spends.
Debates are staged rivalries, of course, and these two revolutions are not really competing for the same ground. But the exercise was worth the hour: open science has mostly not had to answer the sustainability case, and it will have to.
