Not To Share

For this week I am examining the findings reported in a paper by Caroline Savage and Andrew Vickers about data sharing published on PLos. The authors of the study devised an experiment to request that researchers whom have published a paper on PLos carry through with the data requirements stipulated by the publishing agreement and actually make the data referenced in their study available.

It should probably come as no surprise that out of the ten researchers contacted only one was willing or able to actually produce the dataset, and then only after talking with the requester about what would be done with the data.

This raises so many of the key issues concerning scientific data sharing that it could almost act as a summary of why instituting a data sharing methodology has been so problematic. Some researchers were “too Busy” to fulfill the data requests, even though “we pre-specified that any request must not create undue work for the investigators”. Others claimed that they were “forbidden” to release the dataset. Some simply did not respond to the request at all.

The reasons given for not sharing data are easy to sympathize with, gathering a data set, annotating it so that others will find it useful takes an investment in time. It is difficult to persuade anyone to take a few hours out of their busy day to give you something they feel attached to, when they have no idea of your intentions or how important the data is or even who you are. It is illuminating that the one researcher who complied with the request did so after an interview process; almost like a vetting procedure to see if the requesters intentions were true.

It must be pointed out that however reasonable the denials are they come from a group of people whom have published in a journal which explicitly states the requirement that datasets need to be accessible. This active participation in the journal should nullify any claims of being “too busy”. It does not, and this demonstrates one of the key faults with data sharing as it is: journals lack the teeth and methodology to ensure that datasets be made available.

The main reason for this is that journals are not and should not be concerned with enforcement, journals should not be the police ensuring that everyone play nice with data. The best a journal could do is to deny further publication. The real and effective enforcement should come from the community in general. There needs to be a ground up call for data sharing. This sort of change in the way science works will only take place slowly and if there is a proper system in place to reward data sharing.