Exploring the disappearance of research in discontinued online journals, the reuse (or lack thereof) of clinical trial data, and the revolt against Nature’s new machine learning journal.
When research disappears via Wired
This article follows the story of the ill-fated Online Journal of Current Clinical Trials, which was founded in 1992, as one of the earliest online-only journals. The journal was ahead of its time and found little success, shutting down only a few years later but not before it had published 80 research articles that effectively ‘disappeared’ when the journal was taken offline. One of the authors of such an article attempted to find these lost papers and make them publicly available. Her ongoing struggles are captured here.
A manifesto for preregistration via PNAS
A lot of things seem easier with the benefit of hindsight; from the privileged position of the future, it is much easier to see why things did or didn’t work. In the context of scientific research, however, this hindsight may undermine the objectivity of a study. This paper, published in PNAS, outlines the argument for the growing trend of trial preregistration, which is intended to help researchers to control for hindsight bias.
Calls for the sharing of clinical trial data, particularly from studies funded by the pharmaceutical industry, have maintained a high media profile for several years now. This research letter looks at the availability of trial data to researchers wanting to use them, and what was done with the data once a successful request had been made. The authors found 3255 available trials, for which only 113 data-sharing agreements had been formally completed. Of these, only one led to a successful publication. The authors suggest various reasons as to why these data, even when made available, are rarely sought out and reused: the platforms are too segmented, there is little incentive for researchers to conduct validation studies as they are more difficult to publish, and the process is long-winded.
Unlike in other technologically focused economies such as Japan and the USA, the copyright laws in the EU do not, at present, permit text and data mining. The new Copyright in the Digital Single Market Directive, which has been devised to bring the EU copyright law in line with today’s digital realities, includes an article introducing an exception that would permit text and data mining for strictly non-commercial research purposes. This article argues that, although the EU proposal has been prepared with the right goal in mind, the exception would not do enough to help researchers in academia and industry to make the most of text and data mining.
This week, Nature announced the launch of a new machine learning journal. Such a journal doesn’t sound too controversial but it has elicited complaints from thousands of researchers and several tech giants, such as Google and Facebook. The offence? Its subscription publishing model. In a field as young as machine learning, researchers are used to publishing in open access journals and do not intend to move back towards paywalls.