Geoscience Reference
In-Depth Information
change from existing practice - but this does not imply they do not reflect genuine difficulties. A
key requirement of reproducibility is that data and code used in GC become publicly available - this
raises a number of objections (and potential problems):
1. Intellectual property : Algorithms created may be regarded as an individual's or institu-
tion's intellectual property - and to protect this, they should not be disclosed.
2. Commercially sensitive data : Some data may be commercially sensitive, and therefore
those using these data for their research may be unwilling (or not permitted) to share.
3. Confidential data : This is a similar situation to that given earlier, although in this case,
data cannot be shared since they contain confidential information about individuals (such
as medical histories).
Points 1 and 2 identify some key tensions between commercial and publicly funded research -
clearly, financial benefits from designing an algorithm or collecting information will be limited if
the outcomes become public knowledge. There are of course situations where such information may
be required to be made available outside the realm of a company - for example, a pharmaceutical
company may be required to give details of pre-commercial trials of a drug Godlee (2012), but there
may well be unwillingness to share unconditionally. However, despite the objections to data sharing
of this kind, there is growing demand for data from publicly funded research to be made available.
For example in the United States, the National Institute for Health (NIH) requires that research it
funds should be
… available in a timely fashion to other scientists, health care providers, students, teachers and
the many millions of Americans searching the web to obtain credible health-related information.
(Stodden, 2009)
Although these examples relate to the field of health-related research, they could certainly encom-
pass health-related GC.
There are also arguments that the sharing of information is an essential building block of scien-
tific discovery - and as Stodden (2009) also notes
… copyright also establishes rights for the owner over the creation of derivative works… using copy-
righted work in derivative research typically requires obtaining the permission of the copyright holder,
thus creating a block to the generation of new scientific discoveries.
This suggests that it is harder for commercially funded research to contribute to a more open phi-
losophy of scientific discovery and also maintain commercial sustainability - at least for companies
whose business model is strongly based on the exclusivity of their data or code.
The other objection to data sharing relates to point 3 given earlier - some data may contain
confidential information about individuals. In this case, the objection to sharing is motivated by the
protection of those people who are the subjects of the data - and the principle of anonymity clashes
with the principle of openness expressed at various points in this chapter.
17.5.2 w ayS f orward
Trends towards openness in research suggest that the need for some action to ensure reproduc-
ibility is needed. If the quotes relating to CRU and more general calls for reproducibility are
indicative of a more general trend in computational research, then GC may have little alternative
but to adopt these practices. Yet the arguments given earlier suggest that there are well-voiced
objections to an unconditional adoption of reproducibility. Therefore, plans to adopt a repro-
ducible approach should be done in a realistic way and consideration of the way forward must
include reflection on these issues.
Search WWH ::




Custom Search