On Fri, 2012-07-20 at 10:22 +0100, Stefan Decker wrote: > The discussion seem to point to a deeper question: how to enable crowd > sourcing of the analysis of these kind of data sets? This may involve > running of analysis code or maybe even manual work. > What kind of computational infrastructure would we need to enable > this? And how do we validate and aggregate results?
Unfortunately, in the USA at least, the biggest barriers are not technical, but social, because: (a) health information privacy laws such as HIPAA http://www.hhs.gov/ocr/privacy/ make it difficult or impossible to publish the raw data that would be most useful for research; and (b) researchers do not have the incentive to publish their data that might allow other researchers to make discoveries. There is a tension between privacy and the usefulness of data for research, because full de-identification removes information that can be critical to determining cause and effect, such as dates, times and locations. We need better ways -- both bottom-up, such as http://weconsent.us/, and top-down, such as legal changes -- to both encourage the availability of research data and to facilitate appropriate access to it, such as establishing well-defined tiers of access for different purposes. We need technical solutions that will help us work through and around these social barriers. -- David Booth, Ph.D. http://dbooth.org/ Opinions expressed herein are those of the author and do not necessarily reflect those of his employer.
