[e2e] was double blind, now reproduceable results
touch at ISI.EDU
Wed May 26 12:23:06 PDT 2004
David P. Reed wrote:
> If there is no way to verify that the data is real, why bother to
> publish it?
There are valid results that can be derived from data that cannot be
anonymized. That means the author should be willing to run tests for
others at least, IMO.
> If the issue is prevarication, anonymizing protects the prevaricator.
If there exists an anonymization that doesn't kill the experiment, sure.
But that presumes:
a) such anonymization exists for that experiment
b) the data source is comfortable with the anonymization
Neither necessarily applies.
> If the issue is verifying the calculations, the anonymizing transform
> may be more unreliable than the raw data.
> Once transformed and released, there is neither revocation, nor logging
> of use, and so if the anonymizing transform is faulty, there is no way
> to monitor misuse.
> Why not put the raw data in a secure place guarded by the original
> investigators, where people can submit statistical queries against it?
If escrow were feasible that'd be a way to go to. But that's not always
> Of course that can leak info, too, but no more than anonymizers can, and
> at least the query programs actually run and the credentials of their
> authors can be recorded and logged as evidence of bad intentions, and
> vetted to a certain extent.
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 250 bytes
Desc: OpenPGP digital signature
Url : http://www.postel.org/pipermail/end2end-interest/attachments/20040526/5389b22a/signature.bin
More information about the end2end-interest