To make a long story short, in the name of Real Academic Research
(involving simulations and analysis that will be run on our Connection
Machine 5, if that makes it sound more impressive), we're thinking
about building logging mechanisms into the Mosaic tools that track a
number of interesting information access parameters:
which documents are visisted
how long each document is visited
how many times each document is visited
... and in what sequence
how hyperlinks are used
how multimedia features are used
how search features are used
(in the future) how intelligent assistance is used
... and how it is not used
etc.
First, this would be entirely voluntary -- when you start Mosaic for
the first time, you would get a registration screen that, among other
things, would let you either say, ``Yes, I Want To Help NCSA'', or
``No NCSA, Stay Out Of My Cyberspace''.
Second, this would be entirely under user control -- it could be
turned off at any time, via a menu option and/or X resources and/or
whatever other methods we can come up with. (Or, if data accidentally
gets collected and sent to us when you didn't want it to, drop us a
note and we'll purge it from our systems.)
Third, it would be entirely automatic in that the log reports would be
sent, via email, by the program without the user having to do
anything.
Fourth, all information would be held in the strictest confidence --
it would go right to the project's social scientists and stay on
secure machines inside NCSA; even I (your humble hacker) would
probably never see it. We'll swear to this on a stack of religious
texts of your choice. Data will, of course, someday be published in
the generalized, analyzed aggregate -- but there will be no way for
anyone to trace back anything to any user anywhere.
Fifth, and this is the best part -- cooperating with us on this and
helping us gather this data would essentially mean more support for
the Mosaic project, enabling us to build more and better tools for
you. And, the information we learn from the analyzed data will be fed
back into our development -- a particularly exciting concept for us,
because we want to start building intelligence into the systems, and
we need this type of analysis to have something to build on.
So... what do you think? Should we do it? Should we not do it? Will
anyone rebel and refuse to use our software outright? Given the
parameters outlined above (esp. the completely voluntary & under user
control parts), could it still be considered to be an invasion of
privacy, or possibility thereof? Does it make you hopping mad to even
think that we'd consider it? What?
Thanks for your help,
Marc
-- Marc Andreessen Software Development Group National Center for Supercomputing Applications marca@ncsa.uiuc.edu