As I mentioned in an earlier message, I am an optimist, by nature and
by nurture, so Jay's pessimistic "only choice" of either global dieoff
or dictatorial world government seems utterly wrong to me. If I was
inclined to argue the matter I'd probably start with Clarke's First Law:
When a distinguished but elderly scientist states that something is
possible, he is almost certainly right. When he states that something
is impossible, he is very probably wrong.
In this dictum Arthur C. Clarke is suggesting that it is wise to bet
on human ingenuity, which often finds a way to do what others said was
impossible.
But I am NOT inclined to argue the point, for as I said in the message
announcing my social technology mailing list, I prefer algorithms to
argument. So instead of arguing the point that we will find a way out
of the mess, I'd like to present a sketch of my own efforts in that
direction. I've just posted this to the social technology list, so
some of you may have seen it already, minus this preamble -- while
the rest of you should visit http://www.island.net/~dpwilson/stml.html
to find out about the list.
So, here it is, the first step in my own plan to get us out of this mess:
If any of you have read my web pages you may be aware that I have an
interest in matching and optimization for social purposes. That is
something of an understatement -- it might be more truthful to say I
am obsessed with using matching and optimization for social purposes,
and claim we can solve all of our leading social problems that way.
If you haven't at least tried to read some of my web pages, I am
afraid that the rest of this message will make no sense whatsoever, so
you might as well skip it!
So far my work on matching and optimization has all been theoretical,
though I've done a few experiments with simulated data, but in the
near future I plan to collect and process real data via the web.
This is a brief survey of what I am planning to do, and I hope you'll
give me some feedback on it. I am especially interested in solving
the privacy and security problem, and so I'd greatly appreciate
comments on that part of the plan.
The basic idea is to offer a range of matching services via the web.
These will, of course, be free services, but to take advantage of them
the uses will have to provide some data about themselves. The object
of this exercise is to demonstrate what can be done with that data.
My hardest problem is to find a way of ensuring privacy or anonymity,
so that people will be willing to provide this data. It is not easy
to guarantee privacy and still be able to provide useful suggested
matchings.
I would prefer to do this in such a way that I have no access at all
to anything that can identify the person providing the data. For
example, at one time there was an anonymous remailing service
operating out of Finland. Some service like that could be used to
hide from me the source of an e-mailed reply form. It is a little
more difficult to then get the processed data and suggestions back to
the submitter, but for purposes of explanation let's assume I've found
a way of doing that, and have nothing more than a code word or number
to identify each user.
Eventually I'd like to be able to solve very important matching
problems like matching people with jobs, or matching people with
potential spouses, but to get the process underway it will be
necessary to start with much simpler problems.
Indeed, the first step probably does not involve matching at all. As
you may guess, it is very difficult to do any kind of matching until
you have some data, so the first step is to provide a very limited
service, some form of data analysis and reporting.
There are several well-known sets of descriptors for personality, such
as the Myers-Briggs or Cattell descriptors. As a first step I plan to
put up a web page that lets the user fill out a questionaire then
provides a personality analysis or description which the user can save
and may find useful. There are already several such services
available on the web, and to begin with I plan to just offer another
one -- but it will also collect a small amount of other information,
for use in step two.
In step one I will ask a few more questions than are necessary for
personality analysis: I plan to ask for information about a person's
taste in music, movies, art, books, and other preferences. Then in
step two I will be able to provide another useful service, one that
recommends music, movies, art, and so on. Again these are also
services that are currently available on the web, and the underlying
mathematics is fairly simple. It's basically a question of finding
people who share your taste (by comparing your questionaire responses
with theirs), then recommending what those people said they liked.
Step three is more interesting -- I plan to provide some form of
limited matching service for individuals, as friends, possible
co-workers, and eventually even potential spouses or sexual partners.
Again, there are services on the web which do this, though I don't
think they've been very successful.
The underlying mathematical problem in this case is much more
difficult, involving combinatorial optimization and a lot of
processing, but there are well-known matching algorithms that can be
used. My plans involve the use of such algorithms, together with what
I hope are new ideas for conditioning the data, and I have some
reasons to believe I can make it all work.
Step four is the hardest, since it involves trying to match people
with jobs. Even this has been done before, sometimes with reasonably
good results, but I think it can be done better.
These are ambitious plans, but each individual task has been done
before and it seems obvious to me that combining them can only make
them easier. The real problem in doing anything like this is to coax
the user into providing lots of good data. I hope that providing
multiple services will encourage people to experiment with them, and
perhaps to provide a little bit more data each time.
I have several goals here, including the obvious one of using these
services myself. But mostly I just want to acquire some empirical
evidence to back up my extravagant claim that we can solve all of our
leading social problems with matching and optimization methods.
Obviously this is just a sketch, but before I provide any more details
I'd like to run this by you and see what kind of reaction I get. For
more information about these ideas, please see my home page at the URL
given below.
dpw
Douglas P. Wilson [EMAIL PROTECTED]
http://www.island.net/~dpwilson/index.html