Hello, one and all, how are things? I find I can only do so much
programming before my urge to write plain English overpowers me, so
I started to rewrite my home page. Not so plain English, yet, I'm
sorry to say; it still needs a lot of work. But I'd like to run
it past you, to see if anyone can suggest a better approach, or
catches one of the many errors. It's actually got a lot to do with
catching errors, as you will see. I will be quite grateful for any
comments you might care to make, even the most scathing ones.
dpw
Douglas P. Wilson [EMAIL PROTECTED]
http://www.island.net/~dpwilson/index.html <==== new text below to replace this?
http://www.SocialTechnology.org/index.html
----------------------
On my old home page I asked readers to imagine a future world which
would be better for everyone:
Imagine a future world in which it is easy to find a good job.
Imagine a future in which it is easy to find a truly compatible spouse
or sexual partner. Imagine a world in which it is easy to find good
long-term friends.
I went on to argue that such a world would be an affluent one, without
poverty, and one with almost no crime, since people with good jobs,
good friends, and a compatible spouse are much less likely to commit
crimes.
All true, I maintain, but a bit too much for most people's imaginations,
so I'll have to add a lot of explanation and present my concrete plans.
What I have sketched above relates to something I call "pure social
technology". People who follow the stock markets hear a lot about
"information technology" and may have been told that the dominant
technology of the new millenium will be some form of information
technology involving the internet. I certainly don't deny the
importance of information technology or its social side effects, but I
think the future belongs to pure social technoloyg, technology that is
intrinsically social, not merely a side effect of better communication
and transportation.
In our twentieth-century culture people somehow find jobs, spouses, and
friends -- or don't. Far too many people never find a good job, and are
either on welfare, begging on streetcorners, or stuck in a dead-end job they
hate. Far too many people never find a truly compatible spouse and are
either lonely and embittered, move from one unsatisfactory relationship
to another, or are stuck in a marriage with someone they've grown to hate.
This is not a satisfactory situation. It is unacceptable. I believe
society must provide a reliable method to guarantee that each person
has a good job -- one they will enjoy and learn from, or at very least
one that compensates them well for their time and effort. I believe
that society must provide a reliable method to guarantee that each
person has a truly compatible spouse or sexual partner. And I believe
that society must have appropriate mechanisms for finding or growing
good friendships.
Much more about social technology will follow here and on other pages,
but I think it important to sketch out the plan I'm working from and
hint at the technical details. A key part of this plan is the
creation of a non-profit society to prototype new social technology
and promote its use. If I'm right about the eventual impact of this
new technology it will transform our world beyond recognition, perhaps
making the business sector less about making money and more about
serving human needs, but in the next few years, as this process gets
underway many businesses will be set up to exploit social technology
and many people will makes lots of money doing so.
Let me be clear about my own role in this. My own list of personal
heroes does not include Bill Gates or anyone else who has made a large
fortune from exploiting information technology -- instead it includes
people like Richard Stallman of the Free Software Foundation and Linus
Torvald, who wrote the Linux kernel and started the remarkable global
development effort that made Linux a success. So I am trying to
create a non-profit society, SocialTechnology.Org, to do research and
develop free software aimed at social change. It will also promote
social technology and help people make use of it. I have no objection
to people making money from social technology and plan to encourage
people who want to do that, but I am not one of them myself.
(In fact I am totally incompetant when it comes to money and couldn't
make a penny from social technology even if I spent the rest of my
life trying).
Last year I started the SocialTechnology mailing list, and I began
this year by registering the SocialTechnology.Org domain name and
putting up a preliminary home page for the (future) organization.
I spent much of last year writing about social technology and putting
up web pages about these ideas, but this year my emphasis has shifted
to data collection and the development of new software.
Social scientists have been collecting social survey data for many
years, but it has been used mainly for research by those same social
scientists, who have written many academic papers about their
findings. Now it is time to make proper use of all that data, making
it the foundation of a new high-tech industry.
Social technology includes many things, but on this page I want to
concentrate on my own plan for pure social technology, the creation
and maintainance of social structure -- the realization of that
utopian future world I've written about, or something like it.
Because this is a plan meant to be taken seriously and acted upon, it
is somewhat technical in places -- the kind of people I most want to
communicate with will want to see some technical details up front. If
that doesn't include you, please jump to a non-technical overview
page.
The first technical point to make concerns the underlying mathematics,
and the word 'mathematics' will immediately make some people turn
away from this page. But please note that the eventual software
products will hide the mathematics from the end users (unless they
want to see it -- anything I produce will be open-source). To write
this software people will need to know some Discrete Mathematics,
particularly Graph Theory and combinatorial optimization, but the
end-users won't need to know what a triangle is.
The word 'optimization' is again a provocative word which will
displease many people for whom it suggests "Big Brother", centralized
organization, and all the other negative connotations of what was once
called "Social Engineering". I emphasis the phrase 'Social
Technology' instead because I envision the creation of tools and
techniques, and explicitly reject all authoritarian attempts to
"optimize" or "engineer" society.
Some of the tools and techniques to be created will indeed use
combinatorial optimization methods, and in a larger sense the
development of social structure is clearly a combinatorial
optimization problem, but nothing authoritarian is intended. On the
contrary the new social technology will permit society to be much
less centralized and will encourage new freedoms.
Discrete mathematics is involved simply because people have only
a limited capacity for social interaction. Today most people have
one or perhaps two jobs, if married have only a single spouse, and
live in a household with an integral number of people. Friendship
is less restrictive and may shade from slight acquaintances to much closer
bonds, but it is still best modelled with links between people
forming a graph called a sociogram.
(Insert GIF of suitable graph here! Something like this, perhaps:
A B C
O<--------->O<-----------O
| |
| |
O---------->O----------->O
D E F
but with many more vertices and edges).
Sociologists have been using sociograms for many years, and discrete
mathematicians have been working with similar looking graphs for
many years, but somehow the obvious connection has never materialized.
-- Sociologists describe small sets of mutual friends as "cliques",
and indeed discrete mathematicians talk about "clique detection
algorithms" for finding mutually interconnected vertices.
-- A familiar combinatorial optimization problem is the bipartite
matching problem, which mathematicians sometimes refer to as
the assignment problem, since bipartite matching algorithms could
be used to match workers to job assignments.
-- Another familiar mathematical problem is the so-called
"stable marriage problem" which addresses the problem of creating
a stable solution to a bipartite matching problem amongst mutually
attracting couples.
But still, the obvious connection has never materialized. Combinatorial
optimization is not used to match workers to jobs or generate stable
marriages. I can assure you, having spend many years trying, that it
is very difficult to get anyone to seriously consider applying such
mathematical techniques for anything involving real human beings.
"But wait", you might ask, "what about computer dating?"
Someone always asks that.
A "dating service", whether computerized or not, is something I don't
like at all, for several reasons:
--- most dating services are scams aimed at making
money from other people's lonely desperation
--- dating services involve a small pool of people, but
the quality of matching depends on the pool size
(there may be a lot of fish in the sea, but not many in
pothole-size puddles)
--- as far as I know most dating services that claim
to use computers don't use computers with any real
understanding of either human personality or discrete
mathematics, and certainly don't use social survey data
--- dating is a terribly high-pressure experience that makes
true communication between people unlikely
--- my own personal experiences with dating are mostly unpleasant
memories, even though I already knew the people I dated
--- most important, to me, but hardest to explain, dating services
attempt to find people "best matches", equivalent to the
use of a "greedy" algorithm and that kind of algorithm just
doesn't work for bipartite matching.
People who truly understand combinatorial optimization recognize that
most good algoriths are equivalent to "local search" in an appropriate
space. Defining the space is the key problem -- it is really a kind
of phase space, not unlike those used in thermodynamics, a space of
possible configurations, which may indeed occur sequentially. Local
search is the search for a configuration which is very similar to the
current one, but better. If you find it, then you search for another
one, also similar-but-better.
What does this mean in the real world? I can give you my answer,
and of course I will give you my answer, if I can persuade you to listen,
but the real solution will involve lot of discussion and careful analysis.
All I am seeking to do right now is get your attention. I want you
to participate in this discussion and analysis -- I don't want you to
pay too much heed to my own ideas.
For me, the answer is a progressive enrichment of each person's social
environment, so that each person is brought together with more and more
compatible people over time, and can therefore spend less and less time
with incompatible people (or working at unsuitable jobs).
I am not just talking about potential "dates", I mean people of all ages,
potential friends, potential co-workers, and so on for all possible
types of social relationship.
Here is a little exercise for the reader: Take a piece of paper and
write down all the people you spend time with on any regular basis,
family, friends, rivals, supervisors, co-workers, and so on. Write down
how many minutes or hours you might spend with them in a typical week
or month. And beside each number, write a quick guess at how compatible
you are with that person -- on a scale from 1 to 10.
In the ideal world I dream about anyone you spend more than a few minutes
with should get a high number, or at least a passing grade. I can remember
a few glorious months when that seemed almost true for me. It wasn't
real or didn't last. "Six billion people in this world, and I have to
spend every day working with him??!!" "Billions of women in this world,
and I end up in bed with her!!??" Oh, the nightmares!
Elsewhere in these pages you will find some discussion of a simulation of
the world economy. That's also social technology, and yes I am working
on such a simulation, but for me it is all part of the same problem.
To properly simulate the economy, you need to simulate human society,
and to simulate human society you need to be able to make predictions
about human social behaviour. In fact we can predict human behaviour
quite well already, on average -- on a statistical basis. We can make
decent predictions about groups of people. It's just individual people
who defy prediction, partly because we all want to defy prediction.
I can't change that, and certainly don't want to change it. Who would
want to live around completely predictable people? But that's OK.
Statistical predictability is just fine. I never want "society" to
tell Dick he must marry Jane, but I'd be quite happy to see society
making a series of suggestions that are pretty good on average, especially
if none of these suggestions involved making a date with some stranger.
People should meet in groups. Indeed that is how people do meet, most
of the time. The best relationships don't start with high-pressure
dates, and I don't think high-pressure job interviews are a good mechanism
either. Over time people should meet many highly compatible people,
in a social setting -- in groups, including many potential spouses and
many potential employers.
Yes, people meet in groups all the time today, and they do meet many
potential spouses and perhaps potential employers. But they don't
meet many compatible ones or appropriate ones. That's what has to
change.
If truth be told, our society is filled with social technology, tools
and techniques for a social purpose, but it is filled with very
"low-tech" social technology. The typical "computer dating service"
of today is paleolithic social technology. Even the high-priced
headhunters who find expert employees for high-tech companies are
offering something from the old stone age.
You might ask what right I have to say this. Good question! Let me
turn the question around and ask what right they have to claim their
services work -- have they proof? How do they measure it?
I'm not talking about something to be taken on faith, I am talking
about something that can be tested and measured. The best way to
illustrate this is by considering employees hired to make decisions
with immediate financial consequences. Suppose, for example, the
employees need to make guesses about the future prices of stocks,
bonds, or commodities.
Some people seem to be very good at this and make a lot of money.
Others loose money, and also their jobs. But rather than thinking
about individual talents, what about pairs or small groups of people
working together. Two heads are better than one, aren't they?
Apparently not. Decisions made by committees are often worse than
those made by any one person working alone. I am sure you recognize
that decisions made by a committee very often poor ones, but this is
not necessarily so -- in theory individual errors could cancel out,
making for better decisions. Many attempts have been made to
change things so that people working together could produce better
decisions as a result of their cooperation. The most famous of these
attempts was the RAND Corporation's work on the Delphi Method, which
made careful use of written submissions and anonymity to try to force
objective evaluation of each other's ideas.
But the Delphi Method ignores error-covariance. If you put together a
committee of people who make similar mistakes, you will get a committee
more likely to make those mistakes than any individual on it.
I've written a lot about this but rather than boring you with my
arguments I'll just state it as a requirement. To count as
high-technology, any method for matching people with co-workers
must meet the committee-test: working together the people must make
better decisions than any of the the people would working alone.
That something that can easily be measured. In the example above it
can be measured in terms of dollars. A committee trading stocks,
bonds, or commodities should not only make more money than any
individual working alone, it should make more money than the total
for all of them working individually.
To put it in a context I'm more familiar with, Brook's Law should
fail -- adding more people to a late software development should NOT
make it later!
Is this possible? I claim it is. It is quite easy to measure
people's error-tendencies: just ask them to make a lot of guesses or
estimates and compare them with reality. We know that people have
quite consistent and long-lasting error patterns. The simplest case
is simply to match people in pairs, and for that we have adequate
bipartite matching algorithms.
Let me try that again with less tech-speak. Get "the computer" to
match up two people who meet this simple criterion: when they agree
they are right, when they disagree they are BOTH wrong. Like this:
Allan's Guess Bill's Guess Cecily's Guess Reality
-----------------------------------------------------------------------------
Acme $1.00 $2.00 $1.25 $1.50
Better $2.00 $2.00 $2.00 $2.00
Central $1.00 $1.00 $1.00 $1.00
Delta $2.50 $1.50 $1.75 $2.00
If you had to hire one person, it would be Cecily, whose guesses are
better than either Allan or Bill. But she didn't guess right all the
time, and there is a better answer. Think about hiring both Allan and
Bill, because when they agree they are right, and when they disagree
the one is high and the other is low, so the average of their guesses
is bang on.
That's the whole thing in a nutshell. Of course it doesn't have to be
right and wrong, doesn't have to involve decisions, and certainly
doesn't have involve money or any business situation. But whether it
is love or money, there's a powerful idea here.
An old idea, of course, and not mine -- it goes back to Shannon and
Weaver, and the synthesis of reliable channels from redundant
unreliable ones. Using combinatorial optimization to do the matching
is a little more original in the sense that I clearly remember
thinking it out for myself -- but again it is probably an old idea.
A lot of good old ideas are out there just waiting to be applied.
I'm not looking for credit for figuring all this out, I'm just looking
ahead to that better society, the one with love and money for all of
us. A lot of people are broke and lonely, but none of us need be.
In the wrong social context almost any one of us could end up hungry,
homeless and without a friend in the world. And in the right social
context even the saddest specimens of humanity could flourish and
contribute something of worth. Am I mad to think these thoughts?
(So sorry -- it seems that I have literally a little alliteration
habit that is quite a handicap for one wanting to write well).
---------------
... much more yet to be written, rewritten, and rewritten again,
sigh ...
dpw
