[ExI] System design for managing too many friends

Bryan Bishop kanzure at gmail.com
Sun May 31 21:53:02 UTC 2009


Hey all,

I am glad to have so many friends. And it's nice to have a reasonable
percentage of them check up on me when I don't say hello in a while.
Because of the way that my brain works, or because I am forgetful or
possibly not that good of a guy, I am not able to clearly remember
what I have told some people versus what I have told some other
people. The way that most people solve this is by using a blog and
just blasting everyone who follows it with the same information, and
an RSS feed might work for me, but I have another idea in mind.

In particular, I have been working on constructing a giant hash table
(or it might become a giant matrix/array) where each person I know and
try to keep up with is listed in the first column. The remaining
columns are used for particular events that I at least tell one person
about. Then by looking at the distribution of what I tell each person,
or when I tell them or whatever, I could then extrapolate who I should
commonly send the same updates to.

http://heybryan.org/mailing_lists.html

That's actually what I was thinking of doing for some bayesian
filtering system (much like spam filters) for the email that I send
out to mailing lists. Since I am on so many mailing lists, it's hard
to figure out what I should send to which mailing lists. For instance,
the amateur genetic engineering emails that I send out to diybio
should sometimes belong on some of the transhumanism mailing lists,
but not all the time because the perspective is slightly off and I'm
writing from a different point of view, or the project updates simply
aren't going to be as interesting, blah blah blah. So, the idea I had
a few months ago was that I would write a spam filter that would
instead of classify my outgoing mail as junk, would simply classify
the responses to each of the emails I send, based off of the MIME
headers and whether or not anyone responds to them, their
positive/negative responses (cross-referenced to WordNet or just using
a simple rating scheme). Then, over time, I would (ideally) develop a
good idea of what type of messages should be sent to which mailing
lists, using some sort of fuzzy probabilistic logic. I haven't written
out this software yet though, mostly because I'm not convinced of the
utility. Although I do hate keeping everyone in the dark as to what
I've been up to.

Back to the people-update matrix/table database dealy. I was also
thinking of doing some sort of monthly cron job schedule where I would
type down into a list form what I remember last talking about with
that person, and then an automated email could be sent out on some
periodic basis to get back in touch with everyone I know. This would
work out reasonably well, since I'm supposed to be sending out these
emails anyway. Should be a simple script. Time well tell if it will
pay off.

Ok. Hope someone got something interesting out of this. I'd be
interested in getting advice on how to scale communication for someone
like me who has too much going on to remember at any one given moment.

- Bryan
http://heybryan.org/
1 512 203 0507



More information about the extropy-chat mailing list