[ExI] Regarding Wickedness (was beowolf)

[David Lubkin]

I wrote:

 > In the best of all possible worlds, I would have an intelligent agent
 > that mined the list and the outer web, built a predictive model of
 > each list member, and tuned my posting so that each reader saw a
 > personal version that optimized the likelihood that they'd read and
 > reply to to it, much as Amazon does.

Bryan replied:

>What sort of data would we have to mine in order to make this possible?
>I suspect that the preliminary data would be what eacher reader tends
>to "read then quit", what ve replies to, etc. Maybe brain scanning, but
>I think we can do better than that with less equipment and certainly
>less algorithms.

There's a lot of digital spoor out there about each of us. I bet 
that, with a hefty budget, one could acquire enough of it from the 
different spheres of our lives to pull together a frighteningly 
accurate predictive model of human behavior.

Even just looking at me, and just my activity on-line in mailing 
lists and newsgroups, I've posted at least 30,000 messages (assuming 
3 a day for 27 years; sanity-checked with a glance at my Out 
mailbox). I'm sure there are tell-tale patterns.

For the limited question of subject lines, it would be easy to 
measure and compare for each of us separately, for the list overall, 
and for lists throughout the net the reply rates achieved by a 
posting in a new thread versus a posting in a pre-existing thread.

Of course, that wouldn't distinguish between latter postings that 
were on-topic and those that were not.

This feels like something someone is or ought to be studying, perhaps 
as part of computational anthropology or the linguistics subfield of 
discourse analysis, alongside the studies examining how rude people 
are in their on-line interactions as compared to in-person conversation.

With a billion people on-line, it might be useful to know something 
of behavior patterns.


-- David.