From: Mark Day (markday@cisco.com)
Date: Fri Jun 02 2000 - 14:21:18 PDT
> Most email programs today have ways of creating rules or
> filters that tell the program what to do when certain kinds
> of messages are received. This is a useful function, but
> it's only the beginning of what's possible with a technology
> such as Priorities.
There is a little bit of a problem here, of course, because most people
don't find the existing rule systems easy enough or trustworthy enough to
use. It's not clear to me that a more sophisticated/subtle system answers
any real need.
> Priorities is designed to practice "courteous computing,"
> the ability to consider certain criteria before bothering a
> recipient with an urgent message. Based on parameters set by
> the user and the program's ability to "learn" based on a
> user's habits, Priorities weighs the urgency of each message
> sent against the probability that a user will be paying
> attention to a noise from a PC, phone, or pager.
The details of how it "learns" are critical: that's usually the place that
systems like this fail. The article doesn't really explain how the system
manages to improve its performance. The obvious approaches (like, have the
user tell you whether you did the right thing) drive people crazy, because
of course the system is grabbing more of their attention now in the name of
eventually grabbing less of their attention, which is not a tradeoff that
most people are willing to make.
> Without getting into the mind-bending mathematics involved,
> here's roughly how Priorities is supposed to work.
Uh-oh. Any time someone tries the "you can't possibly understand this
system, it's so mathematically complex," my bullshit detector goes off. I'm
sure others can supply the names of other well-known companies that take
this angle.
> I'm told the version of Priorities working at Microsoft even
> has an understanding of Microsoft's corporate pecking order,
> so you can bet that messages from Mr. Gates and Mr. Ballmer
> will be assigned top priority.
Note the conceptual inflation here: "an understanding of Microsoft's
corporate pecking order" when I'd bet that the actual implementation
involves (at best) a single numerical factor determined by distance from
Ballmer and/or Gates in the reporting hierarchy.
> For me, living in a world where I'm blasted by more than a
> hundred email messages a day from publicists and
> junk-mailers, this type of application would be a godsend.
Only if it worked exactly the way you wanted, and the trouble with subtle
systems is that they fail in subtle ways. Sometimes they alert you and you
can't figure out why the hell they did that. Sometimes they fail to alert
you and you can't figure out why.
We did some fun stuff with sending email to pagers a couple of years ago at
Lotus. Even the rudimentary decision mechanism we had felt like overkill for
most users (we had a couple of people go hog wild with rules, but they were
clearly exceptions). I'll dig out a citation if anyone cares.
> This kind of innovation is intended to make life easier on
> consumers and tougher on competitors. Contrary to popular
> belief, the Internet age may be the time consumers are more
> dependent on Windows (and its related services) than they've
> ever been before.
Give me a break. How much did Microsoft have to pay for this blurb? See my
comments above about what we were doing years ago. And Lotus already
announced products descended from our research, not this pathetic "maybe
we'll ship it, maybe we won't."
--Mark
Mark Day
Senior Scientist
Cisco Systems, Waltham, MA
+1 (781) 663-8310
This archive was generated by hypermail 2b29 : Fri Jun 02 2000 - 14:14:01 PDT