"J. R. Molloy" wrote:
> Have you noticed the growth rate of litigation in the US?
> It may out distance technological innovation, smothering it in the process.
> Then again, the law may simply appropriate AI and make the phase transition a
> matter of jurisdiction. How about lawyer-bots running everything?
> No, wait! That may have happened already.
Not quite yet.
I currently work for a subsidiary of West Legal Group, the megacorp that
owns the copyright on the reference system most US lawyers use to refer
to US laws and cases. Or, in short: they think they own the law in the
US, and they're arguably right from a certain point of view.
Anyway, I was discussing the idea of lawyer-bots with my group's CTO,
and how we might be able to develop them. The closest thing he had
heard of was a limited use script, used by some Brazilian judges to
quickly decide standard traffic cases (accidents, et cetera). I
proposed, as a first generation, a lawyer's aide that could predict,
better than lawyers themselves with the tools they currently have, what
the likely outcome of a case would be, based on the facts available and
the evidence each side was likely to be able to obtain. (I had to
justify this in monetary terms somehow, and that seemed an easy path
from tech to money. Besides, it was an extension of the Brazilian
script he already knew was working.)
He seemed to like the idea and believed the core tech was feasable, but
soured on the interface: if it's going to be the hundreds or thousands
of pull-down menus and similar widgets needed to properly classify a
case, forget it. I considered free-text entry, say like an advanced
chatbot, but figured that natural language processing probably is not
advanced enough for use by lawyers (except those lawyers who are also
geeks, but that dramatically narrows the target market).
I figured this might be a good way to pump money into practical AI
development, not to mention start automating away a sector that already
has far too many people working in it - changing things to a legal code,
as programmers understand the word "code", and hopefully reducing the
costs of being sued for something one did not commit (which, in and of
itself, seems to be one of the biggest factors stifling research and
experimentation, at least in the US). But unless I can get past the
interface problem (different interface? different product?), no dice in
this approach.
Does anyone here has any practical suggestions on how to solve this
problem?
This archive was generated by hypermail 2b30 : Sat May 11 2002 - 17:44:32 MDT