artificial intelligence

[Terry Reedy]

"Arthur" <ajsiegel at optonline.net> wrote in message
news:mailman.1062394443.2068.python-list at python.org...
> OTOH, the concept of "coincidence" must necessarily be ruled out in
AI, I
> would think.

Statisticians generally assume the null hypothesis of 'coincidence'
until evidence shows elsewise.  However, applied statisticians must
take into account the benefit/cost of being right/wrong when there
is/isn't an association.  I suspect Amazon concluded that the
rate*cost (customer unhappiness) of reporting false associations is
low compared to the benefit (stimulated sales) of true associations
and even perhaps some that were false.  Are you going to boycott
Amazon because they gave you a laugh?  I am sure that Amazon is
counting on most people not being really offended.  On the otherhand,
reporting and stimulating pursuit of non-existent associations if more
costly in science.

Terry J. Reedy