[FOM] paraconsistent logic and computer science

[Andrej Bauer]

I believe the following example from the real world is relevant to this
discussion, and also relevant to the original post.

Cycorp [http://www.cyc.com/] is a long-term AI research project that
turned into a commercial company at some point. One of its main goals is
development of what one might call automated common sense reasoning. To
see how they use logic and knowledge databases to "give machines some
common sense", have a look at the talk at
http://video.google.com/videoplay?docid=-7704388615049492068

Because they are trying to capture human common sense their database is
huge. Reasoning based on proof search in a deductive system becomes
useless in such a situation because it is too slow and because it leads
to very odd conclusions. These arise when one starts with common sense
"facts" and applies four or five steps of deductive reasoning to them.
It is nearly impossible, and even undesirable, to make the knowledge
database consistent. Common sense knowledge just doesn't mix well with
logic, as the readers of this list undoubtely know from their personal
experiences.

For these reasons Cycorp is favoring probabilistc reasoning methods.
"Exact" reasoning, whether classical, intuitionistic, or paraconsistent,
is just not suitable for their purposes.

Andrej