Toward Human-Level Cognitive Adequacy
Our long-range aim is to design and implement common sense in a
computer. Click here
for details.
If you would like to learn more about Active Logic, we suggest
you start with one of our primers.
Conversational Adequacy
One major ongoing application of active logics is that of building a “conversationally
adequate” dialogue agent (see paper on
Conversational Adequacy—
one can view such an agent as having the ability to learn
in McCarthy’s sense of advice-taking, via conversation).
We have accomplished the following goals (see research
report):
- process a few simple queries via voice input, requiring the use of
negative introspection, in the context of the Rochester TRAINS system
rewritten to employ active logic.
- handle apparent contradictions as in “Send the Boston train to NYC...
[here the system does something]... No, send the Boston train to NYC”
where the system has misunderstood the first request and so the user
says “no” and repeats the same request. This requires the system to
recognize that “no” refers to the systems response rather than to the
user’s first request.
The long-range research plan envisions the following steps:
- understand and generate spelling in metadialogue (“The word I am
saying is spelled ‘xyz’...”)
- definitions and/or definitional queries (“What does the word ‘xyz’
mean?”)
- more general metadialogue: You said, I said, you meant, I meant,
etc
- reasoning about the physical and social worlds
- focus, attention and memory
- conversational adequacy