Mystery Number Five - Computational Linguistics and Talking Robots - page 100

6.3.6 C ONSEQUENCE INFERENCE FOR NEGATIVE EXPERIENCE ( CIN )

β
Like all DBS inference rules, CIN derives new content (consequent) from
given content (antecedent), here using the connective csq (for consequence).
Unlike the inference rules 5.2.3, 5.2.5, 5.3.3, and 5.3.5, the consequent of 6.3.5
is not connected to the antecedent by one or more addresses. This is because
the consequent may apply to an instantiation of

rec: α act: β rec: bad csq rec: α act: no

different from that in the
antecedent - though identity at the content level is not precluded.
In a second encounter with a red circle, the antecedent of inference 6.3.6
matches the negative experience 2 in 6.3.5 with a previous unknown. Based
on the alternative blueprint for action provided by the consequent of the CIN
inference, this second encounter will not result in hiding, thus preventing an
unnecessary interruption of the agent's feeding activity. Applying CIN to the
negative experience 4 of 6.3.5 has the opposite effect, causing the agent to hide
on next encountering a red circle.
The positive experiences 1 and 3 of 6.3.5, in contrast, reenforce the action
initially chosen at random, based on the following inference:

α

6.3.7 C ONSEQUENCE INFERENCE FOR POSITIVE EXPERIENCE ( CIP )

rec: α act: β rec: good csq rec: α act: β
In this way, adaptive behavior derives new action patterns in the agent's
writable memory which serve to maintain the agent's balance. 18

6.4 Upscaling from Coordination to Functor-Argument

The combination of old elementary “knowns” into a new complex “unknown”
(cf. 3 in 6.3.1), such as red circle in 6.3.5, goes beyond the method of using
rec and act as the only core attributes of proplets, practiced for simplicity
in the definitions of LA-act1 (6.1.6) and LA-rec1 (6.2.5). For example, given
that red circle is a functor-argument, with the adjective red modifying the
noun circle , we need to refine the LA-act1 and LA-rec1 grammars to handle
functor-argument in addition to coordination.
This refinement resembles the transition from propositional calculus to first-
order predicate calculus in Symbolic Logic. Symbolic Logic constructs the
transition by building formulas of predicate calculus from formulas of propo-
sitional calculus, using the additional constructs of functors, arguments, vari-
ables, and quantifiers. Consider the following example: