Mystery Number Four - Computational Linguistics and Talking Robots - page 72

This may be done highly selectively, at various levels of abstraction and in var-
ious domains. The resetting may also be done centrally by changing the cog-
nition's center of gravity, thereby changing the agent's individual preferences
and dislikes in one fell swoop, from nice to not-so-nice, from squeamish to
matter-of-fact, from generous to stingy, etc., whatever has been implemented.

5.2 DBS Inferences for Maintaining Balance

Inferences may generally be described as deduction rules which derive new
content from given content in a meaningful way. This holds for the deductive
inferences in Symbolic Logic, which derive true conclusions (new content)
from premises (given content). It also holds for DBS.
To maintain balance, DBS inferences take current and stored content, includ-
ing the appraisal of previous action outcomes, as input, and compute blueprints
for new action as output. There are three kinds, called R(eactor), D(eductor),
and E(ffector) inferences. 6
The condition 5.2.4 that the consequent of inference n must equal the an-
tecedent of inference n+1 allows easy self-organization and highly adaptive
data processing. The absence of rule names and rule packages facilitates defi-
nition and automation.
For readability, the following example of an inference chain is simplified
as follows: (i) English words are used to represent proplets, (ii) coreference-
by-address notation is omitted, and (iii) easily programmed details regarding
the iteration of values in the variable restriction (line 3) are not included. The
consequent of inference n always equals the antecedent of inference n+1 .

5.2.1 C HAINING R, D, AND E INFERENCES

1. R: β be hungry K cm β eat food K+1

2. D: β eat food K+1 pre β get food K+2

3. D: β get food K+2 down β get α , K+3 where α {apple, pear, salad, steak}

4. E: β get α K+3 exec β locate α at γ K+4

5. E: β locate α at γ K+4 exec β take α K+5

6. E: β take α K+5 exec β eat α K+6

eat food K+7
Each line begins with the step number, e.g., 1 , followed by the kind of infer-
ence, e.g., R (eactor), the antecedent, e.g.,

7. D: β eat α K+6 up β

be hungry ,the prn value of the
antecedent, e.g., K , the connectives, e.g., cm , the consequent, e.g.,

β

β eat food ,

6 The terminology is intended to distinguish DBS inferences from the inferences of Symbolic Logic.
For example, while a deductive inference like modus ponens is based on form, the reactor, deductor,
and effector inferences of DBS may take content, domain, level of abstraction, etc., into account.