Mystery Number Four - Computational Linguistics and Talking Robots - page 71

reconstruction of intention in a cognitive agent. It plainly follows that “meant
something” in the definiendum is unsuitable for computation as well. 3
DBS treats the notion of intention computationally as the system-inherent
(innate) drive to maintain the agent in a state of balance. 4 The principle of

balance, in turn, is realized by means of inferences .
An inference consists of (i) an antecedent schema, (ii) a connective, and (iii)
a consequent schema. Examples of connectives are cm (countermeasure), impl
(implies), or exec (execute). An inference is triggered by a content matching
its antecedent. Matching and binding the antecedent variables enable the con-
sequent to derive an appropriately adapted blueprint for action (5.2.3).
The agent's balance may be disturbed by a nonlanguage perception, e.g., a
sensation of hunger, or a language content, e.g., a reproach or a demand (hear
mode, recognition), just as the countermeasure for regaining the agent's state
of balance may be realized as a nonlanguage action, e.g., getting something to
eat, or a language content, e.g., an apology or a request (speak mode, action). 5
Because all behavior, including language behavior, is managed by the agent's
autonomous control, it is essential for language production in communication:

5.1.2 T HE F OURTH M ECHANISM OF C OMMUNICATION (M O C-4)

The language as well as the nonlanguage behavior of a cognitive agent
is driven by the goal of autonomous control to maintain a continuous
state of balance vis à vis constantly changing external and internal envi-
ronments. The success of autonomous control, short-, mid-, and long-
term, is defined in terms of survival in the agent's ecological niche.

The niche includes the agent's social group. Within this group, maintaining a
state of balance via language production ranges from ritualized customs like
greeting with small talk to an exchange of opinion to a cry for help.
The overall goal of autonomous control to maintain the agent in a state of
balance is a general method with many applications. For example, the method
can be used not only to try to ensure the agent's survival in its ecological niche,
but also for controlling the artificial agent's behavior for purposes of the user.
The artificial behavior may be reprogrammed by changing the “desired” val-
ues (defaults, Sollwerte ) in the parameters of the agent's appraisal software.

3 For further discussion of Grice's notion of meaning, see FoCL'99, Chap. 4, Example II.
4 The cooperative behavior of social animals, e.g., ants in a colony, may also be described in terms of
balance. Such a decentralized approach to behavior is in line with Brooks (1985).
5 See Chap. 10 for an example which includes the use of balance to drive a dialogue.