Information Technology Reference
In-Depth Information
5.3 Evaluations
Let us assume there is a good understanding of what 'evaluations' are (Chapter 2, in particular
Section 2.2.7), and look in more detail now at the relationships between evaluations, goals,
and emotions.
Evaluations imply goals by definition, in that the latter are a necessary component of evalua-
tions, namely, the second argument of the GOOD-FOR predicate. From a more 'substantialist'
perspective, evaluations imply goals in the sense that they
originate
from them: it is the
existence of some goal
g
(either
X
's or someone else's) that makes the words good or bad,
justifies and motivates both the search for a means
m
to achieve it, and the belief that
m
is (not)
GOOD-FOR
g
. Goals and evaluations endow objects and people with 'qualities' and 'faults'.
The relationship between evaluations and goals is even closer, because
evaluations not only
implies goals, but can also generate them
.Infact,if
X
believes
m
is good for some goal, and
X
has that goal,
X
is also likely to want (possess, use)
m
. So there is a rule of 'goal generation'
which might be expressed as follows: if
X
believes something
m
to be a means for
X
's goal
g
,
X
comes to have the goal of exploiting/using the means
m
.
Evaluations, that is, knowledge about 'what is good for what', and 'why', play a crucial role
in all the cognitive activities that are based upon symbolic and explicit representations, rea-
soning and deliberation. For example, in problem solving and decision making, the particular
advantage offered by evaluative knowledge is precisely a preliminary relationship established
between descriptive knowledge and goals, in terms of beliefs about 'what is good for what',
derived from either one's experience about problems solved in the past, or one's reasoning and
inferences (think for instance of evaluation by standards), or others' communication.
Evaluations make such a relationship explicit
;
they fill the gap between knowledge and goals,
by 'reinterpreting' the properties, qualities, and characteristics of objects and situations in
terms of means for the system's (potential or actual) goals
.
The cognitive network ceases to be neutral and becomes 'polarized' toward goals, that is
ready for problem solving and decision-making.
In a cognitive agent preferences can be internally represented both at the
procedural
and at
the
declarative
(propositional) level.
Having a
procedural preference
means that, at a given level of their processing, a system's
goals present different degrees or indexes of activation, priority, weight, value, importance
(or whatever), that in fact create some rank order among them, which will be followed by
some choice/selection procedure.
Having a
declarative preference
means that the system is endowed with an explicit belief
such as: '
m
is better than
n
(for goal
g
)'. In particular, three types of beliefs are relevant
for preferences: (a) simple evaluations, that is beliefs about how good/useful/apt/powerful
are certain entities relative to a given goal ('
m
is very useful for
g
'; '
n
is quite insufficient
for
g
'); (b) comparative evaluations like '
m
is better than
n
for
g
'; (c) reflective preference
statements, of the kind 'I prefer
m
to
n
(for
g
)'. Generally,
(b)
are based on
(a)
; while
(c)
are
based on
(b)
.
Both procedural and declarative preferences can coexist in a human mind (and would be of
some use in artificial minds too), and each level of preference representation - though having
its own mechanisms of reasoning -
is translatable into the other
. One can derive a 'weight'
Search WWH ::
Custom Search