Modeling Short and Long-Term Cognitive Components
of Agents’ Affective States
Fiorella de Rosis
Intelligent Interfaces,
Department of Informatics, University of Bari
(hints to the discussion)
Although, in the last few years, we’ve seen a
flourishing of initiatives and proposals towards
endowing machines with the ability to recognise and to
manifest a personality, an emotional state, a mood,
research efforts in this domain focused on the external
expression of affect rather than on its cognitive
background. However, reacting appropriately to the
Interlocutor’s (human or artificial agent) manifestation
of an affective state may require reasoning on the
causes of this state; this implies reasoning on the event,
the action or the object (to employ Ortony, Clore and
Collin’s model) which might have produced the
perceived state and on the mechanism through which
the decision to manifest (or to partially hide) this state
was made, by the Interlocutor.
 has the difference between representation of
personality, mood (or attitude) and emotion something
to do with long and short-term user modeling? if so,
how should short and long-term features be integrated?
how might distinct affective aspects be assembled,
within an Agent model, into a cognitively plausible
combination?
To this aim, especially when the interaction context is
complex, some insight on the cognitive background of
the affective state of the Interlocutor is needed. A
typical example: a ‘natural’ dialog, in which the
interpretation of a negative (or positive) reaction of the
Interlocutor to a Speaker’s statement is not immediate
or unique; in this case, the Speaker needs to reason on
the Interlocutor’s mind, to try to interpret the (verbal
and/or non-verbal) communication received and to
respond appropriately rather than by generically
manifesting some ‘empathy’. Another notable example
is the decision of when an Agent should or shouldn’t
manifest an emotion, according to its ‘intensity’, to the
Agent’s personality, to the context where interaction
takes place and so on.
My proposal is to verify the interest and the ideas, about
these topics, of the participants to the Workshop (those
who presented a paper and those who more recently
began to work in the field).
People in the User Modeling and Adaptation area may
find excellent hints, to their work in this domain, in the
theories
proposed
by
well-known
cognitive
psychologists, some of them dating back to several
years ago: the theory of personality and its relationship
with goals by Carbonell; the theory of ‘subjective
importance’ of emotions by Ortony; the emotion model
architecture proposed by Frjida; the theory of cognitive
evaluation of social emotions built by Castelfranchi,
and many more.
What UM people might do, in my opinion, is to
reconsider these theories and proposals in the light of
the methodologies that have been proposed and applied
successfully in User and context-adapted systems, in the
last ten years. Some problems, though, have to be
reconsidered. For instance:
 which are the advantages of the various methods
that might be employed to represent the cognitive
aspects of the affective state and their relationships? For
instance: are belief networks -static or dynamic-,
inference networks or neural networks suited to
represent uncertainty in the arousal of emotions and
their intensity?
A few references
J. G. Carbonell: Towards a process model of human
personality traits. Artificial Intelligence, 1980.
C. Castelfranchi: Affective appraisal versus cognitive
evaluation in social emotions and interactions. Affective
Interactions. A Paiva (Ed), Springer, 2000.
C. Elliott and G. Siegle: Variables influencing the
intensity of simulated affective states. Proceedings of
the AAAI Spring Symposium on Mental States ’93.
1993.
N.H.Frjida and J Swagerman: Can computers feel?
Theory and design of an emotional system. Cognition
and Emotion, 1987
A.Ortony: Subjective importance and computational
models of emotions. In V Hamilton, G H Bower and N
H Frjida (Eds): Cognitive perspectives on emotion and
motivation. Kluwer, 1988.
A. Ortony, G.L. Clore and A. Collins: The cognitive
structure of emotions. Cambridge University Press,
1988.
A. Ortony: On making believable emotional agents
believable. To appear in R Trappl and P Petta (Eds):
Emotions in humans and artifacts. MIT Press.
R Pfeifer: Artificial Intelligence models of emotion. In
V Hamilton, G H Bower and N H Frjida (Eds):
Cognitive perspectives on emotion and motivation.
Kluwer, 1988.