Information Technology Reference
In-Depth Information
2.2 Emotions
In order to represent emotional states, a main assumption made is that the intensity of
different emotions can be represented via a number between 0 and 1, as often done
within emotion models within Artificial Intelligence (e.g., [5]). In the current paper,
only one type of emotion is modeled, namely happiness (with 0 = very sad and 1 =
very happy), but the model can also be used to represent emotions like anger and fear.
As shown in the upper part of Figure 1, the component to represent emotions
interacts with the basic BDI-model in various manners. First, for generation of
emotions, it is assumed that they are the result of the evaluation of particular events
against the individual's own goals (or desires), as described by appraisal theory [5].
This process is formalized, among others, by the following LEADSTO rules (stating
that reaching a goal state S leads to happiness with an intensity that is proportional to
the strength of the desire, and inversely for not reaching a goal state):
S:STATE
I:REAL
desire(S, I)
belief(S)
happiness(S, I)
desire(S, I)
happiness(S, 1-I)
For example, in case an agent desires to score a goal, but misses, then the agent
becomes sad about this. In addition to these event-specific emotional states, the
agents' long term mood is modeled. Moods are usually distinguished from emotions
in that they last longer and are less triggered by particular stimuli. Thus, in the model
a global mood is assumed with a value between 0 and 1 (representing its positive
valence), which depends on the values of all specific emotions in the following way:
belief(not(S))
I1, ..., In, J:REAL
happiness(S1, I1) ... happiness(Sn, In) mood(J)
S1, ..., Sn:STATE
mood(β * J + (1-β) * (w1*I1 + ... + wn*In))
Thus, the new mood state is calculated based on the valence J of the old mood state,
combined with the weighted sum of the emotional values Ik for the different aspects
Sk in the domain of application. Here, the wk 's (which sum up to 1) are weight factors
representing the relative importance of the different aspects, and the β (between 0 and
1) is a persistence factor representing the rate at which mood changes over time.
Finally, various rules have been formulated to represent the impact of emotions
and mood on beliefs, desires, and intentions. These rules are mostly domain-specific;
for instance, a positive mood increases the desire to cooperate with teammates, and a
negative mood increases the desire to behave aggressively. These rules are described
in detail in [6].
2.3 Trust
According to [4], trust is a (dynamic) mental state representing some kind of
expectation that an agent may have with respect to the behavior of another agent or
entity. The authors propose that a belief in competence of the other entity is an
important contributor to the development of trust (in addition to a second aspect, the
belief in willingness; however this aspect is currently ignored). This mechanism is
formalized in the current paper by assuming that trust in another agent is based on a
weighted sum of the beliefs in the agent's capabilities (using a similar formula as
above for mood update):
Search WWH ::




Custom Search