The role of emotions in organizational join decisions:

A formal modeling framework1

Antonio Aguilera-Ontiveros and Julio Cesar Contreras-Manrique.
El Colegio de San Luis, A.C.
aaguilera@colsan.edu.mx; ccontreras@colsan.edu.mx

1. Introduction

The study of emotions and emotional behavior of groups in organizational contexts has been
gaining relevance in the last two decades into the organizational studies literature (Fineman,
1996; Weiss & Cropanzano, 1996). The importance of studying emotion in workgroups
settings derive from the evidence that members’ behavior in an organization varies according
to their emotional condition. Indeed, as Weiss and Cropanzano (1996) argue, workplace
experiences comprise a succession of work events that can be either pleasing and suitable, or
stressful and frustrating. These events affect the way in which a member feels and behaves at
work. These have as consequence functional or dysfunctional effects in the specific roll
within her/him workgroup. There are also evidence that emotional traits, such as negative
affectivity (Watson & Clark, 1984), are mechanisms that affect behavior and decision making
in the workplace context, e.g. emotional intelligence (Salovey & Mayer, 1990) and emotional
labor (Fineman, 1996:554).

Cases studies have been using as the traditional way to study the effect of emotions in
organizational behavior (Fineman, 1996). However, in recent years, emotional modeling has
become a new tool to study the roll of emotions in human collective life (Trappl et al., 2002).
This kind of studies lies into a whole new branch of organizational research known as
computer organization theory (COT) (Prietula et al, 1998). COT sees organizations as
complex and distributed computational entities in interaction (Prietula, et al. 1998; Aguilera
and López, 2001; Lomi and Larsen, 2001).

In this communication, we present a formal modeling framework to study the effect of

emotions in joint decision-making processes in workgroups. The decision-making process in
our model includes two emotional mechanisms (Ortony, et al. 1988; Howard, 1993).

In the following section we explain the relation between emotions and the decision-making
process introducing the Howard’s and Ortony’s ideas. Next, we define modal and temporal
logic formalism for the model. In our conclusions we argue that this mathematical formalism
can be useful to build a computational model.

2. Emotions and decision-making processes

Traditionally, emotions have been conceptualized as a contradiction to rational behavior. In

this context, emotions are seen as dysfunctional factors into the process of decision making,
which is inherently rational (Howard, 1993:121). Nevertheless, in the last years, several
disciplines as sociology and economy have increased their attention in the influence of the
emotions in the rational behavior (Elster, 1996; Loewenstein, 2000). Also, there are a great

1
In Proceedings of The First World Congress on Social Simulation. August 21-25, 2006, Kyoto University,
Kyoto, Japan. Vol. 1. Pp, 351-358.
evidence of historical and cultural variability of the meaning, expression and regulation of the
emotions (Thoits, 1989:319; Fineman, 1996:550). This historical and cultural variability
suggests an important degree, that subjective experiences and emotional beliefs are socially
acquired and structured (Thoits, 1989:319). Actually, emotions are conceptualized as a
prerequisite for rational behavior (Trappl & Payr, 2002:1) and as a factor that chance the
rationality (Fineman, 1996).

There are four basic items of a rational decision-making process: a) desires; b) beliefs; c)
information, and d) action (Elster, 1996:1391). Elster argues that there are at least seven
different ways in which emotions can fit into the decision making process (Elster, 1996:1391)
making it so complex to model it..

2.1. The Howard’s model

To resolve the complex effects of emotions into decision-making process first we must
explain the Howard’s model. Howard argues that a decision-making process can be seen as a
theoretic soft game, where the basic elements of the game are actors, options and scenarios.
Actors have a set of options which it may or may not decide to implement. For each actor a
future decision is a particular scenario (Howard, 1993: 617). The relevance of the model is
the interaction between actors before the options are taken. In this interaction actors try to
influence other’s behavior. In Howard’s model joint decision could be conceptualized as the
coincidence of future scenarios between the participants of the game, i.e. each actor in the
game has a specific output from the desire-beliefs-information-action system which match
with a mutual scenario. Actors propose scenarios to joint implementation of an action.
Emotions play the role of significance schemata which make true or false a message (either
threat or promise) exposed by one actor to another. Emotions either increase or inhibit the
beliefs about the proposed scenario. Decision is then taken.

We found useful Howard’s model but it is necessary to make some specifications. Howard’s
model does not specify a correct ontology about emotions. Also, emotions are simplified to be
preferences about one scenario. We think that model must take in account a true emotional to
create preferences.

2.2. The emotional model of Ortony, Clore and Collins

Ortony, Clore, and Collins (1988), and O’Rorke, and Ortony (1994) propose a model for
emotional behavior. This model is based in the cognitive aspects of human behavior and can
be formulated as a computational model. Cognitive aspects of the emotions imply
appreciation, comparison, classification, inference, attribution, and appraisal. Emotions are
viewed as valence reactions towards events, agents and their actions, as well as towards the
objects (Ortony et al., 1988:13).

According to the model of Ortony, Clore, and Collins (1988), and O’Rorke, and Ortony
(1994:289-291), different kinds of emotions are groups or sets of conditions which cause
emotional behaviors and that can be coded in some computational language. The simplest
emotions are the well-being emotions, labeled as joy and distress. These emotions can be
understood like the positive and negative feedbacks that individuals have before desirable and
undesirable happen events. The second group of emotions is referred to the fortune of others.
This group includes four kinds of emotions: happy for, gloating, resentment and sorry-for.
Each kind in this group is a combination of pleasing or displeasing aspects respected to a
specific event, which is presumed to be pleasant or disagreeable for another person. The third
group is the prospected-based emotions, which includes six specific kinds of emotions: hope,
satisfaction, relief, fear, fears-confirmed and disappointment. Each kind represents a reaction
to a desirable or undesired event that has been either confirmed or not. The fourth group is
concerned with attributions and it covers four kinds of emotions: pride, admiration, shame
and reproach. Each attribution emotion kind generates a reaction (positive or negative) to
either any type of own action or actions conducted by others. The fifth group is related to
attraction. This group refers to reactions that individuals have towards objects. The two
emotions in this group are the momentary feelings of liking or disliking an object. The final
group includes eight elements that constitute the result of the Cartesian product "well-
being"×"attributions". This kind of compound emotions does not correspond to the co-
occurrence of their emotional components. Union logical function is used to resolve when a
compound emotion is valid in the computational model (O’Rorke and Ortony, 1994:289-291).

In general, the conditions of "production" of one emotion are given in terms of the variables
that contribute to increase the activity of emotions. Theory specifies global variables that
affect all the emotions and local variables that affect only subgroups of the emotions (Ortony,
et al., 1988:68-81). The variables have values and weights associated to emotions and theory
suggests that one emotion is experienced simply when certain levels of the thresholds of the
emotion are exceeded.

We will use this classification of emotions to build emotional mechanisms for the Howard’s
models. In the next section we explain our model with its formal framework.

3. The formalism

In this section, we present our formal definition of the join decision-making process in
emotionality contexts. Our formalism uses modal and temporal predicate framework (Bell, et
al., 2001) as is proposed by Wooldridge (2000). An agent is an entity able to communicate
and to do something. We take the simple idea of an organization from Cohen, March and
Olsen (1972). The idea in our simple organization is that each agent has a set of tasks that
must do. Each task needs a set of resources. This can be defined as:

Definition 3.1. Let Θ = {α1 ,… , α k } a group of agents. Let W = {w1 ,… , wl } a set of tasks that
must be done by agents. Let R = {r1 ,… rq } a set of resources to do tasks. We define an
“organization” as a couple of equivalence relationships Θ W ⊆ Θ × W ∪ W R ⊆ W × R.

Organization works as follow. Each agent takes from a reservoir resources to do tasks.
Resources are shared between tasks. If one resource is taken by one agent this resource is
unable for other agents in the system. Each task w has a duration time defined as
W T ⊆ W × T where T = {t0 , t1 ,… , tk } is a set of time periods. We are interested in
conflictive situations, i.e. where scare of one resource inhibit the optimum action of the whole
organization, e.g. when an agent with a task wk that has a duration time tk ' unable a resource
critic for a set of tasks.

Agents are able to try of get resources up for their specific tasks using a negotiation
mechanism. This mechanism works as follows. For an specific period of time, if the resource
is not in the reservoir, then agent A request for this resource sending a message to the agent
B, who is using the resource in that period of time. Agent B can decide that he/she does not
need the resource (may be because it is no enough to begin the task) and then send a message
to agent A and yields the resource. Also, agent B can refuse to yield the resource sending a
negative message to agent B. This can be formalized as:

Definition 3.2. Let A = { Ayield (), Adeny ()} the set of actions to a request, Ayield () = yields
resource, Adeny () = denies resource. Let M = {magent ,resource , magent , yield , magent , deny } a set of
messages between agents. We use the following modal logic predicate formula to determine a
def
request, request (β ,rk ) ≡ ∃ mα ,k ∧ ◊∃ have(α , rk ). Where the modal logic formula for have()
def
is determine as have(α , rk ) ≡ ∃ α ∈ Θ ∧ ∃(α , w) ∈ Θ W ∧ ∃( rk , w) ∈ R W .

Definition 3.3. Any resource can be in one of three stages, a) available() in reservoir; b)
busy() in one task; c) standby() for one task. The busy() stage is defined as
def
busy (rk ) ≡ ∃have(α , rk ) ∧ DoTask ( w). Where predicate DoTask() is defined using temporal
def
logic: DoTask ( w, tm ) ≡ dtm (∀(ri , w) ∈ R W ∃ have(α , ri )). The standby() stage is defined as:
def
standby ( rk ) ≡ ∃have(α , rk ) ∧ ¬DoTask ( w). The available() stage is defined as
def
available( ri ) ≡ ¬busy ( ri ) ∧ ¬standby ( ri ).

Definition 3.4. Any request needs an answer. This answer needs a decision either to yield or
to deny a resource. Decision making process involve to possible actions: Ayield (α i , α j ) and
Adeny (α i , α j ), where α i , α j ∈ Θ are the agent who sends the request and the agent who
receives the request, respectively. The rule to fire yield action can be determined as:
Def
Ayield (α , β ) ≡ ∃β ∈ Θ ∧ (∀request (α , rk )∃ have(α , rk )) → ¬have(α , rk ) ∧ have( β , rk ).

Of course, there area systemic rules that allow or avoid to yield a resource. The decision rules
def def
can be defined as: yield (ri ) ≡ if (available(ri )) → Ayield () and deny (ri ) ≡ if (busy (ri )) → Adeny ().

Emotionality is introduced as an appraisal process that can modify yield() and deny()
predicates. Taking Ortony’s et al. ideas about emotions, we identify three relevant emotional
aspects in decision: a) agent’s relationship based emotions; b) agent’s task based emotions; c)
agent’s action based emotions. First kind of emotions deals with a valuation of relationships
between agents, second kind deals with how many valuable is a task for the agent, and third
kind deals with the response either positive o negative to a request.

Definition 3.5. Emotional decision rules are determined as follows:

def
1. OtherHappiness(α j ) ≡ ◊(if (∃ request (α j , rk )) → Ayield (α j , α i )).
def
2. ValueOfTask ( w) ≡ ◊(if (∃request (α j , rk )) → Adeny (α j , α i ))
To determine true values for these emotional statements we can use a possible worlds
semantics like was proposed by Kripke (cited in Bell, et al. 2001:180). The underlying idea is
that one agent has different emotional states which impact in their emotional response to one
request. For example, one simple possible world can be the existence of a “well-being”
emotional with an emotional value of “joy”, and the “attraction” emotional state with the
emotional value of “love”. These emotions reinforce a yield() decision. Another possible
world can be the same “joy” emotional value for the well-being emotional state, and attraction
emotion with the emotional value of hate. These emotions reinforce the deny() decision.
Formally speaking:

Definition 3.6. Let Ô an Ortony’s ontology, i.e., Ô ≠ ∅ and Oˆ = {Ψ1 , , Ψ n }, where Ψ k is

an emotional state according to it was defined in section 2.2. Let Ω Ψ = {ϕ1 ,… , ϕ r } the domain
of emotional values for emotional state Ψ. A possible world will be m = {x : x = Ψ |ϕ }. The set
of possible worlds will be M. There area a function f :{T , ⊥} → m ∈ M . Of course, researcher
is free to propose combinations of the Ortony’s emotional ontology to define possible worlds
and function f.

To evaluate the “desirability” of an event, the “praiseworthiness” of an action, and the

“appealingness” of an object, agents must have knowledge about their “feels” to their self and
others goals, feels about another agents, and feels about objects. Ortony, et al.’s model is not
explicit about this matter. We belief that is so important be explicit about the affective
relationship between agents. This allows to evaluate the type of “emotion” that must be
related with an event, and agent or and object.

Definition 3.7. Let α i , α j ∈ Θ , there is a affective relationship between α i and α j defined as

∀x, y ∈ Θ, AffectiveRelation( x, y ) ⎯⎯
R
→ r ∈ { friend , enemy, neutral}.

Definition 3.8. Let w ∈ W , and α i ∈ Θ there is an emotional relevance for this goal defined as
∀α i ∈ Θ, w ∈ W , ∃ EmotionalRelevance(α i , w) ⎯⎯
R
→ r ' ∈ {high, medium, low}.

Definition 3.9. Let r ∈ R and let w ∈ W , and w r ∈ W × R, and α i ∈ Θ then, there is an

Attitude of α i towards resource r, so that agent α i w defined as
Attitude(α i , r ) ⎯⎯
R
→ r '' ∈ {indispensable, optional}.

Finally, we define the effect of actions on emotional states. Either yield or deny a resource
will change emotional values for some specific emotional states. For example, deny a
resource can change the emotional value “joy” to “distress” for the emotional state “well-
being emotion”. Indeed, emotional state “liking” can change from emotional value “happy-
for” to “resentment”. This can be formalized as:
def
Definition 3.10. Let Append (Ψ, (ϕ1 , ϕ2 )) ≡ (∃Ψ ∈ m) ∧ Ψ |ϕ1 → Ψ |ϕ2 .
4. An example

We present a simple example in order to understand how our model works. Let α1 , α 2 ∈ Θ a
pair of agents. There are the following relationship between agent α1 and α 2 ,
AffectiveRelation(α1 ,α 2 ) = friend . Agent α1 request to agent α 2 for the resource r5 . This
resource is necessary to do task w3 . Agent α 2 is “busy” doing task w3 . The rational way to
react to this situation is deny the resource. But, the request (an event) of resource r5 generate
an “emotional reaction” in agent α 2 . Agent α 2 has a “low” EmotionalRelevance for task w.
Also, the “low” relevance of task is a cause for the attitude toward the resource r5 . In this
case, attitude is “optional”. All the “emotional calculus” can be visualized as:

AffectiveRelation( Friend ) ∧ busy (r5 ) ⇒ Fortune − of − Others ( sorry − for )

EmotionalRelevance(low) ∧ busy ( r5 ) ⇒ Attribution( shame)
Attitude(optional ) ∧ busy (r5 ) ⇒ Attraction(hate)

All this “emotional reaction” modify the mood of agent α 2 . The Well-Being(joy) is
transformed in Well-Being(distress). The evaluation of the “event” is the classified using the
following formula:

if (Fortune − of − Others(sorry − for) ∧ Attribution(shame) ∧ Attraction(hate) ⇒

desirability(displeased ))

This desirability of event modified the “rational action”. Agent α 2 ”take the decision” of stop
the task w3 , yield the resource r5 , and “be happy” because her/his friend is happy.

As can be see, the formal model allows us to create models of emotional situations. Of course,
this is only a first step in our research and we hope in a short time be able to present a
“computational framework” in which all these ideas can be explored and evaluated their
usefulness.

5. Conclusions

We have presented a model of join decision-making process with emotionality contexts and
have described formalism for designing organizational agent-based systems with explicit
emotional stages. In respect to other formulations about emotion modeling, our formulation
uses a cognitive formulation of emotions (see for example Cañamero, 2002; Ball, 2002;
Perlovsky, 2006). We think that this cognitive formulation is a natural way to think about
emotions. Also it is easy to code in a modal logic framework (For example, O’Rorke and
Ortony, 1994, uses a situational calculus framework which is a derivation from modal logic).
One specific advantage of our formulation with respect to Howard’s model is that emotions
are explicit in our formulation and the emotional mechanism can be reformulated in one easy
way. As the last point, we belief that our formulation can be useful to develop a
computational models using general purpose available software as PROLOG or specific
modal logic language as METATEM (Wooldridge, 2000, p. 168) This is a question that we
will answer in a future work.
Acknowledgments

This research was supported by the Government of the State of San Luis Potosi, Mexico, and
the National Council of Sciences and Technology of Mexico (CONACYT) through the grant
FOMIX-SLP, No. FMSLP-2003-C02-11121.
References

Aguilera, A. & López, A. (2001) Modelado multiagente de sistemas socioeconómicos: Una

introducción al uso de la inteligencia artificial en la investigación social, México: El Colegio
de San Luis, A.C.

Ball, E. (2002) ”A Bayesian heart: Computer recognition and simulation of emotion” in

autonomous agents” in Trappl, R., Petta, P. & Payr S. (Editors). Emotions in Humans and
Artifact. Cambridge, MA: MIT Press, pp. 303-332.

Bell, J.L., DeVidi, D., & Solomon. G. (2001) Logical Options: An introduction to classical
and alternative logics. Canada: Broadview Press.

Cañamero, L. D. (2002) “Designing emotions for activity selection in autonomous agents” in

Trappl, R., Petta, P. & Payr S. (Editors). Emotions in Humans and Artifact. Cambridge, MA:
MIT Press, pp. 115-148.

Cohen, M. D., March, J.G. & Olsen, J.P. (1972) “A garbage can model of organizational
choice”. Administrative Science Quarterly, Vol. 17, No. 1, 1-25.

Elster, J. (1996) “Rationality and emotions”, The Economic Journal, Vol. 106, No. 438, 1386-
1397.

Fineman, S. (1996) “Emotion and Organizing”; in Clegg, S.R., Hardy, C. and Nord, W. R.
(Editors) (1996) Handbook of Organization Studies, London: SAGE, pp. 543-564.

Howard, N. (1993) “The Role of Emotions in Multi-Organizational Decision-Making”, The

Journal of the Operational Research Society, Vol. 44, No. 6 (June), 613-623.

Lomi, A. & Larsen, E. R. (Editors) (2001) Dynamics of Organizations: Computational

Modeling and Organization Theories, Menlo Park, NY: AAAI Press and MIT Press.

Loewenstein, G. (2000) “Emotions in economic theory and economic behavior”. The

American Economic Review, Vol. 90, No. 2, 426-432.

O’Rorke, P. & Ortony, A. (1994) “Explaining emotions”; en Cognitive Science, 18, 283-323.

Ortony, A.; Clore, G. L. & Collins, A. (1988) The Cognitive Structure of Emotions,
Cambridge: Cambridge University Press.

Perlovsky, L. I. (2006) “Toward physics of the mind: Concepts, emotions, consciousness, and
symbols”, Physics of Life Reviews 3, 23–55.
Prietula, M.J., Carley, K.M. & Gasser, L. (Editores) (1998) Simulating organizations:
Computational models of institutions and groups, Cambridge, MA: The MIT Press.

Salovey, P., & Mayer, J. (1990). “Emotional intelligence”, Imagination, Cognition and
Personality, 9, 185-211.

Thoits, P. A. (1989), “Sociology of emotions”. Annual Review of Sociology, Vol. 15, 317-342.

Trappl, R, & Payr, S. (2002), “Emotions: From brain research to computer game
development” in Trappl, R., Petta, P. & Payr S. (Editors). Emotions in Humans and Artifact.
Cambridge, MA: MIT Press, pp. 1-10.

Trappl, R., Petta, P. & Payr, S. (Editors) (2002), Emotions in Humans and Artifacts.
Cambridge, MA: MIT Press.

Watson, D., & Clark, L. A. (1984). “Negative affectivity: The disposition to experience
negative emotional states”. The Journal of Personality and Social Psychology, 60, 927-940.

Weiss, H. M., & Cropanzano, R. (1996). “Affective events theory: A theoretical discussion of
the structure, causes and consequences of affective experiences at work”. Research in
Organizational Behavior, 18, 1-74.

Wooldridge, M. (2000) Reasoning about rational agents. Cambridge, MA: MIT Press.