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 Lpez, 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 Howards and Ortonys 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
In Proceedings of The First World Congress on Social Simulation. August 21-25, 2006, Kyoto University, Kyoto, Japan. Vol. 1. Pp, 351-358.
1

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 Howards model To resolve the complex effects of emotions into decision-making process first we must explain the Howards 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 others behavior. In Howards 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 Howards model but it is necessary to make some specifications. Howards 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 ORorke, 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 ORorke, 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 "wellbeing""attributions". This kind of compound emotions does not correspond to the cooccurrence of their emotional components. Union logical function is used to resolve when a compound emotion is valid in the computational model (ORorke 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 Howards 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 = {1 , , 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 W 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
def

messages between agents. We use the following modal logic predicate formula to determine a request, request ( ,rk ) m ,k have( , rk ). Where the modal logic formula for have() is determine as have( , rk )
def

( , w) 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
busy (rk ) have( , rk ) DoTask ( w). Where predicate DoTask() is defined using temporal
def def

logic: DoTask ( w, tm ) dtm ((ri , w) R W have( , ri )). The standby() stage is defined as:
standby ( rk ) have( , rk ) DoTask ( w). The
available( ri ) busy ( ri ) standby ( ri ).
def def

available()

stage

is

defined

as

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 ( i , j ) and
Adeny ( i , j ), where i , j 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:

Ayield ( , ) (request ( , rk ) have( , rk )) have( , rk ) have( , rk ).

Of course, there area systemic rules that allow or avoid to yield a resource. The decision rules 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 Ortonys et al. ideas about emotions, we identify three relevant emotional aspects in decision: a) agents relationship based emotions; b) agents task based emotions; c) agents 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: 1. OtherHappiness( j ) (if ( request ( j , rk )) Ayield ( j , i )). 2. ValueOfTask ( w) (if (request ( j , rk )) Adeny ( j , i ))
def def def def

Def

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:

an Ortonys ontology, i.e., O and O = { , , }, where is Definition 3.6. Let O 1 n k an emotional state according to it was defined in section 2.2. Let = {1 , , r } the domain of emotional values for emotional state . A possible world will be m = {x : x = | }. 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 Ortonys 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 i , j , there is a affective relationship between i and j defined as
R x, y , AffectiveRelation( x, y ) r { friend , enemy, neutral}.

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

Definition 3.9. Let r R and let w W , and w r W R, and i then, there is an Attitude of i towards resource r, so that agent i w defined as
R Attitude( i , 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 wellbeing emotion. Indeed, emotional state liking can change from emotional value happyfor to resentment. This can be formalized as: Definition 3.10. Let Append (, (1 , 2 )) ( m) |1 |2 .
def

4. An example We present a simple example in order to understand how our model works. Let 1 , 2 a pair of agents. There are the following relationship between agent 1 and 2 , AffectiveRelation(1 , 2 ) = friend . Agent 1 request to agent 2 for the resource r5 . This resource is necessary to do task w3 . Agent 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 2 . Agent 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 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 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 Caamero, 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, ORorke and Ortony, 1994, uses a situational calculus framework which is a derivation from modal logic). One specific advantage of our formulation with respect to Howards 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. & Lpez, A. (2001) Modelado multiagente de sistemas socioeconmicos: Una introduccin al uso de la inteligencia artificial en la investigacin social, Mxico: 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. Caamero, 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, 13861397. 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. ORorke, 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, 2355.

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.