Formal Goal-based Modeling of Organizations

Viara Popova

1,2

and Alexei Sharpanskykh

De Montfort University, Centre for Manufacturing, The Gateway, Leicester, LE1 9BH, U.K.

Vrije Universiteit Amsterdam, De Boelelaan 1081a, 1081 HV Amsterdam, The Netherlands

Abstract. Each organization exists for the achievement of certain goals. To

ensure continued success, the organization should monitor its performance

w.r.t. these goals. Performance is often evaluated by estimating performance

indicators (PIs). In most existing organization modeling approaches the relation

between PIs and goals is implicit. This paper proposes a formal approach for

modeling goals based on PIs and defines mechanisms for establishing goal

satisfaction, which enable evaluation of organizational performance. Analysis

and methodological issues related to goals are briefly discussed.

1 Introduction

The behavior of an organization is usually guided by its strategic and tactical goals.

The performance of an organization is often evaluated by estimating the values of its

qualitative and quantitative performance indicators (e.g., profits, number of clients).

To ensure effectiveness of an organization, the key PIs should be reflected in its

goals. In most existing approaches on organization modeling the relation between PIs

and goals remains implicit. This paper defines a clear and general mechanism for

specifying goals based on PIs. The organization’s performance can then be evaluated

by estimating the satisfaction of its goals. Moreover, often the satisfaction of goals

can only be established in a framework, in which goals are related to other concepts

(such as tasks, roles and agents). A goal-based modeling approach proposed in this

paper constitutes a part of a general organization modeling and analysis framework, in

which organizations are considered from four interrelated perspectives (or views). In

particular, the performance-oriented view addresses PIs, goals and relations between

them; the process-oriented view considers tasks, workflows and resources; the

organization-oriented view defines roles, their authority and interaction relations; the

agent-oriented view identifies different types of agents with their capabilities, and

principles of allocation of agents to roles. The views are linked via relations between

their concepts, e.g., agents are allocated to roles, the roles are assigned tasks, tasks

realize goals, etc. Concepts and relations within every view are formally described

using dedicated languages expressive enough to convey structures and processes of

organizations of most types. To provide the formal meaning for the concepts and to

ensure consistency of specifications, axioms and constraints are defined that establish

relations between concepts within and across views.

Popova V. and Sharpanskykh A. (2008).

Formal Goal-based Modeling of Organizations.

In Proceedings of the 6th International Workshop on Modelling, Simulation, Veriﬁcation and Validation of Enterprise Information Systems, pages 19-28

DOI: 10.5220/0001735800190028

 SciTePress

The formal language, axioms and constraints specific for modeling goals within the

performance-oriented view are described in this paper. Moreover, some of the

verification techniques specific for the performance-oriented view are briefly

considered. Furthermore, the paper addresses methodological issues of creating and

revising goal structures, and considers the process of organizational performance

evaluation based on goals.

The proposed approach was applied for modeling and analyzing an organization

from the security domain within the project Cybernetic Incident Management. The

main purpose of the organization is to deliver security services. It has multi-level

structure, with predefined job descriptions for employees. The examples given in this

paper are related to a part of the organization concerned with planning the allocation

of security officers to locations of customers. The planning process consists of

forward (or long-term) planning and short-term planning. Forward planning is a

process of creation, analysis and optimization of forward plans for the allocation of

security officers for a long term. During the short-term planning, plans for the

allocation of security officers to locations within a certain area for a short term are

created and updated based on the forward plan and up-to-date information about the

security employees. Due to the space limitations only some parts of this case study are

given in the paper. A more elaborated description of the case study is given in [7].

The paper is organized as follows. In Section 2 the main concepts for the goal

modeling framework are specified. The relationships between them are described and

formalized using the dedicated logic-based language in Section 3. Section 4 discusses

how the performance of the organization can be evaluated. Some design principles are

given in Section 5. Section 6 discusses the related work on goal-oriented modeling.

2 Goal Modeling Concepts

Each organization exists for the achievement of one or more goals. This varies

depending on the type of organization, e.g., the main goal of a manufacturing

company can be the realization of maximal profit. Being aware of these goals is a

prerequisite to taking measures for their satisfaction. To ensure continued success, the

organization should monitor its performance w.r.t. formulated goals. To enable goal-

based performance evaluation, organizational goals should be formulated over

performance measures (indicators).

Performance Indicators (PIs) are defined as measures, quantitative or qualitative,

that can be used to give a view on the state or progress of the company, a unit within

the company or an individual (e.g., time to produce a short-term plan, efficiency of

allocation of security officers). Expressions can be formulated over PIs containing

>, =

<, for example for defining target values: an expression over the PI P1:“efficiency of

allocation of security officers”

is defined as P1 = high.

Goal Pattern is a property over one or more PI expressions used to define a goal. A

goal pattern can be checked for a given state/time point or interval for a company or

agent. Goal patterns have a

type specifying the way the property will be checked:

achieved (ceased) – check if the property is true for a given time point; maintained

(avoided) – check if the property is true for the duration of a given time interval;

optimized (maximized/ minimized/approximated) – check if the value of the PI expression

has increased/decreased/approached a target value for the given time interval.

Achieved, ceased, maintained, avoided are used on PI expressions that are evaluated to

Boolean values;

optimized is used on PI expressions that are evaluated to value of any

ordered type (maximized/minimized) or for which a distance measure is defined

(approximated). Examples: “maintained efficiency of allocation of security officers = high” with

maintained pattern type and “achieved that time to produce short-term plan ≤ 48” with achieved

pattern type. Goals are formulated by adding to goal patterns information such as

desirability and priority.

Goal is an objective describing a desired state or development of the company or an

individual. Example:

“it is required to maintain high efficiency of allocation of security officers”. A

goal is characterized by evaluation type, horizon, ownership, hardness, priority,

negotiability, etc

. Evaluation type determines if a goal is based on goal pattern of type

achieved or ceased (achievement goal), or if a goal is based on goal pattern maintained,

avoided or optimized (development goal). Horizon specifies within which time interval

(for development goals) or at which time point (for achievement goals) is the goal

supposed to be satisfied:

long-, mid-long- or short-term goal. Ownership can be

organizational (of an organization/unit/role) or individual (of an agent). Goals of agents

may comply with organizational goals or not.

Priority is defined by the numerical

estimation between 0 and 1; alternatively {very high, high, medium, low, very low} or a (partial)

ordering on goals may be defined. Normally, organizational goals have a higher level

of priority. Priority of individual goals depends on the company policy: one company

might assign lower priority to individual goals; another might decide to involve and

motivate the agents by taking into account their goals and avoiding conflicts between

individual and organizational goals. By

negotiability goals are divided into non-

negotiable and negotiable. This can be used for conflict resolution at the design phase.

Hardness distinguishes soft and hard goals. Satisfaction of a soft goal cannot be

clearly established. We use the term

satisficing to indicate acceptable degree of

satisfaction of soft goals. Labels are given corresponding to the degrees of

satisficing/denial with a natural order:

satisficed > weakly_satisficed > undetermined >

weakly_denied > denied. Satisfaction of hard goals can be established quantitatively. Hard

goals also have labels ordered as follows: satisfied > undetermined > failed. In the example

below

G3 is soft, PI “efficiency of allocation” cannot be objectively established to be

maintained high or not, instead we use a subjective estimation of degree of satisficing.

G3.1 is hard – it can be checked if PI “time to produce up-to-date plan” is at most 48 hours.

Goals are realizable by organizational tasks. A task represents a function performed

by role(s). A role is characterized by a set of functionalities performed by it. Roles are

allocated to agents to perform tasks. Roles and agents are committed to certain goals.

Besides organizational goals, an agent may pursue its own individual goals that

comply or conflict with organizational goals. These and other concepts are only

briefly discussed in this paper and are considered in the descriptions of other views.

Examples:

Goal name:

G3.1

Goal name: G3

Informal definition: It is required to maintain high

efficiency of allocation of security officers to objects

Informal definition: It is required to achieve that

within 48 hours from receiving operational data,

an up-to-date short-term plan exists

Eval. type: development goal (maintain goal pattern) Eval. type: achievement goal (achieve goal

pattern)

Horizon: long-term; Ownership: organizational Horizon: short-term; Ownership: organizational

Hardness: soft Hardness: hard

Priority: high; Negotiability: negotiable Priority: medium; Negotiability: negotiable

3 Formal Goal Modeling

To specify the meta-model for the performance-oriented view the first order sorted

predicate language is used (the graphical representation of the developed meta-model

is given in [7]). In this language, for each concept a special sort is introduced, which

contains all the names of concept instances (e.g., sort

GOAL contains all names of

goals). Using this dedicated language a number of relations between goals and other

concepts are specified. To provide the formal meaning for the introduced relations

and to ensure the consistency and integrity of goal-based specifications, axioms and

constraints are defined along the definitions of relations.

The relations on goals and PIs introduced informally in Section 2 are formalized as

follows:

is_based_on: GOAL_PATTERN × PI: the goal pattern in the first argument is defined over

the PI in the second argument;

uses: GOAL_PATTERN × PI_EXPRESSION: goal pattern

defined over PI expression. For example, the goal pattern GP1 “maintained efficiency of

allocation of security officers to objects = high

” is based on the PI P2 “efficiency of allocation of

security officers to objects

” and uses the PI expression PE1 formulated over P2, (PE1: P2=high):

is_based_on(GP1,P2); uses(GP1,PE1)

is_formulated_over: GOAL × GOAL_PATTERN: The goal is defined over the goal pattern.

For example, goal

G3 defined earlier is formulated over the goal pattern GP1.

Goals are related to tasks, roles and agents by the following relations:

is_realizable_by:

GOAL × TASK_LIST: the goal in the first argument is realizable by the list of tasks in the

second argument; is_committed_to: ROLE × GOAL: the goal is an organizational goal and

the role is committed to the satisfaction of this goal; wishes: AGENT × GOAL: the goal is

an individual goal of the agent. For example role

Planner is committed to goal G3.1

realizable by task T4.4.1:“update short-term plan”, i.e. is_committed_to(Planner,G3.1) &

is_realizable_by(G3.1,L41) & is_in_ task_list(L41,T4.4.1)

, where is_in_task_list: TASK_LIST × TASK.

A goal structure can be built by refining high level goals and by aggregating lower

lever goals into higher level goals. Since goals in the modeling framework can be of

two types: hard and soft, having very different features, different types of refinement

relations should be considered.

Hard goals are refined into and-lists of hard goals (sort

AND_GOAL_LIST), in which

the goals are connected by and relation. A refinement of a hard goal is specified by:

is_refined_to: GOAL × AND_GOAL_LIST: Defines a refinement of a hard goal into a list of

hard goals, which contribute to its satisfaction. When all the goals in the list are

satisfied then the goal in the first argument is satisfied as well. If one or more goals in

the list fail and no other refinement exists where all goals are satisfied, then the goal

in the first argument will fail too. More formally, the predicates

satisfied: GOAL and failed:

GOAL

express the satisfaction state of a goal and the following axioms are formulated:

is_refined_to(g, l) & (∀gi:GOAL is_in_ goal_list(gi,l) ⇒ satisfied(gi)) ⇒ satisfied(g)

∀ l: AND_GOAL_LIST (is_refined_to(g, l) ⇒ ∃gi: GOAL is_in_ goal_list(gi, l) & failed(gi)) ⇒ failed(g)

where is_in_goal_list: GOAL × GOAL_LIST expresses that a goal is in a goal list. Sort

GOAL_LIST is a supersort of AND_GOAL_LIST, which contains the names of all goal lists.

When more than one refinements are defined, they are considered as alternatives

connected by

OR, i.e., they allow a choice, which measures to take to satisfy the goal.

Examples of refinement of hard goals are provided in [7].

Since the satisfaction of soft goals cannot be established in a clear-cut way, their

refinement differs from the refinement of hard goals. Instead of decomposition, we

talk about positive contribution from other goals to the satisfaction of the goal. Such

contribution can vary in its degree expressed by the following relations, in which the

second argument is a soft goal and the first argument can be soft or hard. In

satisfices:

GOAL × GOAL the first goal strongly contributes positively to the satisficing of the

second goal. If it is satisfi(c)ed and no other influences are known then the second

goal is considered satisficed. In

contributes_to: GOAL × GOAL the first goal contributes

positively to the satisficing of the second goal, but might not be enough to satisfice it.

The precise meaning of these relations is defined through the propagation rules for

goals refinement. These rules are used to determine the degree of satisficing of a

higher level goal (specified by a label) based on the available information about the

degrees of satisfaction/satisficing of lower level goals in its refinement list. To

determine the label of a higher level goal, first the labels of its contributing (lower

level) goals are propagated, taking into account the types of the links by which the

lower level goals are connected to the higher level goal. The propagated labels of the

lower level goals are determined using Table 1. Then, the propagated labels of the

lower level goals are combined depending on the type of the list to determine the label

of the higher level goal.

Lower level goals can be combined in lists using and- or balanced contribution

relations, contributing positively to the satisficing of higher level soft goals:

has_influence_from: GOAL × GOAL_LIST: The goals in the list contribute positively to the

satisficing of the soft goal in the first argument. For each goal in the list it is defined

separately what the level is of its contribution (the type of the link) using the above

defined relations

satisfices and contributes_to.

The combination of goals in an and-list implies that if all goals in the list are

satisfi(c)ed then the higher level goal will also be satisficed. In order to ensure this the

following constraint is enforced: at least one of the goals in an and-list is connected

with a link of the type

satisfices to the higher level goal. When lower level goals are

combined in an and-list, the label of a higher level goal is defined by the minimal

label propagated from the goals in this list using the defined order between the labels.

Another kind of relation between goals represents balanced contribution which

gives us the possibility to describe more fine-tuned ways of contributing which favor

the majority influence. The rule that is used to calculate the exact effect first

quantifies the propagated labels of lower level goals and then takes the (weighted)

average which is then discretized again to the closest label, which is the sought label

for the higher level soft goal. The quantification scale for the propagated labels may

look as follows:

satisficed = 2, weakly_satisficed = 1, undetermined = 0, weakly_denied = -1, denied = -

. Then, to fine-tune influences that the lower level goals from the balanced list (and

thus, the propagated labels) have on the label for the higher level goal, weights can be

assigned for the lower level goals in the list. Let the quantified propagated labels from

the goals in the balanced list be

and the weights defined for each goal in the list are

. Then the influence of the balanced list on the higher level goal is calculated using a

formula of the type:

Σ w

/ Σ w

. The weight of a goal in a balanced list is specified by:

has_weight_in_list: GOAL × INTEGER × BAL_GOAL_LIST. In the following an example from

the case study of a soft goal refinement by a balanced list is considered. Examples of

refinement of soft goals by and-lists are provided in [7].

Table 1. The table for determining propagated labels for a higher level goal based on the

satisfaction/satisficing labels of lower level goals and types of contributing links.

Contributing goal label \ Type of link satisfices contributes_to

satisficed/satisfied satisficed weakly_satisficed

weakly_satisficed weakly_satisficed undetermined

undetermined undetermined undetermined

weakly_denied weakly_denied undetermined

denied/failed denied weakly_denied

Example: Consider the following set of goals and relations between them (see Fig. 1):

G7: It is required to maintain optimal number of qualified personnel

G7.1: It is required to maintain high qualification of personnel

G7.2: It is required to maintain up-to-date data on the available and needed capacity and qualifications

G7.3: It is required to maintain timely recruitment and dismissal of personnel according to the data on

available and needed (human) capacity and qualifications.

Let us assume that the degrees of satisficing of the lower-level goals G7.1, G7.2 and

G7.3 are known: G7.1 is satisficed, G7.2 and G7.3 are weakly satisficed. The labels are

quantified and the degree of satisficing of G7.1 is considered 2 and of G7.2 and G7.3 – 1.

Thus the degree of satisficing of G7 is calculated as (3*2+1+1)/5 = 1.6 which we round up

2 (which corresponds to satisficed in our scale). Thus the label assigned to G7 is 2.

is_in_goal_list(G7.1, L1),

is_in_goal_list(G7.2, L1),

is_in_goal_list(G7.3, L1),

has_influence_from(G7, L1),

satisfices(G7.1, G7) ,

satisfices(G7.2, G7) ,

satisfices(G7.3, G7),

has_weight_in_list(G7.1, 3, L1),

has_weight_in_list(G7.2, 1, L1),

has_weight_in_list(G7.3, 1, L1)

Fig. 1. The refinement of the soft goal G7 into the balanced list consisting of goals G7.1, G7.2

and G7.3; and the conflict relations between G7 and the goals G8 and G9.

When a goal is refined in one list only then the influence calculated using the

described above rules defines the satisficing label of the goal. Sometimes a goal is

refined into several influence lists related by or. This reflects the knowledge that these

lists are in conflict or competition and if one is satisficed then the probability that the

rest will also be satisficed is lower. In such situations we use the following strategy:

first the influences of the and- and balanced lists are calculated separately and the

highest among them label is assigned to the higher level goal.

Apart from the refinement links discussed so far, we can also define conflicts,

which represent negative relations between goals or lists of goals.

conflicts_with: AND_GOAL_LIST × AND_GOAL_LIST: Represents joint negative effect

between lists of goals – the goals in both lists cannot be satisfi(c)ed or weakly

satisficed at the same time. More precisely, if all goals in one list are satisfi(c)ed then

at least one goal in the other is failed or denied; if all goals in one list are at least

weakly satisficed, at least one goal in the other is at most weakly denied.

weakly_conflicts_with: AND_GOAL_LIST × AND_GOAL_LIST: Represents weak joint negative

effect between lists of goals, i.e., the goals in both lists cannot be satisfi(c)ed at the

same time. More precisely, if all goals in one list are satisfi(c)ed then at least one goal

satisfices link

balanced

contribution relation

weak conflict link

soft goal

++ ++

G7.1 G7.2 G7.3

in the other is at most weakly denied; if all goals in one list are at least weakly

satisficed then at least one goal in the other is at most weakly satisficed.

Conflicts can be defined at each two levels of the goal hierarchy, however if the

hierarchy is sufficiently complete then these conflicts should be propagated through

the goal refinement to the lowest level at which the sources of the conflicts can be

found. Conflicts can also be used at the analysis and evaluation phases for

propagating satisfaction labels when only partial information is available. For

example let goals

g1 and g2 be in conflict at the lowest level of the goals structure and

g1 is known to be satisfied. Then if the satisfaction label of g2 is not known it can be

assumed at most weakly denied if g2 is soft and failed if g2 is a hard goal. If however

it is known that g2 is satisfi(c)ed then there is an inconsistency in the specification. In

the example on Fig.1 goals

G8 “It is required to minimize training for personnel” and G9 “It is

required to minimize recruitment of personnel”

are in weak conflict with G7.

4 Goal-based Evaluation of Performance

Every task performed in an organization contributes to the satisfaction of a certain

organizational goal(s). Tasks are realized by processes in the organization’s

workflow. Each goal is formed based on a PI(s). This PI(s) can be measured (directly

or indirectly) during or after the process execution depending on the goal evaluation

type, in the end or during a certain period of time (an evaluation period defined as a

goal horizon). Then, by comparing the measured value(s) with the goal expression(s),

the satisfaction (degree of satisficing) of the goal(s) is determined. In many cases

however it is not feasible (e.g., too expensive or difficult) or not possible to monitor

and measure all necessary PIs. Then the mechanisms for propagating goal satisfaction

values through the goals hierarchy can be used. It is only necessary to evaluate the PIs

of the low-level goals. The obtained goal satisfaction values are propagated using the

rules from Section 3, upwards in the goal hierarchy for determining the satisfaction

(degree of satisficing) of higher-level goals. Thus, the organizational performance is

evaluated by determining the satisfaction (degree of satisficing) of key organizational

goals. The same principles can be applied for evaluation of agent performance.

As illustration of the proposed performance evaluation procedure consider the

following example. For estimating the performance of the organization introduced in

Section 1, the degree of satisficing of the high-level soft goal

G3 (defined in Section 2)

has to be determined.

G3 is refined into an and-list of more specific goals including

hard goal

G3.1 (Section 2) to restrict the duration of the short-term planning process

and other goals on efficiency of forward planning, plans distribution, realization of

plans, etc.

G3.1 is linked to G3 by a satisfices-link, therefore satisfices(G3.1, G3), G3.1 in

turn is refined into an and-list consisting of two hard goals:

G3.1.1 “It is required to achieve

that within 48 hours from receiving a new contract, a new short-term plan is produced”

and G3.1.2 “It is

required to achieve that within 48 hours from receiving data about necessary changes in the short-term plan,

an updated short-term plan is produced”

. The PIs corresponding to G3.1.1 and G3.1.2 are

PI1:“time needed to create a short term plan” and PI2:“time needed to update a short term plan”. These

lower-level goals are related to tasks specified in the task graph:

G3.1.1 is related to the

task T1:“generate a new short-term plan” and G3.1.2 to the task T2:“update short term plan”.

By measuring the actual task execution during the evaluation period defined for

G3.1 (a month), it is determined that values for both PIs corresponding to G3.1.1 and

G3.1.2 (PI1 and PI2) do not exceed 48 hours. Therefore the hard goals G3.1.1 and G3.1.2

are satisfied. Due to the refinement relations, it can be concluded that G3.1 is also

satisfied, therefore it contributes maximally to the satisfaction of G3 – if for example

all other goals in the refinement are satisfi(c)ed then

G3 is considered satisficed. G3 is

at the highest level of the goals hierarchy, thus its satisfaction gives a strong positive

evaluation of the overall organizational performance.

5 Methodological Issues of Goal Design

Usually, high level goals of a company are of a strategic (long-term) type. Such goals

are often made operational by refining them into lower-level tactical (short-term)

goals (a top-down approach). The refinement of goals may proceed until subgoals are

found, which could be realized by lowest-level tasks from the task hierarchy. In

practice, the top-down design approach is often combined with the bottom-up

approach, performed by aggregation of goals. In the goal elicitation approach

described in [4] subgoals are identified by asking “how” questions about the goals

already defined, and parent goals are identified by asking “why” questions. Many of

the strategic and tactical goals can be extracted from organizational documents (e.g.,

organizational strategy, mission statements, policies). Although, goals and objectives

may not be stated very precisely in these documents, still in most cases PI expressions

can be extracted from them. Further, the obtained PI expressions are used to formulate

formal goals. Relations between goals can be identified via relations in the PI

structures [6]. To ensure consistency of goal and PI structures, consistency checks can

be performed as follows. If goals are related by refinement relation, then the PIs

corresponding to these goals are related by causality relation. If the PI expressions for

goals related by refinement, contain comparison functions (i.e., ‘

>’,’<’) or measures of

degrees (i.e., ‘high’, ‘low’), or goal patterns are specified by change functions (i.e.,

‘increased’, ‘decreased’), then the specific type of causality may be determined. For

example, the goal

“It is required to limit the duration of the reviewing process to a month” (with PI

expression

“duration of reviewing process < 1month”) has a subgoal: “It is desired to increase the

number of reviewers”

(goal pattern “increase(number of reviewers)”). Since goals are related by

refinement, the PIs

“number of reviewers” and “duration of reviewing process” should be related

by negative causality relation (e.g., increasing the first PI decreases the second PI).

The identification of conflict relations between goals is of importance for the

design and evaluation of organizations. To identify conflicts, the goal patterns and the

PIs structure can be used: by knowing the type of a causality relation between PIs and

the types of goal patterns, the presence of a conflict between goals can be determined.

For example, the goal

“It is required to maximize the time spent on examining a plan proposal for

correctness”

and the goal “It is required to minimize the time spent on producing a correct plan” are in

conflict, since the PIs

“the time spent on examining a plan proposal for correctness” and “the time

spent on producing a correct plan”

are related by positive causality relation, and the

corresponding goal patterns are based on opposite functions:

maximize and minimize. If

needed, consistency can be achieved applying conflict resolution techniques [9, 8].

6 Discussion

Goal-oriented modeling is given a special place in the area of enterprise engineering.

Some aspects of our definition of a goal are inspired by state-of-the-art approaches in

enterprise modeling and requirement engineering. There are however significant

differences. For the first time in this framework the concept of a goal is defined based

on PIs, to reflect the way the notions are used in practice and provide intuitive

mechanisms for performance evaluation. Though large body of research on PIs exists

in management science, PIs are hardly ever considered in enterprise modeling [6].

In the enterprise modeling framework CIMOSA [2] the notion of objective is used

to represent business goals for a domain of the enterprise. Unlike this framework, no

distinction is made between hard and soft goals. Relations between the objectives are

not defined and no hierarchy of objectives is built. Also in the TOVE framework [2]

hard and soft goals are not differentiated. Goals can be decomposed in AND/OR

subgoal trees. Goals are treated very simply in the Aris enterprise architecture [2]: no

distinction between soft and hard goals is made, also refinement of goals is not

elaborated. The methodology GRAI/GIM [2] models performance indicators in the

context of decision making, without taking into account relationships with goals.

The i* approach [11] focuses on the dependencies relationships between the

actors. The approach models hard and soft goals and goal dependency relationships

between actors w.r.t. goals. In contrast to our approach, goals in i* are specified

informally, no unified representation is enforced and no relation to PIs is established.

The goal and task hierarchies are coupled in i*, tasks are decomposed to goals and

tasks. In the framework proposed, models of tasks and goals can be specified and

analyzed separately. Furthermore, relations between goals and tasks can be

established and used in analysis in the proposed framework. Tropos [1] is a

methodology for agent-oriented software development based on i* and goals are

treated similarly. The KAOS methodology [4] focuses on requirements elaboration

and provides support in connecting high-level goals to operations, objects and

constraints to be implemented by the software. A goal is defined as an objective to be

achieved by the system; an operational objective is called a constraint. Soft goals are

not considered. Goals and constraints are defined formally using the patterns achieve,

cease, maintain, avoid, optimize, which are reused in our approach in the notion of a

goal pattern. A difference is that in our approach a goal pattern establishes a direct

relation to a PI expression. KAOS goals are formalized using temporal logic and

structured and operationalized to constraints in AND/OR graphs. In our approach a

wider set of goals is considered, some of which cannot be expressed as temporal logic

formulae. The NFR framework [5] focuses on representing non-functional

requirements on the software system as interrelated goals. Three types of goals are

defined: NFR, satisficing and argumentation goals. The last two model design

decisions and arguments. The NFR goals are soft goals which can be refined using

different types of relationships describing how the satisficing of the offspring relates

to the satisficing of the parent goal. The label propagation procedure in our approach

is inspired by but different from the one of the NFR framework. We consider only

positive refinement links; negative links are modeled by conflict links. Besides, we

enrich the refinement with a new balanced type of relation for soft goals, not

previously used in literature, for finer definition of joint contributions of sets of goals

to the satisficing of higher-level goals. Furthermore, in contrast to NFR we relate

goals to PIs and other concepts (e.g., tasks, roles, agents) explicitly, which enables

different types of analysis across different views on organizations.

Agent goals in multi-agent systems are specified by declarative logical

specifications that describe states of the agent system, which are desirable and could

be realized by the agent. Only hard goals are considered. Such declarative goals often

have a simple (software-oriented) format and are often operationalized in agent

programming languages by sequences of actions or plans [10]. Then, the distinction

between goals and tasks, essential for our framework, is not tangible any more.

In summary, this paper presents a formal goal-oriented modeling approach in the

context of the performance-oriented view on organizations. The proposed approach is

based on the idea that goals should be defined over organizational performance

indicators. The proposed approach includes a diverse vocabulary to express goal-

related concepts and relations, in particular w.r.t. performance evaluation, e.g.,

organizational or individual, hard or soft goals, how they contribute to or conflict with

each other’s satisfaction, mechanisms for identifying the (level of) satisfaction/

satisficing of goals, as well as guidelines and techniques for building consistent goal

structures. Goals are related to concepts described in other views of the framework as

well, which enables different types of analysis within and between views; they are

mentioned here but will be elaborated and applied on larger case studies elsewhere.

References

1. Bresciani, P., Perini, A., Giorgini, P., Giunchiglia, F., Mylopoulos, J., 2004. Tropos: An

Agent-Oriented Software Development Methodology. Autonomous Agents and Multi-Agent

Systems, 8, 203-236.

2. Bernus, P. et al. (eds.): Handbook on Architectures of Information Systems, Springer-

Verlag, Heidelberg (1998) 209-241.

3. Chan, F.T.S., 2003. Performance measurement in a supply chain International Journal of

Advanced Manufacturing Technology 21(7), 534-548.

4. Dardenne, A., van Lamsweerde, A., Fiskas, S.: Goal Directed Requirements Acquisition.

Science of Computer Programming, 20 (1993) 3-50

5. Mylopoulos, J., Chung, L., Nixon, B., 1992. Representing and Using Nonfunctional

Requirements: A Process-Oriented Approach. IEEE Transactions on Software Engineering,

18(6), 483-497.

6. Popova, V., Sharpanskykh, A., 2007. Modeling Organizational Performance Indicators. In

Proc. of Int. Modeling and Simulation Multiconference, 165-170.

7. Popova, V., Sharpanskykh, A., 2008. Formal Goal-based Modeling of Organizations.

Technical Report 080103AI, Vrije Universiteit Amsterdam, http://hdl.handle.net/

1871/11732

8. Sycara, K., 1988. Resolving Goal Conflicts via Negotiation. In Proc. of the 7

National

Conference on Artificial Intelligence, 245-250.

9. Van Lamsweerde, A., Darimont, R., Letier, E., 1998. Managing Conflicts in Goal-Driven

Requirements Engineering. IEEE Transaction on Software Engineering, 24(11), 908-926.

10. Van Linder, B., van der Hoek, W., and Meyer, J.-J. Ch., 1996. Formalising motivational

attitudes of agents: On preferences, goals and commitments. In Intelligent Agents Volume

II, vol. LNCS 1037, Springer, 17-32.

11. Yu, E., 1997. Towards Modelling and Reasoning Support for Early-Phase Requirements

Engineering. 3

IEEE Int. Symp. on Requirements Engineering, 226-235.