Network Organization Paradigm: ∗ Synergistic Effect on the Productivity of a Collaborative Organization Saad Alqithami Department of Computer Science, Southern Illinois University Carbondale, IL USA alqithami@gmail.com ABSTRACT them. Even though it is rare to find a single paradigm that is Human organizations that have begun to rely on networks the most likely to best describe an organization through its for collaboration are already prolific. Networked collabo- life cycle, the most fitted paradigm (i.e., the style that best ration is highly beneficial in many group activity including describes an organization) guides us to understand an orga- mixed teams of humans and agents. The prospect of under- nization and appreciate its possibilities. However, agents in standing complex interactions on network organizations has an open multi-agent system are self-governed by their own prompted us to develop a paradigm serving as a reference belief systems and have unmanaged and rational behaviors. model for organizations of networked individuals. In this pa- In a previous recent work [5, 4], we explored applications per we present a few salient components suggested to com- that account for spontaneous exigencies in the agents’ ac- prise network organizations. Network properties are central tions to benefit and shape an organization. We found that for incorporating a spectrum of collaboration styles that is traditional organizational paradigms (i.e. hierarchical and outlined in our paradigm. We have introduced synergy as market) lack the representational power in modeling such a specific network effect that embodies collaboration, which spontaneous structure that is formed from frameless actions in turn has the potential to enhance performance at various and connections. The agents in that case seem to collectively levels of an organization as well as the overall productivity form some sort of an organization based their connections of it. over the networks they occupy. For that, we called such formation a network organization, informally described in Definition 1. Categories and Subject Descriptors I.2.11 [Distributed Artificial Intelligence]: Multiagent Definition 1. Network Organization (NO) are large, semi- Systems autonomous, ad-hoc networked individual entities with the aim of automating command and control of distributed com- plex tasks. General Terms Management We aspire to generalize the concept of NO and introduce a novel paradigm that is the best fit to model agents’ ac- tions in an NO that we call Network Organization Paradigm Keywords (NOP) [2]. NOP is one that manifests a network perspective Agents Paradigm, Computational Models, Network Organi- over all aspects of an organization. Although at times an NO zation may exhibit hierarchic feature, it is not characterized by it. NOP guides us to model organizations of large firms working 1. INTRODUCTION on complex, in scope or impact, problems [18]. A significant advancement was established in the network-centric warfare When the agents dwell inside an organization, they form that allowed oversight and control of operations from any repeated patterns of interactions that in result shape the location on the network. Network-centricity stimulates self- structure of their network. There are many existing pat- organization and self-integrating coordination. The US De- terns to describe interactions within organizations, which partment of Defense embraced network centricity paradigm affect their performance features. Horling and Lesser [13] early on to accommodate collaboration and information re- described arrangements and interaction protocols that char- source sharing among distributed military assets and work acterize working relationships among a group of individuals units [1]. Location ignorance is extended in NOP to permit and termed them as paradigms. This included hierarchies, temporal freedom; therefore, operations can be controlled at holarchies, coalitions, teams, etc. Instead, we consider those any time; i.e., asynchronously. Another extension for NOP as features or patterns of interactions that can describe oper- is to allow any credentialed network member node to ex- ating parts of an organization. For us, a paradigm is a term ert influence on operations. In sum, NOP provides a more that capitulates representational power of a more ubiquitous ubiquitously open model. This openness feature may include perspective over its modifier. It is possible for an organiza- transparent entry and exit to the organization. tion to exhibit specific features yet not be characterized by Evolving in the last thirty years, network organizations ∗This paper is an extended version of the papers presented have produced significant impacts on formation and func- in [2, 7] tioning of human organizations. Recent advances in social networking media have accelerated impromptu formation 2. UNDERSTANDING AN NO PARADIGM and adaptations in human populated network organizations There are many actual groups that rely on networks to with benefits from collective pool of human knowledge and organize their activity. Arab Spring and Science Teams are skills. Furthermore, cohesion in human NO is due to com- two examples. The modeling at a more generalized level cuts mon human social traits such as trust and beneficence. We across domains to extricate the model from limited require- have embarked on modeling artificial, agent based network ments of specific domains. A perspective that would model organizations that no doubt will possess features inspired by a generic network organization came to be considered as a human NOs [17]. Although our modeling endeavor aspires paradigm. NOP can model many NO operations that are to endow NO with qualities that are human centric there will applied to open multi-agent systems. Examples are systems remain profound differences. As erected to address specific of river dam control, factory cells, electrical power grids, or- problems, our artificial NO may lack long-term temporal his- ganized labor unions, and traffic control on land, sea, and tory; whereas, human NO often benefit from their collective space. As a paradigm, it does not functionally alter the memories. Even dynamic human NO will possess temporal operations to which it is applied. The paradigm can be un- resilience that is not readily available in agent networks. derstood in terms of the ways it permits arrangement of Earlier studies that focus on the traditional form of or- command and control regimes. Invariably, NO relies on the ganizations was moved by a homologous structure formed network in which it dwells. Thus, a profile of an NOP net- from continuous cooperative interactions among different or- work residence is essential. NO member-nodes (i.e., agents) ganizational entities [9]. In order to address the frequently are critical constituents and will be delineated in separate changing social and economic landscape they operate on, profiles. Target problems (i.e., operations) modeled are im- network as part of the intra-organizational structure was in- portant and will be separately profiled. For simplicity, we troduced. But the impact of networks were not fully consid- would care about flow of data, control, and coordination. ered. On the other hand, the wide use of an inter-organizational The organizations may represent one or more parent insti- structure common among many human NOs is relatively tutions that govern its normative patterns of behavior and neutral and applicable to many real world applications [20]. we will include distinct profiles for them. Broadly speak- Networks strengthen the social communication of an orga- ing, functioning of an NOP can be objective- (i.e., charter-) nization to access critical resources with other organizations driven or pattern driven. Charter-based organizations seek on the network [11] as well as to agilely adapt to environmen- to achieve specific goal(s) such as solving specific problems tal changes [14]. Such properties allow the NO to plastically whereas pattern oriented organizations seek to maintain a transform its internal structure to cope with outside social state such as a flight formation pattern. Either of these or- and information demands which in turn influence behaviors ganization types could be captured in the governance com- of its agents [19]. To this end, we anchor this article on the ponent/profile of the NOP. At this very high level, we sum- intra-organizational structure of NO that is formed among marize an NOP in Definition 2 followed by subsequent de- heterogeneous agents. scription of each component. Since the NO is affected by the structure of its network, one possible effect of the network of interest in this paper is Definition 2. An NOP is a conceptualized tuple consist- synergy among agents. Synergy is instrumental in increas- ing of h networks-profiles, agents-profiles, problems-profiles, ing agents’ efficiency on different tasks by allowing them governance-profiles, institutions-profiles i. to collaborate with each other in an NO. Network effects on the performance of a group have been demonstrated in Profiles in Definition 2 are key concepts in characterizing several recent works. Liemhetcharat and Veloso [15] have the NOP–i.e., the paradigm defines specific NO as profiles studied synergy among agents using a social network frame- change [2]. Those parameters will be introduced in detail work. They built a task-based synergy graph to create an here as informal definitions in order to keep them intuitive ad-hoc team that is efficient in comparison to others with- because symbolism would have created brevity but need- out interfering with existing team structure. The value of lessly obscured the ideas. We emphasize, in this paper, on synergy is determined through agents’ capabilities and dis- describing one important parameter of an NOP: problem tances on the graph where similar agents have similar ca- profile. The process where this profile plays an important pabilities. Parker, et. al. [16] have also used synergy inside role of an NO will be described in a later section. different type of teams in order to improve the efficiency of The network profile is a graph of nodes (i.e., individuals) tasks achievements. From this, inclusion of the synergy in and links among them. The number of links will change as this paper is deployed to improve agents’ performances as a result of not complete graph. The links might richly or well as their network structure. thinly capture ties among individuals because they are most The remainder of the paper is organized as follows. In Sec- likely to be assessed when a mutual event occurs. tion 2, we give a brief introduction to the NOP and focus Definition 3. A network profile is presented in a tuple mainly on one of its key concepts, which is the problem pro- hN , Resource , Pi, where file, and describe the parameters that fall within it. Section 3 introduces one of the important properties that are inherited • N is a set of agents’ profiles who are members of an from the network and affects agents’ behaviors called syn- NO. ergy. Section 4 describes the process of an NOP and how the problem profile plays an important role in navigating among • Resource is the available resources that an NO provides agents when assigning tasks. Finally, we conclude this paper to the agents in order to achieve an organizational and describe some of the future possibilities of this work in charter that is C. Section 5. • P is a set of protocols to govern the activity of an NO that includes norms, rules, and roles. Since the entire network profile might be far larger than an The goal G in the problem profile is generated through NO, members of an NO are required to possess profiles. Each the governance profile of an NOP. Each goal generated will agent will have a public profile that contains all pertinent have different parameters presented in Definition 6 agent attributes including their allegiances with respect to an NO, capabilities, fitness etc. to be compared with other Definition 6. For Every goal Gi ∈ {G} → C where i ∈ agents. This agent’s profile is presented in Definition 4. ~ θperf {x}, there is a tuple: hPlan , IE, EE, ζ, θ, ~ i, where Definition 4. Each agent profile, i ∈ {N }, is a tuple of • Plan is a set of plan(s) needed for the Gi to be achieved. ~i , S~i , Relation hA i , f~it i i ~ i It will be described in detail later on. kill , Pref erence , Aautonomy i. • The agent i allegiance to all things it cares about is • IE is the set of internal events that is a set of planned presented in A. status to be achieved. • Skill is a set of skills that agent i has. It includes the • EE is the set of external events that a giving NO gen- capacity of the agent to handle tasks. erates reactions based upon in order to address certain IE. • Relation is the agent i’s relations with other agents or organizations. • ζ is the mapping function to perceive the relevance of ∀eei → iej , where eei is the ith external event of the • fit is the set of initial fitness values for different types set EE and iej is the j th internal event of the set IE. of tasks based on previous experiences. It helps an NO to decide on which reaction it should perform as a result of a certain outside action. • Pref erence is a set of agent i’s preferences for certain activities. • θ is a set of tasks agents need to handle for executing • Aautonomy is the agent’s autonomy-level at which it can a plan, which is a set of hθ1 , θ2 , . . . , θm i, where m is perform tasks independent from other agents. a unique independent number of tasks. Each task will have its own profile presented next. There are many reasons that compel agents to connect • θperf is an optimal performance threshold for each θ ∈ with each other. The most pertinent reason for our formu- ~ If, at a certain time, performance is lower than θ. lation is to gather in an NO in order to solve a common problem. The problem can be large or small based on the these expected performances, the agents can be evalu- goal that agents aim to achieve. Each distinct goal will cor- ated and reassigned. respond to a distinct associated problem profile that is used The comparison of θperf with an actual task’s performance- in selecting best-fit agents to perform certain tasks. A prob- level is used for two purposes: (a) it allows agents to report lem profile must contain task decomposition detail that pro- problems that they may face as well as (b) it allows assign- vide task precedence and coordination requirements. With ment and in some cases reassignment. θperf does not only enough problem details, a plan can be retrieved from storage depends on the type of task but also on the problem profile of prior plans. If no plans match, a new plan is conceived. provided, the plan to achieve them as well as the agent’s Most often, problems will have corresponding plans that will level of fitness. be retrieved from a case history. When assuming that we have x set of problems and i ∈ {x}, problem i will have its Definition 7. Each task θm ∈ {θ} has a tuple of hPrecedence, own problem profile presented in Definition 5. Independence, MinFitness, θcurrent i, where Definition 5. A problem profile, i ∈ {x}, is considered • Precedence is the temporal order of this task among all ~ a tuple of hControl , Coordination , Gi , Precedence , Independence i, other tasks in the next set of tasks to be assigned to where agents. • Control stands for controlling participants and available • Independence is to indicate that the task can be achieved positions (i.e., roles). alone without any other requirement of prior tasks or • Coordination is a set of coordination rules for each agent in overlapping task completions. or an agent group based on an agent profile for a pos- • MinFitness is the minimum fitness value required from sible assignment. an agent for this task to be achieved. It will include • Gi is the goal that the problem profile i exists to point minimum values from agent’s skills and autonomy-level. out, which includes a set of tasks and set of plans that • θcurrent is the current task performance measure to should be followed to achieve this goal. More details be compared with the optimal performance (i.e., θperf ) about G are presented in an upcoming definition. presented in the goal profile. • Precedence is the precedence of the problem domain com- paring with others (i.e., the priority level of this prob- In general, we consider a plan to be an and-or graph of lem to be addressed next, must be lesser or equal to 1, tasks. Naturally, mutually dependent tasks and tasks with where 1 is the highest priority.) overlapping durations will not be independent. There are different types of tasks that need to be specified before a • Independence stands for the independence of Gi in the task is assigned; most importantly, the task independence problem-profile from other competing goals that can be from other tasks as mentioned in the task profile. On the executed at the same time. one hand, the independence of one task from others means it does not require a prior task completion in order to complete • Pattern is the way to link different NOs. the current task as well as parallel achievement. This type of • Regulation are partially inherited from the network to tasks is assigned immediately to agents and does not require include a set of roles, rule, and norm that is most likely any further classification or evaluation. On the other hand, inherited by its NOs. some tasks are dependent about their completion on com- pletion of other tasks or to be completed in parallel with An NOP is intended to be a generic, meta-model that others. In such a scenario where dependence matters, we outlines prototypical NO instantiations. As such, NOP is check the performance of the agents continuously to make not a direct recipe to be applied just as a set of architectural sure that they are performing tasks in the expected order. principles does not directly yield artifacts. In a later section, For parallel tasks assigned to three or more agents or in a dif- we describe an NOP functions via processes that connect fusion of a task to more than two agents, we will constantly its components in a running NO. Section 3 will focus on check for the network balance [12] using the simple balance studying in details one type of network effect that exists theory equation, where the network is considered balanced among agent living on network and helps in improving their when the number of balanced cycles over the total number performances and the global NO performance. of cycles gives a balanced percentage that is bigger than threshold. We will provide more details about task assign- 3. SYNERGY EFFECT IN NOP ment and reassignment in a later section when we describe In any organization of networked agents, such as an NO, the processes within an NO. there is a level of inter-agent compatibility in which the The governance profile includes the objectives of an NO agents can work together effectively. Such a measure will (i.e., the organizational charters) aw well as patterns of affect the agents’ performances and, as a result, the global which those organizational charters can be achieved. It does output of an NO. As long as there are continual interactions not interfere with both agents and problem profiles, and it between the agents inside the NO, we describe these levels as governs the network profile. Other possible control are in- synergy [15]. When a part of these interactions are not ac- herited form other institutions trough possibly norms [21]. tive, their synergies will be reevaluated and it may affect the The governance and institution profiles are presented in Def- total synergy of their NO. Volatility has set synergy apart initions 8 and 9 respectively. from the traditional learning styles since an agent will no Definition 8. A governance profile is a tuple of hC, Pattern , longer have a synergy with other agents when its connec- F, Au , Operf i, where tions are lost. There exists a synergy profile for each agent as well as a synergy for the local and global network for each • C is the organizational charter adapted from the net- task that has been assigned. The synergy will change over work to generate goals presented by different problem time due to the scale of dynamism in an NO while perform- domains. ing a certain task. • Pattern stands for the pattern of connecting problem- Synergy has a huge impact on organizational performance profiles provided to satisfy the global charter. as a whole as well as on the agents’ performances. In an NO, the current synergies are derived from the network-profile • F is a set of fitness functions for the whole NO to help and modified or controlled through the governance-profile. in evaluating its functioning over time to make sure it The network profile will provide a list of the agents’ profiles follows in a proper direction. that contains their relations with others inside and outside the NO. The synergy contribution of an agent is of a value • Au is the autonomy level of an NO, where with the of “0” when he first joins an NO; then, it is derived from higher level of autonomy, the more independently the his relationships with others. In order to fully understand NO operates. It is self-declared and not externally de- the way we derive synergy, we will describe relations in the termined. agent profile next. • Operf is an optimal organizational performance to be compared with the current performance to measure the 3.1 Relations formation and contribution to NO progress. synergy When a group of agents form a small world to work on a As has been mentioned before, an NO lives on a network certain problem profile, the value of their relations have a that is often far larger than its scope and there may exist huge impact on the formation as well as the coordination in one or more institutional profile within that network envi- this world [10]. It, in return, affects their performances and ronment. The network will have its own regime and control; productivities. Therefore, the agents are obliged to provide, as well institutions will provide their specific norms, rules in their profiles, a set of their relations whether inside or and roles. Common protocols will be inherited directly from outside the problem domain. Those relations are not static the institutional profile. However, when there is a contradic- and the agents are able to improve or diminish these rela- tion in protocols between the institutions and network, NO tions’ values while performing a task. Also, new relations will most likely stay neutral or might follow the institution’s may be formed from existing ones to help in improving a protocols for the worst-case scenario. Abstract definition of total performance of an agent as well as the performance of institution is presented in Definition 9. her NO. The importance of relations has led us to model the agents’ relations as an important parameter in their profiles. Definition 9. A institution profile is a tuple of hCharter, In order to model dynamic values of relations, we capture Pattern, Regulationi, where relations in a goal-based graph. As we described previously • Charter is much bigger than C of NO to give a general in the problem-profile, there are different goals {G} pro- idea of the institution. vided by different problems-profiles, and each Gi ∈ {G} for a problem i is equivalent to a set of tasks hθ1 , θ2 , . . . , θm i that with that amount (i.e., Benj→i ). When a pair of indi- need to be achieved in order for the Gi to be completed. viduals reciprocate benevolence, we call that synergy The coordination and control of those goals are also pro- between them shown in Equation 1. vided by the problem profile, which is generally based on the network-profile and the agents-profiles. During task achieve- i→j ment, values of agent’s relations ebb and flow depending on Synergy = Beni→j + Benj→i (1) nature of interactions that forms links (i.e., edges) among where i and j ∈ N them. The continual changes in inter-agent connections will be used in detailing synergies. • By the time an entire group benefits from an individ- A sociograph, as a part of the network-profile, will be build ual action, we call that generalized benevolence. De- upon the contributing agents’ profiles in order to model in- gree of i’s contribution to group g ∈ {N } is denoted by teractions among agents in each task assigned. The agents GBeni→g . When a group appreciates i’s benevolence, will be presented with a node and the edges are based on we consider the proportional appreciation of benevo- their provided relations in their profiles. Other parameters lence to be a synergy between i and group g. Appreci- in the problem-profile will have an effect on the total value ation can be measured by the importance of an group and shape of the graph. By the generic assembly, the so- g bestows to the individual i denoted by importancei ciograph is not active. However, when agents start to inter- and synergy is shown in Equation 2. act over existing but not active edges, they form an active edge through successive interaction. There are two different i→g types of interactions: (a) explicit affinities when two or more Synergy = GBeni→g × Importancei (2) agents have interactions with whom they have previous ex- where i is an agent belongs to {g} ⊆ {N } periences over an existing edge in the graph (i.e., the edges of a graph is build upon original relations provided by the • An important property of collaboration is timely and agents-profiles). (b) Implicit affinities are the interactions beneficial contribution of actions. When an individ- in between two agents without any previous experience be- ual recognizes a specific opportunity for a timely and tween them [22]. These edges emerge from transitivity of significant action by i for another individual agent j, relations (i.e., previously un-modeled relationships) to be we capture that in complementary collaboration de- explained shortly. noted by CCj→i . Whereas benevolence is a general Based on the different structural configuration of the agents’ offering of helpful action toward another, complemen- coordination, the interactions of a triad can be either mu- tary collaborative action is much more directed and tual, directed one way, directed in reverse, or null. The appreciated by the recipient since it is a response to classification of these interactions is based on the MAN la- a specific opportunity (i.e. a need fulfilled by the re- beling introduced in [8]. This labeling is a reduction of the cipient). Similar to benevolence, synergy is generated 64 possible configurations of a triadic closure (i.e., 4 possi- when it is reciprocated. bilities for 3 edges in a triadic will yield a value of 43 = 64) used in structural balance [12] to 16 by classifying the classes i→j into mutual, asymmetric and null. Such labeling has been Synergy = CCi→j + CCj→i (3) adopted to model the interactions among agents. We drive to find the value of interactions in order to evaluate current where i and j ∈ {N } values of edges or help in forming new ones. At this point • A variation of complementary action is general com- the structural balance of an NO is not essential but will play plementary collaboration (denoted by GCCi→g ) when a role in monitoring task assignments discussed in section 4. i’s action benefits a group g ∈ {N }. With group ap- 3.2 Determination of a synergistic value of an preciation measured by the importance value we derive agent a measure of synergy captured in Equation 4. In a network environment, confluence of individual actions and decisions often yield collective and residual rewards for i→g Synergy = GCCi→g × Importancei (4) the network that would not exist had the individuals not been active members of the network. These rewards are post where i is an agent belongs to {g} ⊆ {N } mortem markers of successful interaction in the network. Al- though we may not be able to quantify how well a network To this end, it becomes clear that the value of synergy is functions during task performance, we can observe the re- proportional the contributor capability and relation toward sults from time to time whenever rewards are witnessed. The another or toward a group. It is one of the major effects of degree of successful interaction is called synergy [15, 16]. Al- the network in an NO that determine its performance and though, synergy will commonly remain implicit, it is always productivity for that the previous possibilities of measures proportional to the amount of reward observed. Here, we are not exhaustive. will elucidate different ways to exhibit synergy in an NO: • Whereas collective reward is the group reward (i.e., 4. THE PROCESSES OF A PROBLEM PRO- utility), residual reward is the reward (i.e., utility) that FILE belongs to specific individuals. When an individual After the NO parameters (i.e., profiles) have been deter- agent i is a recipient of a reward, we call action of mined, an NO will begin functioning by the processes where others (say j) as benevolent toward i. When actions the NO will effectively achieve problems or produce desired can be quantified, we set the benevolence of j toward i patterns. We focus on synergy as a predominant form of Data: The process of f1 in an NO  Given a C and Pattern of an NO from the governance Plans profile; f1∗  Let i be a random G ∈ / {Gn }; f2∗ while C is not satisfied do Problems Plans/Play C × {ee} → {G} if {Gn } = null then f3∗ Let Gi = {Gn }; Tasks/Roles else f5∗ f4∗ Allocation if Gi ∈ {Gn } then P roblem P rof ile exit; end end Figure 1: The flow process of a problem profile in for i : 1 → n do an NOP MergeSort Gi based on a priority level in {Gn }; end end network effect that changes performances. This change can Algorithm 1: The process of generating and prioritize be at the level of individuals or groups. We will briefly out- goals line network effects at these two levels. However, we post- pone detailed discussions of processes to a latter part of this section. Figure 1 depicts a simplified sketch of flow in the the assortment of different tasks that they collaborate with problem profile, as prescribed earlier in this paper. each other in order to achieve. Employing those synergies At the individual level, process f4∗ (see Figure 1) will con- will enrich the NO structure and connectively, which in turn tinually monitor task performances and reassign tasks to will improve the total performance of NO. However, those each agent as needed. In part, an agent’s performance is synergies are not preserved and will immediately be lost by determined by its synergy with others (i.e., a network ef- the time agents complete their current goal or depart from fect). Reassignments will attempt to augment synergies over one goal to another. This is remedied when agents’ profiles a task. I.e., positive network effects will increase task perfor- are updated continually in order to take into consideration mance. At the group level, process f5∗ will monitor progress the new formed values of synergies. As well, an NO will use on the current goal and plan in order to remedy problems the formed network of synergies to improve its performance. with low performance on goals and plans. By initiating the After a plan has been set up for execution, f3∗ will assign process of goal re-assignment, NO will strive to increase net- tasks while taking into consideration agents’ profiles. The work effect on goal performance. By initiating proper prob- process of f3∗ is presented in Algorithm 2. When a task has lem selection, NO will strive to fortify network effects on low performance, f4∗ is used to reassign tasks for other agents problems. based on their level-of-fitness (i.e., fit ). The task will have In an NO, the problem profiles are provided through the low performance when the comparison of its performance governance profile. Problem profiles are mainly generated to (i.e., θcurrent ) with expected performance presented in goal focus on the organizational charter whereas other problems profile (i.e., θperf ) is low on the case based threshold (i.e., are based on a perception of an external event that requires τ ). The status of an NO is reported through triggers. The NO attention. The governance-profile will generate a set reassignment of tasks/roles using f4∗ is triggered through of goals. Each goal will have its own profile that shows its t1 . The trigger t1 will make sure that the condition ti1 : i i priority among others in the set. This set should be updated θcurrent < θperf is satisfied before reassignment (i.e., the continuously in order to prioritize the set before assignment. current performance is not less than the expected once). The Thus, the use of f1∗ is not only to generate a set of goals that performance of an NO is formed through different stages of partly satisfies the charter, it will also update this set for new process. This initial performance is a domain related and generated goals, as presented in the Algorithm 1. can be represented in a scale of “0” as a minimum to “100” for The governance process does not stop unless the com- the maximum. Using those initial performances, an agent’s pleted goals largely satisfy the NO charter. After it gen- performance at a time interval µ for a random task m ∈ {θ} erates a set of goals based on the available parameters of is measured through Equation 5. the NO, the problem-profile will follow the traditional steps of planning (or selecting a prior plan) for each goal. Those |N | |N | goals will go through the planning phase based on the pri- X X Perf (θm , µ + 1) = Perf (θm , µ) + Synergy (θm , µ) (5) ority levels assigned to them by the generator function in i,i0 i,i0 the governance module. In majority of cases, the problem- profile will use a case based script f2∗ to match and assign a where i, i0 ∈ {N }, θm ∈ {θ}, and µ is a time interval. plan or play. f2∗ may generate a plan based on the exiting In the case of dependent task or task assignment to more agents’ profiles as start up for the NO. Then, it will store than two agents, f4∗ will use balance theory in order to ex- them in the plan database for future reference. When a sim- amine the balance of those agents’ network. The balance ilar new goal is needed to be assigned, f2∗ will invoke similar of the network is the percentage of the number of balanced a plan that has been assigned to similar previous goals and cycles over number of existing cycles [12]. The assignment match the new goal with a best-fit plan. and reassignment of tasks will change over time. It will use When the agents work on a goal, they form synergy from the agents new values of synergy to update and strengthen their connections. Those synergies help in improving agents’ performances, which in result change the plan for a better and faster achievement of goals. Data: TaskAssignment for assigning tasks to agents  Given agents’ profiles that include Skill , Pref erences and Autonomy ;  Given a set of tasks Precedence and Independence;  Let i be a random agent ∈ {N };  Let θj be a task ∈ {θm } that is ready to be assigned; for θj : θ1 → θm do StateOfTask θj ; . Refer to Algorithm 3 for i : 1 → |N | do Data: StateOfTask based on tasks profile i  Assume a level of Precedence of {0, 1.0}, where 1.0 is if θj ∈ {Pref erence } then i i fit = Scale − of(Skill + Aiutonomy ); the optimal precedence of a task to have the highest i if fit ≥ MinFitness(θj ) then priority among others and 0 for the complete opposite; Assign: θj → i;  Assume another scale of Independence of {0, 1.0}, end where 1.0 for a complete independence of one task to be end achieved independently from others and 0 for a total dependent on others; end if Precedence = 1.0 then end if Independence = 1.0 then Algorithm 2: TaskAssignment for agents TasksAssignment θj ; . Refer to Algorithm 2 else while θcount : θ1 → θj do By the time the plan is complete and tasks need to be if End(θcount ) ≤ Start(θj ) then assigned, different types of tasks have different priority and TasksAssignment θcount ; independency levels that, in result, take more time for agents end to complete them. The StateOfTask is a simple comparison θcount + +; function that covers tasks’ Precedence and Independence and end sort them for assignment. This function is used to examine TasksAssignment θj ; the process of assigning different types of tasks, presented end in Algorithm 3. else In Algorithm 3, the “Sort” function applies a traditional if Independence =1.0 then sorting style to prioritize tasks based on their precedences. Sort{θ}; The functions “End” and “Start” are for the time intervals for TasksAssignment θj ; each task that are used to make sure there are no overlapping else in tasks achievements when assigning them. Algorithms 2 Sort{θ}; and 3 are complimentary to each other, and the functions, for θcount : θ1 → θj do “TaskAssignment” and “StateOfTask” help to easily navi- if End(θcount ) ≤ Start(θj ) then gate between them. TasksAssignment θcount ; The problem profile should be informed about the status end of the goal assigned. When the tasks/roles have difficulties θcount + +; even after the reassignment, t2 will trigger f5∗ to report the end current status and ask for possible change in the current TasksAssignment θj ; plan. In a case where the goal is taking longer than expected, end f5∗ is used to update the status and to see if an extra time end can be allowed for this tasks to be completed or assign a Algorithm 3: StateOfTask based on tasks profiles different plan. For the possibility of a goal failure, f5∗ will add the goal to OLDGoal set, and f2∗ is required to perform the comparisons of the priorities between the two goal sets and assign the goal with the highest priority. Each goal will have a history added to its profile so that when f2∗ tries to find a plan for a previously assigned goal, it will avoid using a similar plan as assigned before and entering into an infinite loop. f5∗ will also inform the problem profile when the goal has been achieved. Different tasks will have different performance levels. The cumulative value of those task performances present the per- formance values of the goal, which is also calculated through f5∗ , helps in evaluating the process of the goal assigned. The performance of each goal is determined using Equation 6. NO features with features that will be required for specific 1 X domains that are unforeseen at the moment. Giperf = Perf (θm ) (6) m m where i is the problem profile and ∀m ∈ {θ} REFERENCES f5∗ will compare the current value of tasks performance [1] D. Alberts and R. Hayes. Power to the Edge: with the optimal performance showing the goal profile, and Command and Control in the Information Age. CCRP report it to the problem profile. The status of completion Publication Series, Washington, DC, 2003. or failure of a goal are reported to the NO through outside [2] S. Alqithami. A succinct conceptualization of the triggers that are out of the scope in this paper. foundations for a network organization paradigm. In When the current performance passes the threshold of 29th AAAI Conference on Artificial Intelligence, pages the minimum performance, we can consider the organiza- 4140–4141, 2015. tion productive. Thus, the improvement in the performance [3] S. Alqithami, J. Haegele, and H. Hexmoor. will improve the productivity of an NO. Synergy helps in Conceptual modeling of networked organizations: The improving NO productivity since it improves the perfor- case of aum shinrikyo. In B. Issac and N. Israr, mance of the goals through existing network effects among editors, Case Studies in Intelligent Computing: its agents. Low productivity level forces an NO to adopt or Achievements and Trends, pages 391–406. CRC Press, plastically transform with different pattern to perform bet- Taylor and Francis, 2014. ter, which may require an update to all NO profiles. The [4] S. Alqithami and H. Hexmoor. Spontaneous plastic transformation of an NO,addressed in [6] , can be organizations: Collaborative computing model of a briefly described as a group of processes that change the networked organization. In 8th International NO structure in order to maintain acceptable performances. Conference on Collaborative Computing: Networking, Thus, it is one part of the governance profile for managing Applications and Worksharing, pages 643 – 650, 2012. an NO shape and future directions. [5] S. Alqithami and H. Hexmoor. Modeling emergent network organizations. Web Intelligence and Agent 5. SUMMARY AND FUTURE WORK Systems, 12(3):325–339, 2014. An NO can be a small team of two or more agents work- [6] S. Alqithami and H. Hexmoor. Plasticity in network ing on a common, quick goal that is possibly faster than hu- organizations. Journal of Advanced Computational man perceptual threshold (e.g., aerial coordination at high Intelligence and Intelligent Informatics, 18(4):567–572, speeds) or a large collection of agents made up of thousands 2014. of people (i.e., possibly swarms) working on long term ob- [7] S. Alqithami and H. Hexmoor. Ubiquity of network jectives that are possibly beyond a single human’s cognitive organizations: Paradigmatic perspective and capacity (e.g., detecting climate change). I have briefly in- synergistic effect. In International Conference on troduced a paradigm to best model organizations dwelling Collaboration Technologies and Systems, 2015. on socially connected networks. This paradigm is a collec- [8] P. Bonacich and P. Lu. Introduction to mathematical tion of principles, layouts, and interaction protocols that sociology. Princeton, NJ: Princeton University Press, obviate the network nature of group activity as an organi- 2012. zation. The salient properties that set an NOP apart from [9] S. P. Borgatti and P. C. Foster. The network paradigm other organizational paradigms are: a. Openness, b. Evolv- in organizational research: A review and typology. ing structure, c. Selfish allegiances and community social Journal of management, 29(6):991–1013, 2003. power, and d. Impromptu network topology. [10] L. Chen, G. G. Gable, and H. Hu. Communication Given the volatility of networks, an NOP will allow for and organizational social networks: a simulation rapid depiction and analysis of emerging and evolving net- model. Computational and Mathematical Organization worked organizations witnessed in our connected world. An Theory, 19(4):460–479, 2013. NOP has introduced modular components capturing essen- [11] H. R. Ekbia and R. Kling. Network organizations: tial units to be modularly combined to define NOs. An NOP Symmetric cooperation or multivalent negotiation? replicates many properties and features of virtual working The Information Society, 21(3):155–168, 2005. groups. A specific salient phenomenon is how working to- gether in networks affects their individual as well as col- [12] F. Heider. Attitudes and cognitive organization. The lective productivities. Synergy is one of the main types of Journal of Psychology, 21(1):107–112, 1946. network effects featured in our paradigm to enhanced per- [13] B. Horling and V. Lesser. A survey of multi-agent formance of agents and the organization. organizational paradigms. The Knowledge Engineering Our plans include analyses of naturally occurring network Review, 19(4):281–316, 2004. organizations that illustrate principles indicated in our pro- [14] D. Hovorka and K. Larsen. Enabling agile adoption posed paradigm as well as designs for novel applications practices through network organizations. European that illustrate flexibility of our modular paradigm. We have Journal of Information Systems, 15:159–168, 2006. shown by a case study that the NO paradigm is applica- [15] S. Liemhetcharat and M. Veloso. Weighted synergy ble for modeling real world organizations [3]. An extended graphs for effective team formation with work will cover more details and applications that corrob- heterogeneous ad hoc agents. Artificial Intelligence, orate tenets of NOs in settings such as Net-centric warfare 208:41 – 65, 2014. as well as grid-based disaster responses. Of particular in- [16] J. Parker, E. Nunes, J. Godoy, and M. Gini. Forming terest are the potential issues arising from scaling NOs to long term teams to exploit synergies among medium and large organizations, and augmenting generic heterogeneous agents. Technical report, University of Minnesota, Department of Computer Science and influence in organizations: An integrated approach to Engineering., 2012. knowledge adoption. Information Systems Research, [17] S. K. Shin and W. Kook. Can knowledge be more 14(1):47–65., 2003. accessible in a virtual network?: Collective dynamics [20] M. van Alstyne. The state of network organization: A of knowledge transfer in a virtual knowledge survey in three frameworks. Journal of Organizational organization network. Decision Support Systems, Computing and Electronic Commerce, 7(2-3):83–151, 59:180 – 189, 2014. 1997. [18] C. C. Snow and Ø. D. Fjeldstad. Network paradigm: [21] J. Vázquez-Salceda, V. Dignum, and F. Dignum. Applications in organizational science. In J. D. Organizing multiagent systems. Autonomous Agents Wright, editor, International Encyclopedia of the and Multi-Agent Systems, 11(3):307–360, 2005. Social and Behavioral Sciences, pages 546 – 550. [22] W. Zhou, W. Duan, and S. Piramuthu. A social Elsevier, Oxford, second edition, 2015. network matrix for implicit and explicit social network [19] S. W. Sussman and W. S. Siegal. Informational plates. Decision Support Systems, 68:89 – 97, 2014.