Movatterモバイル変換

Abstract

Organizational routines are collective phenomena involvingmultiple individual actors. They are crucial in helping to understandhow organizations behave and change in a certain period. In this paper,by regarding the individual habits of multiple actors involved asfundamental building blocks, we consider organizational routines froman 'emergence-based' perspective. We emphasise the impacts ofconnections or network topologies among individual actors in theformation of organizational routines, and carry out a multi-agent basedsimulation analysis of organizational routines on complex networks. Weconsider some important factors such as inertia resulted fromindividual memories, component complexity of organizational tasks,turnover of individual actors, the impacts of both heterogeneity andimprovisation of individual actors involved, and the dynamicalproperties of the network topologies within which individual actors arelocated. The results of our research show that network topologies amongindividual actors do determine the dynamic characteristics oforganizational routines. Although the fact is that the mechanismsbeneath this are also influenced by some main factors like the memorycapacity of individual actors and the component complexity oforganizational tasks that these individual actors should deal withrepetitively, and that the total costs for the organization to bearduring their implementation of organizational tasks are variant, theroutine system on scale-free networks can always have a betterperformance, and obtain a much higher coherency and routinization levelof collective behaviours, even in the case of turnover of individualactors. In addition, when individual actors involved are heterogeneous,the routine system on scale-free networks would also exhibit a stronganti-disturbance ability, no matter whether there are minorimprovisations from these individual actors or not. Nevertheless, alarge number of improvisations enable individual actors to act in somemore individualistic manners, and destroy the routine system as aresult.

Keywords:

Organizational Routines, Connections, Complex Networks,Multiple Actors, Individual Habits, Multi-Agent Based Simulation

Introduction

1.1

Recent research indicate that routines play a crucial rolein helping to explain the behaviour of organizations, and are key tounderstanding how organizations accomplish their work. According toFeldman and Pentland (2003),organizational routine is defined as some 'recognizable, repetitivepatterns of interdependent actions, carried out by multiple actors'.Many scholars have yet focused on the role of these multiple individualactors (Abell et al. 2008;Feldman & Pentland 2003),their psychological properties such as procedural and declarativememories (Cohen & Bacdayan1994;Lazaric 2008),motivations and incentives (Grave 2008),experiences (Turner & Fern2012), etc., during the performance of a routine. Becker (2004) and Hodgson (2008) argued that there is anessential relationship between individual actors and the collectiveroutines they are involved in. Organizational routines emerge frominteractions between individual actors (Bapujiet al. 2012). They are not just simply followed orreproduced; rather, individual actors 'have choices between whether todo so, or whether to amend the routine' (Becker2004,2005;Feldman 2000;Feldman & Pentland 2003).However, individual actors involved in organizational routines may belocated in different places and belong to different organizationalunits. Their acts depend on their own knowledge contexts andcapabilities, as well as the roles they play while performingorganizational routines (Becker 2004;Feldman & Pentland 2003).In other words, all of these individual actors involved inorganizational routines are distributed and heterogeneous, connectingwith each other through communications.

1.2

Feldman and Rafaeli (2002)and Turner and Rindova (2012)pointed out that the connections formedin the process of performing organizational routines – which enable thetransfer of information among individual actors – contributed to thedevelopment of common understandings and agreements regarding theperformance of routines in practice; this in turn contributed to boththe stability and adaptability of organizational routines. However,relatively limited attention has been paid to the question of how theseconnections among individual actors affect routine performances.

1.3

Miller et al. (2012)presented multi-agent based simulation as a method of modellingmicro-foundations of organizational routines. Their work identifiedthree kinds of memory – procedural, declarative, and transactive – toexplain routine dynamics. However, they did not consider theconnections or network topologies among individual actors which, fromthe perspective of complex networks theory (Barabâsiet al. 1999;Dorogovtsev& Mendes 2003;Watts1999), play a crucial role in determining the dynamicproperties of routine systems.

1.4

In this paper, we appraise organizational routines througha 'bottom-up' approach, and investigate their evolutionary trajectorieson complex networks. By regarding the habitual behaviours of individualactors (namely, the individual habits of multiple actors involved inorganizational routines) as 'fundamental, individual-level buildingblocks' (Hodgson 2008;Knudsen 2008;Gao et al. 2014), we considerroutines as spontaneous emergences from repeated interactions thatindividual actors engage in on an 'increasingly habitual basis' (Witt 2011), and build up amulti-agent based simulation model to formulate this process. Someimportant factors such as inertia resulted from individual memories,component complexity of organizational tasks, and turnover of theindividual actors. These factors, as well as the impacts of bothheterogeneity and improvisation of the individual actors involved, andthe dynamical properties of the network topologies within whichindividual actors are located, are all considered in our simulationmodel.

1.5

From the simulation results of variant concrete situations,we conclude that network topologies among individual actors dodetermine the dynamic characteristics of organizational routines. Weobserved that for all the different combinations of both the memorycapacity of individual actors and the component complexity oforganizational tasks that these individual actors need to deal withrepetitively, the total costs that the organization bears for theperformance of organizational tasks vary. However, the routine systemon scale-free networks can always perform better, and obtain a muchhigher coherency and routinization level of the organizationalroutines, even in the case of individual actors' turnover. In addition,we found that when individual actors involved are heterogeneous, theroutine system on scale-free networks exhibit a strong anti-disturbanceability and achieve a relatively stable, higher routinization level ofcollective behaviours over time, regardless of whether or not there areminor improvisations from these individual actors. Nevertheless, alarge number of improvisations enable individual actors to act in moreindividualistic manners, which may destroy the routine system. Thus, abetter balance is required for the emergence of a routine systemthrough all the levels of the crucial factors previously mentioned.

1.6

The rest of this paper is structured as follows. In thesecond section, we provide a brief overview of literatures related toour work. Based on a summary of the relationship between the two mainaspects of organizational routines, the ostensive and the performative,we take individual habits as the basic units of analysis, and discusssome of the characteristics in network topologies inherited from theindividual actors involved in organizational routines. In the thirdsection, we build up a multi-agent based simulation of organizationalroutines through a 'bottom-up' approach. Through an analogy between theindividual actors involved in organizational routines and a specifictype of active agents, we offer a multi-agent based simulation modelvia Swarm package (Swarm developmentGroup [SDG] 2000), and achieve the visualization andreproduction of both the ostensive and performative aspects oforganizational routines with the aid of computer tools. Next, wepresent the simulation results for variant concrete situations, anddiscuss factors such as the network topologies of the individualactors, their memory capacity and the component complexity oforganizational tasks, and both the turnover and improvisations fromindividual actors involved, that do have impacts on the evolution oforganizational routines. The last section includes our conclusion anddiscussion.

Theoreticalfoundations

The ostensive and performative aspects of organizationalroutines

2.1

As collective phenomena that involve multiple individualactors, or the generative system conceptualised by Feldman and Pentland(2003), it has beenwidely accepted that there are mainly two aspects in organizationalroutines (Feldman &Pentland 2003;Pentland& Feldman 2005). The first is the ostensive aspectthat shapes our perception of what the term 'organizational routine'should refer to. It constitutes sets of particular, overlappingnarratives or functional events which refer to the 'abstract patterns'of action, or the 'ideal/schematic form' of the routines, and providesroad maps for individual participants to carry out their organizationaltasks (Pentland & Feldman2007,2008).The second is the performative aspect that represents the specific(actual) performances taken by specific people, at specific times, inspecific places (Pentland & Feldman, 2005). Pentland andFeldman (2005,2008) argued that theseperformances are particular sequences of events experienced by theindividual actors involved in the routines.

2.2

However, there is a recursive relationship between the twoaspects. On one hand, the specific performances of individual actorscreate and recreate the ostensive aspects (namely, the narrativenetworks) of the routines. On the other hand, the ostensive aspectconstrains and enables the performances in practice. These two aspectsmutually constitute an organizational routine as a whole (Feldman & Pentland 2003),as shown in Figure1.

Figure 1. Internal structure oforganizational routines

2.3

Here, we define the term 'narrative network' as acollection of functional events that relate to each other through'their sequential occurrence in a story or set of stories' (Hendricks 1972,1973). Pentland and Feldman (2007,2008) indicated that thenarratives as the basic nodes within a narrative network are indeedsome fragments of stories, while connections between these fragments(namely, ties with the narrative network) imply their sequence orrelationship in occurrence. As a whole, the narrative network can beregarded as 'an explicit representation' of the ostensive aspect of anorganizational routine (Pentland& Feldman 2008).

2.4

In addition, there are individual actors involved in everynarrative, ranging from a few to several hundreds or even more. Theseindividual actors may work together or communicate with each other,sharing information and knowledge about their organizational tasks. Putsimply, individual actors involved in narratives are interconnectedthrough certain relationships.

From individual habits to organizational routines

2.5

Organizational routines emerge from repeated interactionsbetween individual actors engaged in the implementation oforganizational tasks. Becker (2004)argued that the personal behaviours of these individual actors areinherently habitual. That is, they have tendencies to behave or thinkin a particular way 'in a particular class of situations', and 'repeatthe same act in similar material conditions' (Felinet al. 2012;Hodgson 2008).However, such a tendency 'may, or may not, be expressed in currentbehaviour or thought', depending on the presence of an appropriatestimulus or context (Knudsen 2008).That is, a habit is not necessarily used all the time, but may betriggered by a certain cue. In this sense, we can view a habit as 'asubset of actual behaviours or thoughts that realised from a larger setof possible behaviours or thoughts in a particular class of situations'(Hodgson 2004;Knudsen 2008).

2.6

Knudsen (2008)and Bapuji et al. (2012)considered habits of individual participants as the fundamentalbuilding blocks from which organizational routines emerge, whileorganizational routines are special structures that operate through thetriggering of these interlocking individual habits (Hodgson 2008). Weick's (1990) work of the documentationof a historical airplane disaster showed that organizational routinesmay be disrupted when all the individual actors involved act in moreindividualistic – rather than collective – manners. Thus, Becker (2004) pointed out that a finebalance 'between individual habits of the actors and organizationalroutines' needs to be maintained so as to avoid the breakdown oforganizational coordination.

2.7

However, these habitual behaviours of individual actors arenot always stable. Witt (2011)argued that as both the factual and procedural knowledge of routines isshared and its meaning is often socially constructed, individual actorsinvolved in organizational tasks are not entirely autonomous; they arealso influenced by their nearest neighbours and/or the environment inwhich they are located. During the repetition of actions, even a verysmall novelty may lead to the development of another absolutelydifferent pattern of behaviours among the individual actors, and resultin the emergence of new routines at the organizational level (Hodgson 2008;Witt 2011).

Network topologies for individual actors involved in theroutines

2.8

Hodgson (2008)argued that individual habits are not developed in isolation, butemerge within social life. That is, individual actors may interact witheach other through certain relationships in the performance oforganizational tasks. If we use the nodes to represent the distributedand heterogeneous actors involved in organizational routines, and theedges connecting the nodes to refer to the connection betweencorresponding actors, then the connectivity structure of the individualactors can be characterised by a number of topological properties ofnetworks which play a fundamental role in determining the systems'dynamic properties (Barabâsi etal. 1999;Dorogovtsev& Mendes 2003;Watts1999). Typically, two measures are adopted to characterisestructural properties of the networks. One is the average path length,measuring the average separation between two nodes in the graph; theother is the clustering coefficient, which indicates the probabilitythat two nearest neighbours of a node are also nearest neighbours ofeach other.

2.9

Four typical but different network topologies areconsidered in this paper, namely, regular or lattices, random networks,and small-world and scale-free graphs. These different networktopologies may share the same number of individual actors (i.e. nodes),but differ in terms of the pattern of connections.

2.10

Regular: A regular network is defined asa nearest-neighbour coupled network in which every node is only joinedby a few of its nearest neighbours, and the distances of theseconnections do not exceed a given value. This type of network topologyis clustered and covers long average path lengths. In this paper, a 2-Dlattice is used.

2.11

Random: The random network is placedat the opposite end of the spectrum from the regular network, whereconnections between nodes are not limited to the immediate neighbours,but randomly drawn from the entire population (Erdös& Rényi 1959). Thus, the nodes own short average pathlengths and are not clustered.

2.12

Small-world: A small-world network canbe created from a regular lattice by randomly rewiring each of thelinks between individual actors with some probabilityp(Watts & Stogatz 1998).The rewiring aims at substituting some short-range connections withlong-range ones, which may drastically reduce the average distancebetween randomly chosen pairs of nodes, and produce a small-worldphenomenon characteristic to many social networks (Milgram 1967). Small-worldnetworks are highly clustered and have relatively short minimum pathlengths.

2.13

Scale-free:A scale-free network ischaracterised by its connectivity distribution, which is in a power-lawformκ^−r,independent of the network scale (Barabâsiet al. 1999). This power-law distribution, in comparison withthe familiar Poisson distribution obeyed by connection sequences ofboth the random and small-world networks mentioned previously, fallsoff more gradually, allowing for 'a few nodes of very large number oflinks' to exist (Wang & Chen2003), but most nodes have 'only a few incoming or outgoinglinks' (Jeong 2003).Because of this, we affirm that the scale-free network owns a smalleraverage path length and a typically larger clustering coefficient.

Developmentof the multi-agent based simulation model

Design of the agents

3.1

Our simulation model comprises four kinds of agents, namelyactions, individual actors, narratives, and the environment, asdiscussed further.

3.2

Actions: In this paper, we regardactions involved in organizational routines as particular kinds ofagents that provide particular ways for individual actors to performtheir organizational tasks. For every performance of the routines, someof these actions are selected and adopted by individual actors todispose off their organizational tasks. We assume that these actionagents possess at least three of the following basic attributes.

1. actionID: A unique ID number fordefinite identification of each action agent.
2. actionType: The coefficientactionTypeis used for discriminating the performance outputs between differentaction agents. In our simulation model, however, we just assign thisattribute of the action agents with some constant values which refer tocertain narratives that may occur subsequent to the present ones andindicate the event sequences of the stories called organizationalroutines.
3. actionCost: The performance ofactions related to organizational tasks involves essential expensessuch as material and human resources, money, time, and so on. Here, weonly take the coefficientactionCost to representall the expenses borne for each action performed in the routines.

3.3

Individual actors: Based on thediscussion in the section entitled 'From individual habits toorganizational routines', we suppose that each individual actor agentin our simulation model owns a unique ID number (actorIDNum)for definite identification and the following four basic attributes:

1. A general goal of actions (G).All the individual actor agents involved in the organization have ageneral goal of actions (G). Forexample, under the hypothesis of limited rationality (Simon 1945), it may be expectedfor their actions to cost minimum possible expenses. This general goal (G)determines the actions chosen for individual actors from their possibleaction sets each time when performing their organizational tasks.
2. A set of possible actions (S).As there are possibly various but limited ways for individual actors toperform their organizational tasks, we use a setSto represent the collection of possible actions that individual actorsmay or may not choose in a particular class of contexts.

   S = {action1,action2, …,actionn}

   (1)

   Here, the size of the setS– i.e.n in formula (1) – refers to some capacitiesof individual actors, and each of the elementaction_i(i = 1, 2, …,n) represents acertain action agent that indicates a particular way for individualactors to handle their organizational tasks.

3. A list of habits (H).For treating individual habits as dispositions, we use a specific setof conditional probabilities (H)– the probability that 'a particular behaviour will occur in responseto a stable contextual cue' (Knudsen2008) – to explicitly express the concept 'individual habits'in our simulation model.

   H = {p1,p2,…,pn}

   (2)

   The coefficientsp_irefer to the conditional probability for individual actors to take acertain actionaction_i astheir repetitive actions in response to certain stimuli (Ωs) fromenvironments:

   pi=p(actioni|Ω),   (i= 1, 2,…,n)

   (3)

4. A list of memory (M).Miller et al. (2012)emphasised the importance of (individual) memory in both the formationand persistence of routines, and argued that 'over time, as individualagents discover and remember successful actions', the need to searchmay be reduced or eliminated, thereby improving efficiency and'producing recognizable, repeated problem-solving patterns'. In thispaper, we use the 'first-in-first-out' (FIFO) stackwith a fixed length (memoryLength) to representmemory of individual actors (M).During each repetition, some actions occur in accordance with theindividual actors' list of habits (H).These actual actions are then added to the memory list (M),which would affect the probability values of the habits (H)in the following performances. If we take the variableperformNum_ito denote the actual number of occurrences for each action agentaction_istored in the memory list (i = 1, 2, …,n),then we can observe the relation that:

   (4)

3.4

Figure2 provides asketch of the primary interactions between these four main attributesof the individual actor agent.

Figure 2. Internal structure ofthe 'individual actor' agent

3.5

Figure2 shows thefollowing: every individual actor agent responds to the particularsituations of the environment in accordance with his/her goal ofminimizing the performance expenses of organizational tasks. That is,individual actor agents are expected to seek and/or imitate theirneighbours to derive a more cost-effective approach. As a result, acertain action may be selected from the possible action setS,and be adopted to cope with the work that the individual actor agentsare entrusted with. Then, within a stable context, these individualactor agents repeat particular actions, and gradually evolve their ownhabits through the reinforcement of memory.

Narratives

3.6

Narratives are considered as a type of agent thatrepresents some cognitive regularities and/or expectations individualactors use to 'guide, account for, and refer to specific performancesof a routine' (Feldman &Pentland 2003;Pentland& Feldman 2005). According to Pentland and Feldman (2007), each narrative agentcomprises several individual actors and certain kinds of actions thatoccur with them. The following are some main attributes:

1. narrativeID: a unique ID numberfor definite identification of each narrative agent.
2. A list of individual actor agents(actorList):we use a set namedactorList to represent acollection of individual actor agents that are involved in eachnarrative. All these individual actor agents are distributed on a2-Dimension Raster, interacting with each other and performing thegiven organizational tasks together.
3. A list of action agents (actionList):ThisactionList in the simulation model denotes acollection of action agents that are possibly selected by individualactor agents to process their organizational tasks. The size of thiscollection of action agents (actionNum) indicatesthe number of different approaches that these actor agents arepermitted to adopt, which Wood (1986)labelled in his work as 'component complexity' of the organizationaltasks.
   We observe that every individual actor agent's set of possible actionsare indeed a subset of the collection of action agentsactionListof the narrative agent that he/she belongs to. That is,n≤actionNum, andS ⊆actionList.
4. A cost coefficient (narrativeCost):We use a variablenarrativeCost to evaluate thenarrative agent's performance, which denotes the total expenses thatall the individual actor agents should expend at a time.

3.7

Further, we use a first-order Markov matrix (T)(Pentland et al. 2010)to describe explicitly sequential adjacencies between these differentnarrative agents, as shown in the following figure:

(5)

3.8

Here, the subscript variablem informula (5) denotes the number of narratives that are involved in theroutine. The element w_ijindicates the connection probability from narrativeito narrativej, which numerically equals theproportion of individual actor agents engaged in narrativeithat adopt the action related to narrativej astheir current approach in performing the organizational tasks (1 ≤i,j ≤m). Thus, the values ofthese coefficients satisfy the following conditions:

(6)

3.9

The matrixT shows a specific structureof the narrative network, and gives us an integral description of thepossible frequencies of event-sequences (that is, transitions from somenarratives to others) that exist in practice. Based on Pentland's (2003) fundamental work, weuse the Euclidean distance (D) of the observedaction network from a uniform, random network for describing thevariation of the actual behavioural sequences during performance of theroutines:

(7)

Here, the variableD's range of valueis 0 to (m − 1), demonstrating the routinizationlevel of organizational behaviours. That is, larger values of thevariableD may reflect less variety of thenarrative network, and vice versa.

The environment

3.10

Cohendet and Llerena (2003)and Howard-Grenville (2005)pointed out that organizational routines are essentiallycontext-dependent, and that their performances are often embedded inspecific environments. It is the environment that offers a specificcontext within which individual actor agents are located and interactwith each other. Although it has been widely recognised that changes inenvironment are common occurrences for today's organizations, and thereare a few literatures on organizational routines in dynamicenvironments (Gilbert 2005;Grote 2009;Howard-Grenville2005), we still know little about micro-mechanisms underlyingthe question of how individual actors' understanding and performance ofthe routines are affected by the organizational environment (Kozica et al. 2014).

3.11

In this paper, we assume that the environment is relativelystable during the simulation process, and treat it as a constant agentthat provides fixed conditions for individual actors to act andinteract with each other in performing their organizational tasks. Thatis, for each simulation tick, some specific stimuli are alwaysexistent, repetitively invoking related action agents for individualactors to perform their organizational tasks.

Variation activities of actions in performance of theorganizational tasks

3.12

Some literature from both empirical and simulation research(Aarts & Dijksterhuis 2000;Cohen & Bacdayan 1994;Hodgson & Knudsen 2004;Wood & Neal 2007;Yin et al. 2004) haveshown that it is repetition that immediately causes the formation ofindividual habits. With repetition, the control of behaviour of all theindividual actors involved in organizational routines may shift 'fromcognitive mechanisms to an automatic mode' (Knudsen2008), and a persistent habit gradually emerges.

3.13

Nevertheless, in the repeated performance of organizationaltasks, individual actors may vary their habitual behaviours to achievetheir primary goals (i.e. minimizing the operational expenses of theiractions). First, when confronted with new organizational tasks,individual actors with limited capacities have powerful incentives tosearch (from the set of possible actions) for new alternatives with lowexpense to substitute their current actions (Hodgson2008;Knudsen 2008).However, when all operations function well, this search foralternatives will stop, and 'the current way of doing things isrepeated' (Grave 2008;Winter 2003). In addition,immediate or delayed rewards (in our example, referring to the lesserexpense that individual actors must be paid) may reinforce therepetition of the given actions (Knudsen2008;Postrel &Rumelt 1996). In our simulation model, we take everysimulation tick to represent a certain performance of the routine, andformulate this mechanism as follows: at each simulation tick,individual actor agents decide whether to search for new alternativesrandomly. If the searching activity does occur, the individual actoragent searches from his/her collection of possible actions for anotherdifferent action with a much lesser expense than the current one (ifthis action exists, of course) as his/her new alternative in the nextperformance. Otherwise, there is nothing the individual actor agentshould do.

3.14

Second, individual habits may be replicated throughimitation (Hodgson 2008).The fact is that within a relatively stable context, individual actorsmay imitate some particular actions of their neighbours eitherconsciously (e.g. duplicating their neighbour's action of lesserexpense) or unconsciously (e.g. reproducing an action that is popularamong their neighbours). Hence, some specific actions are propagatedand reinforced by recurrent performances as this can intensifyfamiliarities, and the habits giving rise to them are replicated aswell (Schulz 2008). As aresult, these imitation activities may bring about changes of habits (H)as well as the memory (M) ofindividual actors, or even mutation of organizational routines. Werepresent this mechanism in our simulation model as follows: at eachsimulation tick, an individual actor agent decides whether to imitatehis/her neighbours in an apparently random manner. If the answer is'Yes', then the individual actor agent selects one of his/her neighbouragents randomly, and reproduce that neighbour's current action as thealternative in the next iteration if it exhausted a lesser expense thanhis/her own.

3.15

Figure3 gives anoverview of the procedure that an individual actor agent goes throughin the completion of his/her variation activities.

Figure 3. Flow chart for actionsearching and/or imitation activities

3.16

However, the imitation activities of the individual actoragents are often also influenced by the behaviour of their neighbours.That is, as previously mentioned, individual actor agents are inclinedto adopt the actions that are popular with their neighbour agents, aswell. To depict this situation, for each individual actor agentactor_i(i = 1, 2, …,actorNum, wherethe parameteractorNum denotes the total number ofactor agents that participated in each of the narrative agents), we letcActionCost_i denote theexpense that he/she should undertake in performing the current action (actionCost)andneighborList_i to denotethe collection of his/her neighbour agents, and define a cost functionfor this individual actor agent in the performance of organizationaltasks, as shown in the following:

(8)

3.17

Here, the coefficientδ_iis a penalty coefficient, representing the degree of differencesbetween the current action of the individual actor agentactor_iand that of their neighbour agents. In this paper, we assume that thevalue ofδ_iis equal to the number of neighbour agents fromneighborList_iwho currently adopt actions different from the one that is employed bythe individual actor agent actor_i(i = 1, 2, …,actorNum).

3.18

Then, for each narrative agentnarrative_j(j = 1,2, …,m),we describe the attribute parameternarrativeCostas follows:

(9)

3.19

Here, the coefficientactorList_jrepresents the collection of individual actor agents participating inthe narrative agentnarrative_j(j = 1, 2, …,m).

3.20

Thus, we use a variabletotalCost todescribe the performance efficiency of the routine, which equals thesum of all the values of the coefficientnarrativeCostof the narrative agents:

(10)

Construction of the simulation model via Swarm package

3.21

The multi-agent based simulation model was constructedusing Swarm package (SwarmDevelopment Group [SDG] 2000). The code is available athttps://www.openabm.org/model/4468/version/2/view

3.22

The Unit Testing method is used for verification of thesimulation model. North and Macal (2007)argued that 'it can be a powerful way to test and retest code'. Byconducting tests first on each function or method, then on each module,then on each area, and finally on the entire system, we make sure thatthe specifications are complete and that there are no mistakes made inthe construction of the multi-agent based simulation model. Inaddition, the Specific Scenarios Testing method, which Gilbert (2008) viewed as a very weakyet effective way for verification and validation of multi-agent basedsimulation, is also used for validation of the simulation model.Through the test and analysis of a series of simple scenarios, weaffirm to some extent that our simulation model is valid.

3.23

We suppose that there are 12 narrative agents involved inthe routine (m = 12), each consisting of a number ofaction agents and individual actor agents, as well. All the individualactor agents participating in a narrative are distributed on a2-dimension lattice. For analysis of the organizational routine oncomplex networks, two classical models from the complex networks theoryare employed in our study. The first is the WS model (Watts & Stogatz 1998),which is adopted to generate the regular lattice, random network, andthe small-world by setting the rewiring probabilitypequals to 0, 1 and a very small value in the interval (0, 1),separately. The second is the BA model (Barabâsiet al. 1999;Barabâsi& Bonabeau 2003), which is adopted to generate thescale-free network via two important mechanisms (i.e. random growth andpreferential attachment). We started with an initial set of three fullyconnected nodes (namely, the coefficient in theBAmodeln₀ = 3), and let thedegree of the new added vertices equal 3 (that is, the coefficient intheBA modelm₀= 3). As a result, we obtained the four typical but different kinds ofnetwork topologies shown in Figure4.

Figure 4. Four typicalconnections/network topologies used in simulation

Simulationresults

4.1

We discuss our simulation results in four stages. First, wecompare the evolutionary trajectories of the organizational routine onthe four typical but different kinds of network topologies mentioned inthe section entitled 'Network topologies for individual actors involvedin the routines' (namely, regular lattice, random, and small-world andscale-free networks, as shown in Figure4).

4.2

Second, we identify two important factors that affect theformation of organizational routines. The first is the memory capacityof individual actors, which we represent by the length of the memorylist (memoryLength). The other is the componentcomplexity of organizational tasks, represented by the number ofactions that need to be executed in the performance of the routine (actionNum).

4.3

Third, in order to analyse the influence of turnover onindividual actors, we suppose that at every simulation tick, individualactor agents may leave the organization randomly with a givenprobabilityturnover_P (0 ≤turnover_P≤ 1). However, to facilitate easy discussion, we assume that the totalnumber of individual actors involved in the routine remains unchangedduring the simulation process. That is, new participants would join theorganization, filling in locations of the individual actor agents whohave left.

4.4

Fourth, we consider the role of both heterogeneity andimprovisation of individual actors in the performance of the routines.First, when individual actors involved in organizational routines areheterogeneous, it means that all the individual actors possess variantmemory capacities (that is, with different values of the coefficientmemoryLength),and deal with organizational tasks of different component complexities(that is, with different values of the coefficientactionNum).Second, we discuss the impact of improvisation from individual actors,based on the fact that organizational practice is inherentlyimprovisatory, with plenty of scope for adjustments and variations that'make it possible to get things done in diverse situations' (Pentland & Feldman 2005,2007).

4.5

(1) For the former two stages, we assume that the number ofindividual actor agents (actorNum) that participatedin every narrative agent is the same (in this paper, according to theresults of sensitivity analysis technique about input parameters of thesimulation model, we letactorNum = 36), and thatboth the coefficients ofmemoryLength andactionNumare constants. This means that while individual actor agents possessthe same level of memory capacities, all the narrative agents own thesame number of possible actions. With default settings as shown inTable1 (which we callScenario I), we vary the value ofmemoryLength from1 to 6, and that ofactionNum from 3 to 8,respectively, and execute the simulation model on all of the four typesof network topologies, each with 120 simulation ticks (representing 120performances of the routine, iteratively), successively, arriving atthe simulation results shown in Figure5.

Table 1:Overviewof model parameters (Scenario I)
Variable	Default settings	Description
actorNum	36	Number of individual actor agents that participatedin each narrative agent.
actionNum	3, 8	Number of action agents involved in each narrativeagent (the size of the possible action setS).
m	12	Number of narrative agents constituting the routine.
actionType	U(1,narrativeNum)	This attributeactionTyperefers to certain narratives that may occur subsequent to the presentones. Its value is randomly assigned with a uniform distributionU(1,narrativeNum).
actionCost	U(1,9)	Expense that should be paid when the relatedactions are being undertaken. Its value is randomly assigned with auniform distributionU(1, 9).
memoryLength	1, 6	Length of the memory list (M),referring to the memory capacity of individual actor agents.
Note: underlined values aredefault settings of the coefficients

(a) The total number of individual actor agentssearching for new alternatives

(b) The total number of individual actor agentsthat imitate their neighbours

(c) Values of the coefficienttotalCost

(d) Values of the coefficientD

Figure 5. Simulation resultswith default settings for Scenario I

4.6

Based on the simulation results shown in Figure5, we can state that networktopologies among individual actors do impact the formation process ofthe routine at the organizational level. However, these impacts areclosely influenced by both the memory capacity of individual actorsthat were involved in the routine and the component complexity of theorganizational tasks to be performed. First, when confronted withorganizational tasks of lower component complexity (actionNum= 3), there being less alternatives to adopt, the total numbers ofindividual actors both searching for and imitating activities reducesto zero quickly, and the routine system goes into a stable state. Thereare no significant differences between the routine systems on differentnetwork topologies, as shown in the upper-left portion of Figure5 (a) and (b). Nevertheless, asindividual memory and cognition may result in inertia among actorsduring their performance of the organizational tasks, individual actorswith a higher level of memory capacity (memoryLength= 6), would experience this convergent tendency for a much longerperiod of time, relatively. The routine system on random networks isespecially more reflective of this case, followed by those on regularlattice. The routine systems on both small-world and scale-freenetworks require a much shorter period of time for converging to astable state, as shown in the bottom-left portion of Figure5 (a) and (b).

4.7

Second, when individual actors are dealing withorganizational tasks of higher component complexity (actionNum= 8), they have to struggle between inertia of memories and some morefeasible alternatives. Thus, the total numbers of individual actorsperforming both searching and imitation activities fluctuate with time,and the routine system would gradually maintain relatively dynamicstability. However, there are differences between organizationalroutines on different network topologies. On one hand, for individualactors with lower levels of memory capacity (memoryLength= 1), the routine system on small-world networks exhibits the mostobvious fluctuations, as shown in the upper-right portion of Figure5 (a) and (b). On the other hand, forindividual actors with higher levels of memory capacity (memoryLength= 6), it is the routine system on random networks, as shown in thebottom-right portion of Figure5(a) and (b).

4.8

Third, regarding the influences of network topologies onthe total costs that the organization should undertake for performingthe organizational tasks, we can assert from Figure5(c) the following: (i) For organizational tasks of lower componentcomplexity (actionNum = 3), there being lessalternatives for individual actors to adopt, the routine system onsmall-world networks has to pay a minimum expense, followed by theroutine system on both regular and scale-free networks; the routinesystem on random networks bears the maximum, as shown in the leftportion of Figure5 (c). (ii)For individual actors handling organizational tasks with highercomponent complexity (actionNum = 8), the routinesystem on scale-free networks pays the most, followed by that on bothrandom and small-world networks, and finally followed by that onregular lattices, as shown in the right portion of Figure5 (c). (iii) When the value of thecoefficientmemoryLength varies from 3 to 6, thevalue of the coefficienttotalCost fluctuates witha large amplitude due to inertia among individual actors, which resultsfrom the increase of memory capacity. However, the routine system onscale-free networks shows a much stronger sensitivity of thecoefficientmemoryLength, which depicts the memorycapacity level of all the individual actors involved in the routine, asshown in the bottom portion of Figure5(c).

4.9

Fourth, Figure5(d) shows that for all the four different situations ( relating to thefour different combinations of the two coefficientsmemoryLengthandactionNum), the routine system on scale-freenetworks owns the highest value of the coefficientD(representing a much more coherent routinization level of collectivebehaviours at the organizational level), followed by the routine systemon both the regular and random networks. Nevertheless, the routinesystem on small-world networks displays more sensitivity to theconcrete situations, as shown in the bottom-right portion of Figure5 (d).

4.10

(2) In the third stage, for analysing the influence ofturnover of individual actors during the formation process of theroutine, we let the turnover probabilityturnover_P= 0.001, keeping all the other input parameters settings the same asthose for Scenario I; we called this Scenario II. Then, we executed themulti-agent based simulation model on all of the four types of networktopologies, each with 120 simulation ticks (representing 120performances of the routine, iteratively), successively, and obtainedthe simulation results shown in Figure6.

(a) The total number of individual actor agentsthat searching for new alternatives

(b) The total number of individual actor agentsthat imitate their neighbours

(c) Values of the coefficienttotalCost

(d) Values of the coefficientD

Figure 6. Simulation resultswith default settings for Scenario II

4.11

We affirmed that the turnover of individual actors wouldaggravate the fluctuation of the routine system on all four differentnetwork topologies. Furthermore, as shown in Figure6,although there are subtle differences between combinations of differentvalues of both the coefficientsmemoryLength andactionNum,we were still able to derive some general patterns from the simulationresults. That is, compared to the vibration of the routine system onthe other three kinds of networks (namely, the regular, random, andsmall-world networks), the routine system on scale-free networks tendsto converge over time and eventually reaches a stable state.Essentially, it can more effectively cope with fluctuations resultingfrom turnover of individual actors. Furthermore, inertia as a result ofmemory capacity of individual actors exacerbates this phenomenon, asshown in Figure6 (a) and (b).

4.12

Further, with the simulation continuing, the routine systemon scale-free networks owns a relatively lower value of the coefficienttotalCost (with an exception only of thesituation whenmemoryLength = 1 andactionNum= 8, as shown in the upper-right portion of Figure6(c), where there are no significant differences between the routinesystems on the four distinct network topologies), but a higher value ofthe coefficientD. This means that the routinesystem on scale-free networks is easier to achieve at a much morecoherent routinization level of collective behaviours at theorganizational level, and the expenses are relatively minimal.

4.13

(3) In the fourth stage, we differentiated the initialvalues of both the coefficientsmemoryLength andactionNumfor each of the individual actor agents, to consider the role ofheterogeneity of individual actors in performance of the routines.Then, to discuss the impact of improvisation from individual actors, weintroduced a coefficientimprovi_P (0 ≤improvi_P≤ 1) to describe the improvisation activity probability of individualactor agents. When improvisation occurs, the particular courses ofaction (namely, the way that organizational tasks are executed) whichindividual actor agents choose in performance of the organizationaltasks are not simply followed and/or reproduced, but are substituted byabsolutely new ones that they either created randomly or by imitatingtheir neighbours, regardless of the resulting costs.

4.14

At the beginning of the simulation, we assigned thecoefficientmemoryLength for every individual actoragent randomly with a uniform distributionU(1, 9),and assigned the coefficientactionNum randomly aswell, with a uniform distributionU(2, 8). We letthe value of the coefficientimprovi_P equal 0.00,0.01, 0.10, and 0.20, respectively, keeping all the other inputparameters settings the same as those in Scenario II (except for boththe coefficientsmemoryLength andactionNum,as previously mentioned). We call this situation Scenario III, andexecute the multi-agent based simulation model on all of the four typesof network topologies, each with 120 simulation ticks (representing 120performances of the routine, iteratively), successively, arriving atthe simulation results as shown in Figure7.

(a) The total number of individual actor agentssearching for new alternatives

(b) The total number of individual actor agentsthat imitate their neighbours

(c) Values of the coefficienttotalCost

(d) Values of the coefficientD

Figure 7. Simulation resultswith default settings for Scenario III

4.15

From the simulation results shown in Figure 7, we canconclude the following:

1. When individual actors involved in the routine areheterogeneous (that is, each individual actor agent has differentvalues of both the coefficientsmemoryLength andactionNum),the routine system on all four network topologies is convergent andeventually reaches a stable state, as shown in the upper-left portionof Figure7 (a) and (b).However, the routine system on scale-free networks would obtain a muchhigher value of the coefficientD, which representsa higher coherency and routinization level of organizationalbehaviours.
2. If there are minor numbers of improvisations fromindividual actors (improvi_P = 0.01, for example),we can deduce from the upper-right portion of Figure7(a), (b), (c) and (d) that the routine system on all four types ofnetwork topologies can be converged over time, arriving at a relativelydifferent yet stable state, as well. This result is in accordance withHutchins's (1991,1995) work on a collectivemanual navigation study, showing that improvisation from participantscan also result in the emergence of new routines in organizations.Although there are no significant differences between these four typesof network topologies, the routine system on scale-free networks obtaina much higher value of the coefficientD (whichrepresents a higher routinization level of organizational behaviours),but expends a relatively higher value of the coefficienttotalCost.Further, the routine system on random networks gains a relatively lowervalue of the coefficientD, indicating itssensitivity to improvisations from the individual actors involved.
3. Nevertheless, with the increase of the improvisationprobability valueimprovi_P (improvi_P= 0.10 and 0.20), there is an increase in individual actors exhibitingmore individualistic manners. In such a situation, we reach the sameconclusion affirmed by Weick (1990), that organizational routines maybe disrupted, as shown in the bottom portion of Figure7 (a) and (b), resulting in thehigher value of the coefficienttotalCost, but thelower value of the coefficientD, as shown in thebottom portion of Figure7 (c)and (d).

Conclusionand discussion

5.1

Organizational routines are some collective phenomena inwhich multiple individual actors are involved. In this paper, byregarding the habitual behaviours of individual actors as elementarybuilding blocks (Bapuji et al. 2012;Knudsen 2008), weconsidered organizational routines from an 'emergence-based'perspective. We emphasised the impacts of connections or networktopologies among individual actors in the formation of organizationalroutines, and developed a multi-agent based simulation model forformally expressing and visually depicting the emerging process oforganizational routines on complex networks. The conclusions of ourwork are as follows.

5.2

First, the multiple, individual actors play fundamentalroles during the formation of organizational routines. By regardingorganizational routines as the emerging results of individual habits ofthe multiple actors involved, we observe a mutual relationship betweenboth the ostensive and performative aspects of the routines (Feldman & Pentland 2003),and provide a comparatively new but effective approach to investigatethe underlying dynamics of organizational routines at the micro-level.

5.3

Second, we created a multi-agent based simulation model oforganizational routines on complex networks. Our simulation resultsindicate that network topologies among individual actors do havesignificant impacts on the dynamic characteristics of organizationalroutines. However, their impacts on the evolutionary trajectory of theroutines are closely influenced by two main factors. The first is thememory capacity of individual actors that are involved in the routines.The second is the component complexity of organizational tasks relatingto the organization. Nevertheless, for all four situations withdifferent combinations of the levels of these two factors (in thispaper, we use the values of the coefficientmemoryLengthandactionNum to represent, respectively), we findthat the organization (which constitutes a certain number of individualactors) on scale-free networks can always obtain a much highercoherency and routinization level of collective behaviours that we callorganizational routines.

5.4

That is, scale-free networks possess inherent superiorityin shaping strong organizational routines. The reason may be thatscale-free networks are beneficial for imitation activities amongvarious individual actors, as imitation is one of the crucialmechanisms through which individual habits are transmitted (Hodgson 2008), leading to theemergence of routines at the organizational level. In addition, theimpacts of network topologies on the total costs that the organizationbears during the implementation of organizational tasks are variant,depending on different combinations of both the memory capacity ofindividual actors and the component complexity of organizational tasksthat the individual actors need to deal with repetitively.

5.5

Third, the turnover of individual actors may result in morefluctuations on the routine system. In such a circumstance, the routinesystem on scale-free networks performs significantly different fromthose on the other three network topologies (i.e. the regular, random,and small-world networks).

5.6

Fourth, when the individual actors involved in the routineare heterogeneous, the routine system on scale-free networks exhibits astrong anti-disturbance ability, and achieves a relatively stable,higher routinization level of collective behaviours over time,regardless of whether or not there were minor improvisations from theseindividual actors. Nevertheless, a large number of improvisationsenabled individual actors to act in more individualistic manners, whichmay destroy the routine system.

5.7

Therefore, we confirm that the routine system shouldmaintain a balance between the inertia resulted from individualmemories, component complexity of organizational tasks, the impacts ofboth turnover and improvisation from the individual actors involved,and the dynamical properties of network topologies within which theseindividual actors are located.

5.8

All of these findings could help open up the 'black box' oforganizational routines that emerge from repeated interaction amongindividual actors, and to further investigate their formation andmicro-dynamics at the micro-level. Nevertheless, we should state thatour simulation results are not validated in practice. This is a richfield of research, and there are some additional significant pointsabout the multi-agent based simulation model that could be improvedupon in the future. For example: (1) Although environment changes arecommon occurrences for today's organizations, we only refer toorganizational routines in a relatively stable environment. (2)Artifacts are not involved in our simulation model, but according to aparticular set of studies (Feldman& Pentland 2003;Pentland& Feldman 2007,2008),artifacts are one of several very important parts of organizationalroutines. (3) Identity of the individual actors involved is also acrucial concept in the organizational routines theory (Hutchins 1995), but is notreferred to in this paper. (4) We only treat the memory capacity ofindividual actors as a rough concept, not identifying the distinctionbetween different kinds of memories (i.e. procedural, declarative, andtransactive (Miller et al. 2012).These would be expanded in a more complete version. In addition, fielddata are needed for validation of the simulation model. This may seemchallenging, but could lead to more fruitful findings in futureresearch.

Acknowledgements

The work for this paper is supported by National Natural ScienceFoundation of China (under Grant No. 71471007, 91224007 and 71302188),Beijing Natural Science Foundation (under Grant No. 9142013).Humanities and social sciences research planning project from theEducation Department of Shandong Province (under Grant No. J12WF04),Youth Foundation from Shandong Institute of Business and Technology(under Grant No. 2014QN004). The authors would like to thank the editorand four anonymous reviewers for their constructive comments, whichhave helped us to improve the manuscript.

Note

This manuscript is a further improved version of the work that wassubmitted to the 44th International Conference on Computers &Industrial Engineering (Istanbul, 14–16 October 2014)

References

AARTS, H., &Dijksterhuis, A. (2000). Habits as knowledge structures: automaticityin goal-directed behavior.Journal of Personality and SocialPsychology,78(1):53–63[doi:10.1037/0022-3514.78.1.53]

ABELL, P.,Felin, T., & Foss, N. J. (2008). Building micro-foundations forthe routines, capabilities, and performance links.Managerialand Decision Economics,29(6): 489–520[doi:10.1002/mde.1413]

BARABÂSI, A.L., Albert, R. & Jeong, H. (1999). Mean-field theory for scalefree random networks.Physica A,272(1):173–187[doi:10.1016/S0378-4371(99)00291-5]

BARABÂSI, A.L., & Bonabeau, E. (2003). Scale-free networks.ScientificAmerican,288(5):60–69[doi:10.1038/scientificamerican0503-60]

BAPUJI, H.,Hora, M., & Saeed, A. M. (2012). Intentions, intermediaries,and interaction: examining the emergence of routines.Journalof Management Studies,49(8):1586-1607[doi:10.1111/j.1467-6486.2012.01063.x]

BECKER, M. C.(2004). Organizational routines: a review of the literature.Industrialand Corporate Change,13(4):643–677[doi:10.1093/icc/dth026]

BECKER, M. C.(2005). The concept of routines: some clarifications.CambridgeJournal of Economics,29(2):249–262[doi:10.1093/cje/bei031]

COHEN, M. D.,& Bacdayan, P. (1994). Organizational routines are stored asprocedural memory: evidence from a laboratory study.OrganizationScience,5(4):554–568[doi:10.1287/orsc.5.4.554]

COHENDET,P., & Llerena, P. (2003). Routines and incentives: the role ofcommunities in the firm.Industrial and Corporate Change,12(2):217–297[doi:10.1093/icc/12.2.271]

DOROGOVTSEV,S. N., & Mendes, J. F. F. (2003).Evolution ofNetworks: From Biological Nets to the Internet and WWW.Oxford: Oxford University Press

ERDÖS, P.,& Rényi, A. (1959). On random graphs.PublicationesMathematicae (Debrecen),6:290–297

FELDMAN, M.S. (2000). Organizational routines as a source of continuous change.OrganizationScience,11(6): 611–629[doi:10.1287/orsc.11.6.611.12529]

FELDMAN, M.S., & Pentland, B. T. (2003). Reconceptualizing organizationalroutines as a source of flexibility and change.AdministrativeScience Quarterly,48(1): 94–118[doi:10.2307/3556620]

FELDMAN, M.S., & Rafaeli, A. (2002). Organizational routines as sources ofconnections and understandings.Journal of Management Studies,39(3):309–331[doi:10.1111/1467-6486.00294]

FELIN, T.,Foss, N. J., Heimeriks, K. H., & Madsen, T. L. (2012).Microfoundations of routines and capabilities: individuals, processes,and structure.Journal of Management Studies,49(8):1351–1374[doi:10.1111/j.1467-6486.2012.01052.x]

GAO,D., Deng, X., & Bai, B. (2014). Theemergence of organizational routines from habitual behaviours ofmultiple actors: an agent-based simulation study.Journal of Simulation,8(3):215-230[doi:10.1057/jos.2014.1]

GILBERT, C.G. (2005). Unbundling the structure of inertia: Resource versus routinerigidity.Academy of Management Journal,48(5):741–763[doi:10.5465/AMJ.2005.18803920]

GILBERT, N.(2008).Agent-Based Models, London: SagePublications

GRAVE, H.(2008). Organizational routines and performance feedback. In: Becker,M. C. (ed.),Handbook of Organizational Routines,Edward Elgar: Cheltenham, 187–205[doi:10.4337/9781848442702.00016]

GROTE,G, Weichbrodt, J. C., Günter, H., Zala-Mezö, E., & Künzle B.(2009). Coordination in high-risk organizations: the need for flexibleroutines.Cognition, Technology & Work,11(1):17–27[doi:10.1007/s10111-008-0119-y]

HENDRICKS,W. O. (1972). The structure study of narration: Sample analyses.Poetics,1(3):100–123[doi:10.1016/0304-422X(72)90040-X]

HENDRICKS,W. O. (1973). Methodology of Narrative structural analysis.Semiotica,7(2):163–184[doi:10.1515/semi.1973.7.2.163]

HODGSON, G.M. (2008). The concept of a routine. In Becker, M. C. (ed.),Handbookof Organizational Routines, Edward Elgar: Cheltenham, 15–30[doi:10.4337/9781848442702.00007]

HODGSON, G.M., & Knudsen, T. (2004). The complex evolution of a simpletraffic convention: the functions and implications of habit.Journal of Economic Behavior & Organization,54(1):19–47[doi:10.1016/j.jebo.2003.04.001]

HOWARD-GRENVILLE,J. A. (2005). The persistence of flexible organizational routines: therole of agency and organizational context.OrganizationScience,16(6):618–636[doi:10.1287/orsc.1050.0150]

HUTCHINS, E.(1991). Organizing work by adaptation.Organization Science,2(1):14–39[doi:10.1287/orsc.2.1.14]

HUTCHINS, E.(1995).Cognition in The Wild, Massachusetts: TheMIT Press

JEONG, H.(2003). Complex scale-free networks.Physica A,321:226–237[doi:10.1016/S0378-4371(02)01774-0]

KNUDSEN, T.(2008). Organizational routines in evolutionary theory. In Becker, M.C. (ed.),Handbook of Organizational Routines,Edward Elgar: Cheltenham, 125–151[doi:10.4337/9781848442702.00014]

KOZICA, A.,Kaiser, S., & Friesl, M. (2014). Organizational routines:conventions as a source of change and stability.SchmalenbachBusiness Review,66(3):334–356

LAZARIC, N.(2008). Routines and routinization: an exploration of somemicro-cognitive foundations. In: Becker, M. C. (ed.),Handbookof Organizational Routines, Cheltenham: Edward Elgar, 205–227[doi:10.4337/9781848442702.00017]

MILGRAM, S.(1967). The small world problem.Psychology Today,2:60-67

MILLER, K. D.,Pentland, B. T., & Choi, S. (2012). Dynamics of performing andremembering organizational routines.Journal of ManagementStudies,49(8):1536–1558[doi:10.1111/j.1467-6486.2012.01062.x]

NORTH, M. J.,& Macal, C. M. (2007).Managing Business Complexity:Discovering Strategic Solutions with Agent-based Modeling and Simulation,Oxford University Press: New York

PENTLAND, B.T. (2003). Conceptualizing and measuring variety in the execution oforganizational work process.Management Science,49(7):857–870[doi:10.1287/mnsc.49.7.857.16382]

PENTLAND, B.T., & Feldman, M. S. (2005). Organizational routines as a unitof analysis.Industrial and Corporate Change,14(5):793–815[doi:10.1093/icc/dth070]

PENTLAND, B.T., & Feldman, M. S. (2007). Narrative networks: patterns oftechnology and organization.Organization Science,18(5):781–795[doi:10.1287/orsc.1070.0283]

PENTLAND, B.T., & Feldman, M. S. (2008). Designing routines: on the follyof designing artifacts, while hoping for patterns of action.Informationand Organization,18(4):235–250[doi:10.1016/j.infoandorg.2008.08.001]

PENTLAND, B.T., Hærem, T., & Hillison, D. (2010). Comparing organizationalroutines as recurrent patterns of action.Organization Studies,31(7):917–940[doi:10.1177/0170840610373200]

POSTREL, S.,& Rumelt, R. P. (1996). Incentives, routines and self-command.In Dosi, G., & Malerba, F. (ed.),Organization andStragety in the Evolution of the Enterprise, Basingstoke andLondon: Macmillan Press Ltd, 72–102[doi:10.1007/978-1-349-13389-5_4]

SCHULZ, M.(2008). Staying on track: a voyage to the internal mechanisms ofroutine reproduction. In: Becker, M. C. (ed.),Handbook ofOrganizational Routines, Edward Elgar: Cheltenham, 228–255[doi:10.4337/9781848442702.00018]

SIMON, H.(1945).Administrative Behavior, New York: FreePress

SWARM DEVELOPMENT GROUP[SDG]. (2000).A tutorial introduction to Swarm,http://www.swarm.org ,accessed 15 October 2009

TURNER, S. F.,& Fern, M. J. (2012). Examining the stability and variabilityof routine performances: the effects of experience and context change.Journalof management Studies,49(8):1407–1434[doi:10.1111/j.1467-6486.2012.01061.x]

TURNER, S.F.,& Rindova, V. (2012). A balancing act: How organizations pursueconsistency in routine functioning in the face of ongoing change.OrganizationScience,23(1):24–46[doi:10.1287/orsc.1110.0653]

WANG, X. F.,& Chen, G. (2003). Complex networks: small-world, scale-freeand beyond.Circuits and Systems Magazine, IEEE,3(1):6–20[doi:10.1109/MCAS.2003.1228503]

WATTS, D. J.(1999).Small Worlds: The Dynamics of Networks between Orderand Randomness, Princeton University Press

WATTS, D. J.,& Stogatz, S. H. (1998). Collective dynamics of "small-world"networks.Nature,393:440–441[doi:10.1038/30918]

WEICK, K. E.(1990). The vulnerable system: an analysis of the Tenerife airdisaster.Journal of Management,16(3):571–593[doi:10.1177/014920639001600304]

WINTER, S. G.(2003). Understanding dynamic capabilities.StrategicManagement Journal,24(10):991–995[doi:10.1002/smj.318]

WITT, U. (2011).Emergence and functionality of organizational routines: anindividualistic approach.Journal of Institutional Economics,7(2):157–174[doi:10.1017/S1744137410000226]

WOOD, R. E.(1986). Task complexity: Definition of the construct.OrganizationalBehavior and Human Decision Processes,37(1):60–82[doi:10.1016/0749-5978(86)90044-0]

WOOD, W.,& Neal, D. T. (2007). A new look at habits and the interfacebetween habits and goals.Psychological Review,114(4):843–863[doi:10.1037/0033-295X.114.4.843]

YIN, H. H.,Knowlton, B. J., & Balleine, B. W. (2004). Lesions ofdorslateral striatum preserve outcome expectancy but disrupt habitformation in instrumental learning.European Journal ofNeuroscience,19(1):181–189[doi:10.1111/j.1460-9568.2004.03095.x]