PreprintPDF Available
Preprints and early-stage research may not have been peer reviewed yet.

Abstract

In this study, we unpack the micro-level processes of knowledge accumulation (experiential learning) and knowledge application (problem solving) to examine how task allocation structures influence organizational learning. We draw on untapped potential of the classical garbage can model, and extend it to analyze how restrictions on project participation influence differentiation and integration of organizational members' knowledge and consequently organizational efficiency in solving the diverse, changing problems from an uncertain task environment. To isolate the effects of problem or knowledge diversity and experiential learning, we designed three simulation experiments to identify the most efficient task allocation structure in conditions of 1) knowledge homogeneity, 2) knowledge heterogeneity, and 3) experiential learning. We find that free project participation is superior when the members' knowledge and the problems they solve are homogenous. When problems and knowledge are heterogeneous, the design requirement is on matching specialists to problem types. Finally, we found that experiential learning creates a dynamic problem where the double-duty of adapting the members' specialization and matching the specialists to problem types is best solved by a hierarchic structure (if problems are challenging). Underlying the efficiency of the hierarchical structure is an adaptive role of specialized members in organizational learning and problem solving: their narrow but deep knowledge helps the organization to adapt the knowledge of its members while efficiently dealing with the problems at hand. This happens because highly specialized members reduce the necessary scope of knowledge and learning for other members during a certain period of time. And this makes it easier for the generalists, and for the organization as a whole, to adapt to unforeseen shifts in knowledge demand, because they need to learn less. From this nuanced perspective, differentiation and integration may have a complementary, rather than contradictory, relation under environmental uncertainty and problem diversity.
Differentiation and Integration in Organizational Learning: A Garbage Can Model
Sangyoon Yi
Graduate School of Future Strategy
Korea Advanced Institute of Science and Technology
Nils Stieglitz
Frankfurt School of Finance and Management
Thorbjørn Knudsen
Danish Institute for Advanced Study
&
Strategic Organization Design Group
Department of Marketing and Management
University of Sothern Denmark
1
Differentiation and Integration in Organizational Learning: A Garbage Can Model
ABSTRACT
In this study, we unpack the micro-level processes of knowledge accumulation (experiential learning)
and knowledge application (problem solving) to examine how task allocation structures influence
organizational learning. We draw on untapped potential of the classical garbage can model, and
extend it to analyze how restrictions on project participation influence differentiation and integration
of organizational membersknowledge and consequently organizational efficiency in solving the
diverse, changing problems from an uncertain task environment. To isolate the effects of problem or
knowledge diversity and experiential learning, we designed three simulation experiments to identify
the most efficient task allocation structure in conditions of 1) knowledge homogeneity, 2) knowledge
heterogeneity, and 3) experiential learning. We find that free project participation is superior when the
membersknowledge and the problems they solve are homogenous. When problems and knowledge
are heterogeneous, the design requirement is on matching specialists to problem types. Finally, we
found that experiential learning creates a dynamic problem where the double-duty of adapting the
membersspecialization and matching the specialists to problem types is best solved by a hierarchic
structure (if problems are challenging). Underlying the efficiency of the hierarchical structure is an
adaptive role of specialized members in organizational learning and problem solving: their narrow but
deep knowledge helps the organization to adapt the knowledge of its members while efficiently
dealing with the problems at hand. This happens because highly specialized members reduce the
necessary scope of knowledge and learning for other members during a certain period of time. And
this makes it easier for the generalists, and for the organization as a whole, to adapt to unforeseen
shifts in knowledge demand, because they need to learn less. From this nuanced perspective,
differentiation and integration may have a complementary, rather than contradictory, relation under
environmental uncertainty and problem diversity.
Keywords: Coordination, Differentiation, Garbage Can Model, Organizational Structure,
Organizational Learning, Problem Solving, Specialization
2
INTRODUCTION
In the modern knowledge economy, it is probably difficult to over-state the importance of
organizational learning for the survival and success of firms. A notable fact about such learning
processes is that the division of labor shapes the trajectory of organizational learning (Argote, 1999;
Lawrence & Lorsch, 1967; Levinthal & March, 1993; Levitt & March, 1988; Simon, 1991a).
Different organizational structures thereby have long-term consequences for the development and
distribution of knowledge within the organization and, in turn, for organizational performance.
Despite the significance of organizational learning and stark differences in organizational structure,
the current literature only offers a rudimentary perspective on how organizational structure impacts
the micro-level process and aggregate-level performance of organizational learning. Critically, the
approaches in prior research largely abstract from differentiation and integration, the two defining
dimensions of organization design research (Burton & Obel 2004; Puranam, Alexy, & Reitzig 2014).
This limitation has severe implications, including identification problems in empirical research and
developing a solid basis for useful managerial advice.
In this work, we aim to advance useful principles for understanding and enhancing the organization of
internal processes of knowledge accumulation and application. We focus on how a particularly
important aspect of structure, namely the internal division of labor (differentiation) and the subsequent
coordination (integration) of the divided labor, influence organizational learning and efficiency in
uncertain task environments. To do so, we identify untapped potential of the Garbage Can Model
(GCM), the classic agent-based model of organizational decision-making processes (Cohen, March, &
Olsen, 1972), and build on it to develop a formal model of organizational problem-solving and
learning. Our model formalizes micro-level processes of integration and learning of organizational
members who jointly solve diverse problems arising from an uncertain task environment. Resolution
of different problems demand that different types of specialist knowledge are applied. For this reason,
the organizational structure has a critical role in task allocation, i.e. in regulating who does what.
From a dynamic perspective, the task allocation structure has the added role of regulating learning
processes as it effectively distributes learning opportunities among organizational members. That is,
the task allocation structure determines who deals with what kinds of problems (e.g., restrictions on
project participation), and thereby shapes the distribution of learning opportunities and the future
distribution of knowledge that results from that distribution. The task allocation structure, therefore,
not only determines organizational efficiency in solving today’s problems but also influences its
capacity to solve tomorrow’s problems.
3
We consider organizational structures in terms of the degree of differentiation (and thus specialization)
of organizational members with respect to the kind and range of problems assigned to them.
Organizational members under alternative structures therefore face different integration challenges in
their joint problem solving – the more differentiated, the more coordination needed for matching
members and problems, and vice versa (Lawrence & Lorsch, 1967; Mintzberg, 1980).
To stay loyal to the aim of examining structural effects on the processes of problem-solving and
learning, our model and analysis deliberately abstract from other consequences of differentiation such
as conflicts of interest and coordination costs (Becker & Murphy, 1992; Conner & Prahalad, 1996;
Lawrence & Lorsch, 1967; Williamson, 1975). We concentrate solely on the way that structural
differentiation shapes the distribution of learning opportunities and the integration of joint activities in
problem-solving.
To isolate the effects of problem or knowledge diversity and experiential learning, we designed three
simulation experiments to identify the most efficient task allocation structure in conditions of 1)
knowledge homogeneity, 2) knowledge heterogeneity, and 3) experiential learning. We find that free
participation is superior when the membersknowledge and the problems they solve are homogenous.
When problems and knowledge are heterogeneous, the design requirement is on matching specialists
to problem types. Finally, our main result is that experiential learning creates a dynamic problem
where the double-duty of adapting the memberslevel of specialization and matching the specialists to
problem types is best solved by a hierarchic structure (if problems are challenging).
Hierarchical task allocation or restriction of project participation creates diversity and overlap in the
task scopes of members. That is, some members may be allowed to deal with more kinds of problems
and develop shallower but broader, more general knowledge, whereas other members are dedicated to
fewer kinds of problems and allowed to develop narrower but deeper, more specialized knowledge.
Underlying the efficiency of the hierarchical structure is an adaptive role of specialized members in
organizational learning and problem solving: their narrow but deep knowledge helps the organization
to adapt the knowledge of its members while efficiently dealing with the problems at hand. This
happens because highly specialized members reduce the necessary scope of knowledge and learning
for other members during a certain period of time. And this makes it easier for the generalists, and for
the organization as a whole, to adapt to unforeseen shifts in knowledge demand, because they need to
learn less. Hierarchic restriction of participation therefore enhances organizational capability to cope
with unpredictable exceptions in a turbulent task environment. From this nuanced perspective,
differentiation and integration may have a complementary, rather than contradictory, relation under
environmental uncertainty and problem diversity.
4
Our results have important implications for advancing theory on organization design, and for
understanding the conditions under which limiting access to project participation increases
organizational performance. In the concluding section, we elaborate on these.
THEORETICAL BACKGROUND
Knowledge-Based View of Competitive Advantage
Knowledge is regarded as the most important source of competitive advantage (Barney, 1986; Grant,
1996a; Kogut & Zander, 1996). Human productivity is knowledge dependent, and technologies and
tools can be viewed as embodiments of knowledge. Unlike other value-creating resources (e.g.
financial resources), strategically important knowledge is often difficult to recognize, understand, and
transfer. Such characteristics of valuable knowledge serve as an important isolating mechanism for the
persistence of performance differences among competing firms (Lippman & Rumelt, 1982).
Knowledge-based competitive advantage therefore constitutes the core of the resource-based view of
the firm (Barney, 1996; Conner & Prahalad, 1996; Demsetz, 1973).
At the heart of such knowledge-based view of competitive advantage lies the issue of differentiation
and integration in organizational learning. The salient question here is how valuable specialist
knowledge can be developed at the individual level and integrated through organizational means (e.g.,
task allocation, incentive system, coordination rules) into organization-level capability (Grant, 1996b;
Winter, 2000). A basic tenet of the knowledge-based view is that firms exist as an institution where
such knowledge differentiation and integration are facilitated by organizing principles or mechanisms
that markets cannot provide (Ghoshal & Moran, 1996; Grant, 1996a; Kogut & Zander, 1996; Spender,
1996). Such knowledge processes can be influenced and directed through the formal aspects of the
organizational structure (Foss, 2007; Nickerson & Zenger, 2004).
Organizational Structure and Learning
While research has found that organizational structure has a significant impact on organizational
learning (e.g. March 1991; Gupta et al. 2006; Boumgarden et al. 2012), it is an open question what
structures best support learning. In a recent contribution on this question, Fang et al. (2010) showed
how subgroup isolation and intergroup connectivity influence organizational learning. Related work
shows how social networks within an organization influence knowledge sharing (Hansen et al. 2005;
Borgatti & Cross 2003; Reagans & McEvily 2003) and thereby impacts learning processes (Lazer &
Friedman, 2007; Miller, Zhao, & Calantone, 2006). At a higher level of analysis, another stream of
5
literature on industry architecture has clarified how concerns about organizational boundary (e.g.,
organizing to minimize transaction costs), has long-term consequences for learning and development
of capabilities (Jacobides and Winter 2005; Jacobides et al. 2006). It is clear from this line of work
that different ways of organizing influence learning processes in important ways, which in turn shapes
organizational performance, especially the quality of organizational solutions (e.g., Carley, 1992;
Ethiraj & Levinthal, 2004; Fang et al., 2010; Knudsen & Levinthal, 2007).
However, the causal links between specific organizational forms and the micro-level processes of
organizational learning (e.g., organizational members’ learning and their interaction and integration)
have not been systematically examined. In particular, it is not clear how organizational structures
influence organizational learning and performance conditional on the level of differentiation and
specialization. Neither is it clear how organizational structures shape the way that specialized
knowledge and capabilities develop and contribute to organizational efficiency in coping with
uncertain task environments.
Critically, prior research neglects differentiation as a core feature of organizational structure.
Differentiation implies that the overall organizational task is subdivided among specialists (Lawrence
& Lorsch, 1967) and, as Drazin (1995: 137) noted, “to be organized is, by definition, to be
differentiated.” A notable exception in that regard is the study by Bunderson and Boumgarden (2010)
who show how team structure – specialization, hierarchy, and formalization – reduces conflict, fosters
psychological safety, and increases information sharing, thereby encouraging team learning. Related
work on organizational search considered how different elements of the organizational structure such
as (de)centralization, modularity, and firm-level incentive promote or stifle explorative search and
thereby improve organizational performance in complex, uncertain task environments (e.g., Ethiraj &
Levinthal, 2004; Knudsen & Levinthal, 2007; Rivkin & Siggelkow, 2003; Siggelkow & Levinthal,
2003; Siggelkow & Rivkin, 2005). But those studies, focusing on search rather than problem solving,
do not examine micro-level processes of knowledge accumulation and application, in which
specialization and coordination of organizational members are shaped by organization structure.
Indeed, in a recent survey on the topic, for example, Argote and Miron-Spektor (2011: 1127) singled
out the organization’s structure as an area ripe for further research on organizational learning.
In the present paper, we analyze how two fundamental dimensions of organization – differentiation
and integration (Lawrence & Lorsch, 1967; Mintzberg, 1979) – impact the micro-level processes of
organizational learning and how they shape organizational efficiency in solving problems arising from
an uncertain task environment. Differentiation is defined as the internal division of labor, i.e., the
allocation of subtasks to organizational members (such as individuals or groups). That is, in terms of
6
Lawrence and Lorsch, different members are assigned differentiated sets of tasks to cope with
corresponding sub-environments. Integration ensures that the efforts of specialized members are
coordinated and jointly contribute to overall organizational performance (Lawrence & Lorsch, 1967;
March & Simon, 1958; Thompson, 1967).
Our key idea is that differentiation allocates opportunities for learning from task experiences (Argote
& Miron-Spektor, 2011; Levitt & March, 1988): the allocation of tasks influences what problems or
projects organizational members encounter and deal with, and these work experiences feed into the
development of specialized, task-related knowledge on the individual level. Differentiation in tasks
and problem solving thereby creates differentiation in experiences and knowledge, and thus
specialized expertise. At the same time, individual differentiation precipitates organizational
integration (Lawrence & Lorsch, 1967). This is because specialists often have to contribute jointly to
solving problems (Nickerson and Zenger 2004; Knudsen and Srikanth 2014) and their efforts
therefore need to be coordinated (Cohen et al., 1972; Grant, 1996a; March & Simon, 1958). Crucially,
organizational integration becomes more demanding the more specialized and diverse the knowledge
required for solving problems while effective coordination often requires generalists with broader
(less specialized) knowledge (Knudsen and Srikanth 2014). From the organization design perspective,
the dilemma is that organizational members could gain diverse expertise under a differentiated
structure while more general knowledge is better fostered in less differentiated structures that provide
a broader range of task experiences. Taken together, differentiation and integration create a
fundamental tradeoff for organizational learning, as well as for organization design, that prior
literature has not forcefully and formally examined.
Knowledge Accumulation and Application
Organization design that aims to promote differentiation of labor and integration of effort produced by
the divided labor encounters two fundamental challenges. First, integration of effort is challenged by
difficulties relating to knowledge sharing (Demsetz, 1988; Grant, 1996a; Nickerson & Zenger, 2004;
Osterloh & Frey, 2000). The ability or motivation to share knowledge is impeded mainly by two
conditions: humans are cognitively constrained in the speed with which they learn and they are prone
to self-interest (March, 1991; Simon, 1991b, 1991a). Even if valuable knowledge is identified, the
person who possesses such knowledge may not want to share it and/or considerable time and efforts
may be needed to absorb and apply it. Although difficulty in transferring valuable knowledge across
firm boundaries constitutes the theoretical underpinning of the knowledge-based view of competitive
advantage, it should be somehow overcome within the organizational boundary to gain enough
advantage from such knowledge. Crucially, full knowledge transfer is not often feasible and desirable.
7
The whole point of gains from the division of labor is that specialists differ in what they know and
what they can do (Demsetz, 1997; Grant, 1996b). Coordination for knowledge application is therefore
often best achieved through the direct involvement or participation of individual specialists. In this
regard, mechanisms such as team- or project-based organization (Epstein, Ketcham, & Nicholson,
2010; Nickerson & Zenger, 2004) have become more prevalent. This principle of organization,
however, suffers from a fundamental challenge of identifying, classifying, and matching various kinds
of problems to organizational members with corresponding knowledge in a timely manner. Such
processes cannot be perfect but error- or resistance-prone for such fundamental reasons as tacitness of
knowledge, ambiguity in classification, and incentives for commitment (Epstein et al., 2010;
Lawrence & Lorsch, 1967; Levinthal & March, 1993; Osterloh & Frey, 2000; Polanyi, 1966).
The second, relatively less explored, issue is that knowledge accumulation can hardly be decoupled
from knowledge application. Valuable knowledge is usually garnered through context-specific
learning-by-doing or experiential (as opposed to cognitive) learning processes. Organizational
members engage in experiential learning – they learn most effectively from what they do, experience,
and reflect on (Argote, 1999; Arrow, 1962; Levitt & March, 1988; Lundberg, 1961). Since the
organization’s task structure allocates learning opportunities, it is a major determinant of experiential
learning. This implies that the way in which specialized members are assigned to specific tasks shapes
the accumulation of knowledge, which in turn influences the demands on the ways they coordinate
their effort (Kogut & Zander, 1992). A true challenge of organization design from a learning
perspective is therefore rooted in such inherent interdependence between knowledge differentiation
and accumulation at the individual (or group) level and knowledge integration and application at the
organization level.
Untapped Potential of the Garbage Can Model
To examine these fundamental issues, we develop a computational model by building on the Garbage
Can Model (GCM), the classic agent-based model of organizational decision making or problem
solving, introduced by Cohen, March and Olsen (1972). In this paper, we shed light on the untapped
potential of GCM as an extensible, formal platform to study the micro- or problem-level processes of
organizational performance and learning that are influenced by various dimensions of organizational
structure. GCM depicts organizational choice as a dynamic integration challenge, in which multiple
decision-makers attempt jointly to solve problems that the organization faces. The model has been
picked up widely for the empirical study and for the theoretical analysis of organizational decision
processes (see the volume by Lomi and Harrison (2012) for a recent appraisal of the model). Crucially,
the original GCM, while highlighting the coordination of differentiated decision makers in solving
8
decision problems, does not consider how (joint) problem-solving impacts individual learning and the
distribution and dynamics of individual competence (i.e., problem-solving knowledge).
In this study, we tap the potential of this foundational model by extending it to accommodate task-
related learning and knowledge differentiation or specialization of organizational members. Our
model depicts an organization that faces an exogenous and uncertain influx of diverse problems.
Organizational performance depends on the efficiency of solving problems. Organizational members
participate in projects to tackle the associated problems by exerting differentiated efforts, and their
problem-specific competence determines how much their efforts contribute to project accomplishment
or problem resolution. In doing so, organizational members learn from the problems they work on and
this experiential learning (i.e., learning by doing) enhances their future competence in solving similar
kinds of problems. The organizational structure determines who gets to work on what, and therefore
influences the internal processes of knowledge accumulation and application. Our focal interest here
is in how alternative structures (varying along the degree and the way of differentiation) impact the
problem-solving efficiency of an organization and how the result depends on the characteristics of its
task environment.
A GARBAGE CAN MODEL OF ORGANIZATIONAL LEARNING
The model is designed to examine how organizational structure for task allocation influences the
micro-level processes of organizational members’ learning and coordination in joint problem solving.
It builds on and extends the Garbage Can Model (GCM) of organizational decision processes (Cohen
et al., 1972) that formalizes a dynamic integration process in which organizational structure influences
the matching between decision makers and decision problems (Fioretti & Lomi, 2008). To address our
research question, we extend the classic GCM to accommodate knowledge diversity (i.e.,
differentiation and specialization) and experiential learning (i.e., learning from solving problems) of
organizational members.
Given our interest in tacit, difficult-to-transfer knowledge and experiential learning, and to be faithful
to the well-respected organizational principles behind the original GCM, our model builds on three
basic assumptions: (1) an organization deals with an influx of unanticipated, diverse problems arising
from its task environment, (2) various types of problems require corresponding types of knowledge
to be solved, and (3) the organization has a limited number of organizational members whose
knowledge adapts to the problems they deal with. An interesting and important aspect of the problem-
solving process is, as noted above, the interdependence between experiential learning at the individual
9
level (knowledge accumulation) and coordination or matching at the organizational level (knowledge
application).
Our analysis examines the efficacy of an organization’s task structure in balancing the two kinds of
knowledge processes (Argote & Miron-Spektor, 2011; Arrow, 1962; Epstein et al., 2010; Lawrence &
Lorsch, 1967; Levinthal & March, 1993). To provide focus, and to comply with GCM principles, we
deliberately rule out complexity relating to the task structure by assuming decomposability or
modularity in the division of tasks (problems) and knowledge (Baldwin & Clark, 2000; Ethiraj &
Levinthal, 2004; Sanchez & Mahoney, 1996; Simon, 1962) – this assumption is relaxed in the
sensitivity analysis to examine its implications, in relation to the issue of interdependence in projects.
The organizational structure determines who deals with and learns from what kinds of problems,
which in turn shapes organizational efficiency in problem solving as well as knowledge distribution
across organizational members. We consider alternative organizational structures varying along the
degree and the way of differentiation in such task allocation and learning. The following subsections
specify each element of the model.
Task Environment and Organization
An organization is represented as a collection of a given number (M) of organizational members (e.g.,
individuals or groups), which operates in a task environment where various kinds of problems arise
continuously. We assume a finite number of problem types (P).
Organizational members are specialized in different types of problems. Specifically, a member m has
a knowledge vector Kn = {km,1, km,2, …, km,P}, with each element representing the depth of knowledge
for problem type p (1 ≤ pP). Knowledge depth is related to problem types. It is the amount of
specialized effort that a member can produce in each time-period to solve a problem of type p. The
deeper the knowledge, the more effective is the member in solving a problem of a given type. Without
loss of generality, we assume that an organizational member’s knowledge capacity is fixed to unity:
km,1+ km,2+ … + km,P = 1 where 0 ≤ km,p ≤ 1.
The task environment is characterized by a stream of problems, which is denoted by three,
collectively exhaustive, aspects: 1) intensity, 2) diversity, and 3) turbulence. To keep the model simple,
we fix the problem stream to a constant arrival rate of one problem per time-period. Given this base
rate, the intensity of knowledge demand, or problem difficulty, is modeled as a given amount of
specialized effort (E) required by each problem for its resolution. The diversity of knowledge demand
is represented by the problem types (P), and turbulence is formalized by assuming that the type of the
problem that arrives in each time-period is randomly determined. Environmental turbulence is
10
therefore conceptualized as unpredictable changes in the demand of certain kinds of problem-relevant
knowledge over time.
Organizational Structure
We focus on the organizational structure for task allocation that determines projects that
organizational members may participate in. Projects are organizational-level constructs that allow
multiple members to contribute to solve problems of a certain type p. Participation in a project
represents allocation of attention to the associated problems, which provides opportunities for
mobilizing and matching capabilities (membersknowledge) to resolve the problems (Cohen, March
and Olsen 1972). We examine three stylized organizational structures that represent different
constraints on project participation, and thus represent different degrees and ways of differentiation of
organizational members: unsegmented, hierarchic and specialized structures (Cohen et al., 1972).
Under the unsegmented structure, organizational members can be involved in any kinds of projects,
solving any kinds of associated problems, while only the projects and problems that match the
member’s type are accessible under the specialized structure. The hierarchic structure represents in-
between cases where some members may deal with a broader range of projects and problems than
others. That is, the specialized structure and the hierarchic structure capture member heterogeneity in
the range of accessible projects and problems, respectively, while member homogeneity is assumed in
the unsegmented structure. Technically, the specialized structure assigns members of type m to
problems of type p = m, where 0 ≤ m, p P, by allowing for their participation in the projects of the
same type. Under the hierarchic structure, members of type m can participate in projects and deal with
problems of type p m.
Joint Problem-Solving
To deal with the problems at hand, the organization initiates projects that allow its members to work
together, i.e. apply specialized knowledge and integrate their joint efforts. An organizational principle
behind the original GCM is that problems and projects (‘decision opportunitiesin GCM) appear
exogenously (Cohen et al., 1972). We respect this principle because it is a reasonable simplifying
assumption. For instance, there may be (unpredictable) time lags between the emergence of problems
and the initiation of concerted efforts to deal with them. It can take time to identify and understand
certain problems and set up the projects required to handle them. And given various kinds of problems,
the organization may prioritize different projects, which will delay project initiation.
Our model formalizes this unpredictable, exogenous aspect of project initiation in the simplest
possible way: while a new problem arrives in each time-period, in every other time-period, the
11
organization initiates a new project dedicated to deal with a randomly chosen type of problem. This
assumption captures the common situation where the organization is confronted with an excess of
problems, relative to its coordination capacity (e.g., projects) for providing solutions. Once a project
has been initiated, any existing problem of the same type is, along with the members who are allowed
to work on it, assigned to this project for resolution. These assumptions are both realistic and
consistent with the original GCM.
When assigned to a project, the participating members jointly attempt to resolve the associated
problems and thereby finish off the project. The time spent on the project depends on the problem
difficulty. Organizational members and problems continue to be mixed and matched with respect to
available projects. This depicts a process of dynamic coordination in decision making where
participants and problems vary over time (Bendor, Moe, & Shotts, 2001; Cohen et al., 1972). The
process of matching members and projects is imperfect as it is done by boundedly rational human
agents (Cohen et al., 1972; Levinthal & March, 1993; Nickerson & Zenger, 2004; Simon, 1991a).
Consistent with the original GCM, this matching process is operationalized with a simple behavioral
rule: a member will choose to work on the project which, as a result, becomes most likely to be solved
among the accessible ones (determined by organizational structure). Put differently, a member looks
for a project with easily solvable problems, an assumption that is both consistent with a behavioral
bias toward exploitation (March, 1991), and a general tendency to “make a difference” in near future.
In sum, our modeling of matching participants (members) with problems and projects captures an
aspect of coordination that is formalized as joint problem solving.
1
As in the original GCM, the
matching is a result of the interplay of top-down managerial control (i.e., participation restriction) and
bottom-up selection of projects by the participants (i.e., preferring those closer to resolution) – a form
of (structurally) constrained fluid participation to make collective efforts for organizational
performance.
1
Our basic model set-up is based on the intuition applied in prior work (Cohen et al., 1972, 2012; Lomi &
Harrison, 2012; Masuch & LaPotin, 1989; Padgett, 1980, 2013). In this class of models, there is no explicit cost
scale. As in coordination games, the cost scale is set by opportunity costs, i.e. in game theory, the implicit cost
of taking actions that miss out on more profitable actions. The GCM captures a dynamic coordination problem
where the underlying payoff structure changes endogenously. Rather than measuring performance on a
predefined payoff matrix (as in conventional game theory), performance in the GCM is measured in terms of the
organization’s problem-solving efficiency. Foregone opportunities to take actions that are more efficient thus
characterize coordination costs. Per our discussion of the way coordination costs are defined (as opportunity
costs), we adhere to the traditional specification in the general class of GCM.
12
With respect to a project, the specialized effort of the participating members is accumulated and
contributes to resolving the associated problems. Specifically, the specialized effort of a current
participant for a project of type j (the j-th element of her knowledge vector) is added to the past stock
of accumulated effort for the project. If this cumulative sum of specialized efforts is greater than the
sum of the required efforts of the associated problems, and if there is at least one participant
2
, the
associated problems are solved and the project dissolved. The idea of the stock of specialized efforts
captures the intuition that current problem-solving is a function of prior efforts as well as the efforts of
current participants.
Learning and Specialization
A key assumption is that organizational members’ knowledge adapts to the problems they deal with,
while the total capacity of knowledge (i.e., the sum of the elements of the knowledge vector) and the
learning rate are constant. This assumption is operationalized as follows. At a given time point t, the
knowledge vector K(t) of an organizational member who is working on a project of type j adapts to a
unit vector Uj with the j-th element of 1, the other elements set to 0. This adaptation process occurs
with a learning rate of r (0 r ≤ 1), with K(t) as an exponentially weighted average of prior
knowledge: K(t) = rUj + (1 - r)K(t-1).
3
This implies that faster learning from new experiences
coincides with faster forgetting of old ones. In models of experiential learning, this is a widely used
formulation (e.g., Christensen & Knudsen, 2010; Denrell & March, 2001; Levinthal & March, 1981;
Puranama et al., 2015).
This formulation captures two important aspects of experiential learning. First, an organizational
member will experience diminishing returns to working on known problems at hand. Second,
organizational members gradually forget what they are not working on – their knowledge is
depreciated unless continuously developed and used. Consequently, learning about problems of a
certain type and unlearning (or depreciation of knowledge) about problems of other types coincide
2
In the original garbage can model (Cohen et al., 1972), a decision can be made without a decision maker
(decision by flight), which Bendor et al. (2001) criticized as “ghost energyand found it behaviorally
implausible. We address their concern by adding this requirement for decision making. Our results are, however,
robust with respect to this requirement.
3
We use a standard reinforcement learning function, capturing the role of exploitation and exploration in
problems of experiential learning and knowledge adaptation (March, 1991; Sutton & Barto, 1998). The
parameter r determines to what extent newly acquired information overrides old information. Low values of r
are associated with a mode of exploitation, while high values of r are associated with a mode of exploration.
Thus, if r = 0, the organizational member (agent) learns nothing and is exclusively exploiting prior knowledge –
and when r = 1, the agent is in a mode of pure exploration where only the most recent information is considered.
As this learning function is well known and widely used in the literature, we find this is a reasonable choice.
13
(Dierickx & Cool, 1989). This assumption is particularly relevant when it comes to knowledge of a
broad range of specialized professions, including managers, engineers, and lawyers. Such specialist
knowledge can only be achieved and kept up-to-date through continuous commitment to learning by
doing.
SIMULATION EXPERIMENTS AND RESULTS
Our simulation experiments and analysis are designed to examine how the alternative task allocation
structures serve the double-duty role of allocating opportunities for experiential learning across the
members (knowledge accumulation) and matching the specialists with the projects or problems to
generate organizational performance (knowledge application). Our argument is that this double-duty
role is best accomplished by the hierarchic structure. Because the theoretical underpinning of this
argument relies on assumptions about knowledge heterogeneity and experiential learning, we
deliberately design two simplified experiments that omit one, or both, of the two assumptions. This
aids in understanding the underlying mechanisms of our arguments.
In experiment 1, we abstract away from the effects of both knowledge heterogeneity and experiential
learning, and assume that all members have homogenous knowledge and can deal with any problem –
note that their participation in a certain project is still regulated by the organizational structure.
In experiment 2, we assume heterogeneous but constant knowledge of organizational members with
which they are capable of handling the problems they face under the given organizational structure.
For the purpose of presentation, we report here the results for ten problem types (P = 10), but the
qualitative patterns are robust against the number of types. Since other model settings are identical
except for the heterogeneity assumption, any differences in the results can be considered as the
consequence of heterogeneity in member knowledge. Since both experiments abstract from
experiential learning and knowledge accumulation, the two experiments simplify the exposition of the
results by focusing only on the dynamic coordination (mixing and matching) of heterogeneous
knowledge or problem-solving efforts of the organizational members.
In experiment 3, we finally bring in experiential learning by assuming that the knowledge of
organizational members becomes specialized into the kinds of problems they deal with. Therefore,
organizational structure, which constrains who can deal with what kinds of problems, influences the
intra-organizational distribution of individual knowledge and problem-solving efficiency in the future.
The two simplified models in the experiment 1 and 2 provide a baseline to examine this dynamic and
endogenous process of knowledge accumulation and application, which is shaped by the
organizational structure of participation restriction.
14
Experiment 1: Homogeneous Knowledge
The simplified model for this experiment assumes homogeneous and constant knowledge of
organizational members. Specifically, every member can apply a problem-solving effort of 1 (full
knowledge capacity) in a time-period to any problem, and every problem requires an amount of effort
E for resolution which is parameterized as problem difficulty. The organization has ten members (of
ten types, respectively, i.e., M = 10) and encounters a new problem (of a random type) in each period.
Therefore, problem difficulty is calculated as E/10.
Figure 1 presents the problem-solving rate (= % of problems solved) of the three organizational
structures, observed over 100 time periods (same for all the figures). The result shows that the
unsegmented structure outperforms the hierarchic and specialized structures over the whole range of
problem difficulty. The intuition is straightforward. Given the amount of available problem-solving
effort of organizational members, problem-solving efficiency can be enhanced by reducing wasted
effort. As the kinds of problems and projects arising from the environment vary over time, structural
segmentation or participation restriction could generate idle members who fail to find an accessible
project, which results in the waste of their available efforts. In contrast, unsegmented structure allows
organizational members to join any project, and thus they are unlikely to remain idle. By this logic,
the amount of wasted energy is expected to increase with the degree of structural segmentation or
differentiation so that specialized structure performs worst. The result in Figure 1 confirms this
conjecture.
< Figure 1 >
Experiment 2: Heterogeneous Knowledge without Learning
In this experiment, we examine the effect of knowledge heterogeneity at the individual level on
problem-solving efficiency at the organization level. The second simplified model assumes that
problems of each type require corresponding knowledge to be solved, and organizational members
have diverse knowledge determined by the organizational structure (i.e., structurally differentiated
knowledge). Under the unsegmented structure, every member has a knowledge vector of {0.1, 0.1, …,
0.1}. Under the specialized structure, a member of type m has a unit knowledge vector with m-th
element of 1 and others of 0. The scope of knowledge varies across the members under the hierarchic
structure, depending on their type. A member of type m has a knowledge vector with m-th to P-th
element of 1/(P-m+1) and the other elements of 0. This distribution of knowledge is an
operationalization of the assumption of limited knowledge capacity – i.e., the sum of the elements of
the knowledge vector is equal to 1. This implies that organizational members with broader knowledge
15
are less efficient in solving problems than are members who possess narrower, specialized knowledge
relating to the problems. The remaining model settings are identical to the first experiment so that any
differences in the results are either caused by structural differentiation or heterogeneity in the
members’ knowledge. As none of the model changes affects the model dynamics under the specialized
structure, its problem-solving rate can be used as a benchmark across the experiments.
Figure 2 presents the results. Two qualitative patterns are observed. First, the overall problem-solving
efficiency of the unsegmented and hierarchic structures has declined below that of the specialized
structure. Second, knowledge heterogeneity among organizational members gives rise to a reversal of
the unsegmented structure’s performance. In the previous experiment, without knowledge
heterogeneity, it was superior, but it is now the worst performing.
The first result is intuitive as knowledge heterogeneity reduces overlap of the organizational members
knowledge. Increased knowledge heterogeneity implies that organizational members have more
specialized, or deeper, knowledge that is also narrower. As the scope of the organizational members
knowledge decreases, fewer agents possess the knowledge to match the requirements of the problems.
The matching between members and projects (or problems) becomes more difficult and therefore
induces waste of available efforts – for instance, mismatching may happen when specialized members
seek for and participate in easy projects rather than those that need their expertise. As the specialized
structure is designed to improve the matching of specialists with the requirements of the problems, it
is naturally superior. The second result is more interesting.
< Figure 2 >
Why does the unsegmented structure perform worse than the hierarchic and specialized structures?
The unsegmented structure in the previous experiment had an advantage regarding knowledge
integration because it allowed all organizational members to participate in, and contribute to, any
project. The premise was that every member had homogenous knowledge that may be applied to
solving any problem. The presence of knowledge heterogeneity, however, undermines the advantage
of the unsegmented structure. Although the members of the unsegmented structure can access every
type of project and collaborate to solve the associated problems, they are inferior to any specialist
working on a project within her scope of knowledge. The poor performance of the unsegmented
structure implies that its advantage in coordination (scope of knowledge) is dominated by a
disadvantage in specialization (depth of knowledge). This also explains why the specialized structure
performs best. The hierarchic structure falls in between. It has members who can access a broad range
16
of projects (and solve a broad range of problems) while there is also someone with a narrow but deep
knowledge specialized in a certain kind of problems or projects.
Experiment 2 is a stark setting, highlighting the benefits of specialization in a world of diverse
problems. In conjunction with experiment 1, in which the decisive requirement was on coordination
and integration of effort, the question is: How can an organization overcome the apparent tension
between specialization (knowledge differentiation) and coordination (knowledge integration), and
benefit from both as it aims to enhance its capability to deal with diverse problems? The next
experiment suggests that hierarchical participation restriction could be a useful principle when the
diverse problems an organization faces reflect the uncertain nature of its task environment and
organizational members learn from the problems they deal with.
Experiment 3: Heterogeneous Knowledge with Experiential Learning
This experiment is designed to examine how the organizational structure affects both the knowledge
accumulation process (specialization) at the individual level and the knowledge application process
(coordination) at the organization level. In contrast to the previous experiment, we are interested in
the origin of members’ knowledge and assume that they learn from their work experiences. Thus, the
organizational structure not only constrains members’ participation in projects but also allocates
learning opportunities. An important change is made to the model used in the previous experiment:
the organizational members’ knowledge vector adapts to the kinds of problems they deal with, at a
constant learning rate r (0 r ≤ 1). As described in the model section, learning about a problem type
that a member works on comes at the cost of gradually forgetting about any other kinds of problems.
For this reason, the scope and depth of the members’ knowledge will change over time. This model
change (adding experiential learning) does not alter the model dynamics under the specialized
structure because organizational members are fully specialized in only one type of problems. As the
membersknowledge remains stable over time in the specialized structure, its problem-solving rate
can be used as a benchmark to examine how the problem-solving rate of the other two structures is
affected by experiential learning.
Figure 3 shows the result of problem-solving rates corresponding to that of Figure 1 and Figure 2. The
result is quite surprising in the sense that the hierarchic structure, which in the two previous
experiments was inferior, rose as the dominating structure in dealing with diverse and difficult
problems. Furthermore, its efficiency and advantage are enhanced when organizational members learn
faster (i.e., higher r). In the environment of easy problems (i.e., low E), the specialized structure
remains superior.
17
< Figure 3 >
Why does the hierarchic structure outperform when organizational members learn from what they
work on? The intuition is that it both has the beneficial feature of the unsegmented structure discussed
in experiment 1, and the beneficial feature of the specialized structure discussed in the experiment 2.
The hierarchical structure provides the necessary tradeoff to combine specialists with the capacity of
fast problem solving and generalists who can address a broad range of problems. The former role of
the specialists accelerates problem solving when they are matched to problems that lie within their
knowledge scope. The latter role of the generalists provides flexibility in coordinating or matching
specialistsknowledge, which reduces the waste of available problem-solving efforts within the
organization. What individual-level learning adds to the value of hierarchic structure is that it helps
adapt to unpredictable shifts in knowledge demand by accordingly shape the distribution of individual
knowledge. Put differently, the hierarchic structure endogenously develops a useful distribution of
specialists and generalists as required by the joint conditions of the internal and external environments.
The detailed account of this adaptive capability of the hierarchic structure, for instance, is as follows.
Some members concentrate their efforts on particular types of problems, and therefore become
efficient at solving them. Other members spread their efforts over a range of different kinds of
problems. In this divide-and-conquer process, learning helps accelerate resolution by deepening some
of the participantscorresponding knowledge in handling the current problems, while other members
remain generalist who spread their effort over a broad range of problems. As a result, there is an
emergent, dynamic division of labor where organizational members can efficiently work in parallel on
a changing range of problems. A key role of the generalists in this adaptive process is that they help
deal with unanticipated streams of problems (e.g., highly demanding certain kinds of knowledge) that
cannot be addressed efficiently only by the specialists. Put differently, in an uncertain task
environment, the organization needs generalists to cope with exceptional situations (e.g.,
(unanticipated shift in knowledge demand), and the specialists aid in this role of the generalists by
reducing the range and amount of problem-solving and experiential learning that they need to
accomplish. In sum, individual-level learning adds to the complementarity between narrow and broad
knowledge, which enables the hierarchic structure to benefit from both specialization and
coordination.
Learning also benefits the unsegmented structure and strengthens the advantage of free or
unconstrained participation. However, it is not enough to offset the cost of excessive commitment to
solve easy projects. As organizational members prefer easy projects with problems closer to resolution,
they are likely to reduce the scope of projects they participate in and therefore make redundant or
18
wasteful efforts – a dark side of freedom. In comparison with Figure 2, the result shown in Figure 3
implies that the unsegmented structure, without any participation restriction, benefits from learning,
but does not overcome the problem of excessive commitment towards easy projects.
With diverse but easy problems, strict participation restriction based on specialization provides the
highest return. This is because an organizational member with deep specialized knowledge can easily
solve the problems within her knowledge scope without any help from other members. Put differently,
unpredictability in the external task environment does not cause that much exceptional situations (e.g.,
concentration on certain kinds of problems or knowledge demand) within the organization, which
necessitates the adaptive role of the generalists discussed above. For this reason, the premium is on
the specialized structure, which is an organization design that solves the coordination problem related
to matching specialists and problem types. Difficult problems, however, increase the cost of member-
problem mismatch and the resulting exceptional situations such as excessively skewed distribution of
problems in the organization. This is because it takes longer time for specialists to deal with the
problems within their knowledge scope, and thus unanticipated concentration of knowledge demand
on certain kinds of problems will have more long-term consequences and thus could be more
detrimental to overall organizational efficiency in problem solving.
Robustness of the Results
In this subsection, we examine the robustness of our results against the assumptions about the nature
of learning and the nature of coordination. The former is related to the issue of learning rate in
research on organizational learning, and the later to the issue of interdependence in projects (another
aspect of the organizational structure in the original GCM).
Learning rate. Figure 4 shows the problem-solving efficiency of the three organizational structures
over the full range of the learning rate parameter (r). The results confirm the positive impact of faster
learning on the advantage of hierarchic restriction of participation in the face of uncertain, diverse
problems from the task environment. In our model, learning means adaptation to the knowledge
requirements of the problems at hand, so faster learning is better (if coordination or matching is done
well). One might contrast this aspect of experiential learning in problem solving with the benefits of
slow learning in organizational search for better solutions (e.g., March, 1991), as compared to
problem-solving efficiency as the dependent variable of this study. In the problem of organizational
search (through collective learning), diversity or exploration matters so that faster learning (greater
exploitation) is not always better, but rather the (spatial or temporal) balance of exploration and
exploitation is critical for high organizational performance (Siggelkow & Levinthal, 2003; e.g.,
Tushman & O’Reilly III, 1996).
19
< Figure 4 >
Interdependence. In the baseline model, we assumed “decomposabilityin problems and knowledge,
as in the original GCM, i.e., decomposition of a complex task into simpler sub-tasks or problems. We
now relax this assumption to examine the implications of interdependence in projects. First of all,
think of two aspects of coordination: coordination across different types of members and coordination
across different types of problems. The baseline model captures the first aspect of coordination that is
formalized as joint problem solving as a result of the interplay of top-down managerial control (i.e.,
participation restriction) and bottom-up selection of projects by the participants (i.e., preferring those
closer to resolution). In Experiment 3, our model demonstrates the importance of coordination (across
different types of members to deal with a single type of problems in a project): the hierarchic structure
may outperform the specialized structure. Furthermore, we theorize the contingent role of learning
rates in determining the efficacy of coordination by showing that the advantage of the hierarchic
structure (over the specialized structure) holds only when organizational members learn fast enough
(i.e., high r). This implies that, in the presence of task environment uncertainty (turbulence in
knowledge demand), fast learning may complement the effect of top-down control for coordination,
especially when the participation of members in projects is restricted in a hierarchic manner.
In the extended model, we capture the second aspect of coordination (across different types of
problems within a project), which is necessary because of interdependence in a project (i.e.,
coexistence of diverse types of problems). We do so by relaxing the assumption of full
decomposability and allowing for a broader range of problems to be associated with a certain kind of
project, which requires the corresponding types of members (their coordination across problems) to
accomplish the project. The new model has a new structure of problem-project association that
defines the range of problem types that a certain kind of project may be composed of – this
corresponds to the “access structurein the original GCM. We assume that this structure is aligned
with the organizational structure that constrains project participation of members – i.e., an aspect of
rationality reflected in organization design. Specifically, in the extended model, problems of type p
can be associated with projects of type j p – in the baseline model, j = p. Note that this corresponds
to the “hierarchic access structurein the GCM. To be consistent with the assumptions (e.g.,
exploitation bias) in our baseline model, and in the original GCM, problems tend to get associated
with the project closest to accomplishment.
The results from the extended model (Figure 5) show that the hierarchic structure dominates,
regardless of problem difficulty and learning rate, in the presence of interdependence between
problems in projects. This is not surprising in the sense that the baseline model setting is the most
favorable to the specialized structure because problems may be associated only with the projects of
the same type and thus coordination across different types of problems is not needed within a project
20
(i.e., no interdependence, so easier coordination). Our results in Experiment 3 show that even in such
an unfavorable condition, the hierarchic structure may outperform the specialized structure when
members learn from their problem-solving experiences.
< Figure 5 >
In sum, with uncertainty in the task environment (e.g., unpredictable turbulence in knowledge
demand), the hierarchic restriction of participation – representing a form of structural differentiation
with fluid participation – may have an advantage in dynamic coordination to match projects,
participants and problems, and in coping with exceptional situations.
CONCLUSION AND DISCUSSIONS
Task allocation structures are central in the theories of organization design. This is because any
organization creates a context where its members are assigned tasks and learn from their work
experiences. In this study, we have unpacked the micro-level processes of knowledge accumulation
(experiential learning) and knowledge application (problem solving) to examine how task allocation
structures influence organizational learning. We tap the potential of the classical garbage can model
and extend it to analyze how restrictions on project participation influence differentiation and
integration of organizational membersknowledge, and, consequently, organizational efficiency in
solving the diverse, changing problems from an uncertain task environment. To isolate the effects of
problem or knowledge diversity and experiential learning, we designed three simulation experiments
to identify the most efficient task allocation structure in conditions of 1) knowledge homogeneity
(homogeneous problems and knowledge), 2) knowledge heterogeneity (diverse problems and
knowledge), and 3) experiential learning (coincidence of knowledge accumulation and application).
We found that free participation in the unsegmented structure is superior when the members
knowledge and the problems they solve are homogenous. When problems and knowledge are
heterogeneous, the design requirement is on matching specialists to problem types. This is achieved
by the specialized structure, which forgoes coordination of idle resources to force a matching of
specialized members to projects that require their expertise. Finally, we found that experiential
learning creates a dynamic problem where the double-duty of adapting the memberslevel of
specialization and matching the specialists to problem types is best solved by the hierarchic structure
(if problems are challenging).
The results from the comparative analysis of the simulation experiments provide several important
implications. First, the results, especially those from Experiment 2, confirm the conventional wisdom
regarding the merit of specialization (i.e., division of labor) and the demand on coordination that
21
follows when the divided efforts must be integrated. Our analysis also sheds light on an adaptive role
of specialized members in organizational learning and problem solving: their narrow but deep
knowledge helps the organization to adapt the knowledge of its members while efficiently dealing
with the problems at hand. This happens because highly specialized members reduce the scope of
knowledge and learning needed for other members during a certain period of time. And this makes it
easier for the generalists, and for the organization as a whole, to adapt to unforeseen shifts in
knowledge demand, because they need to learn less. For this reason, hierarchic structure has an
advantage in coping with unanticipated, exceptional circumstances. Given significant uncertainty of
knowledge demand from the task environment, knowledge heterogeneity (problem diversity) and
experiential learning are shown to be the necessary conditions for our theory to hold. From this
nuanced perspective, differentiation and integration may have a complementary, rather than
contradictory, relation under environmental uncertainty. Such relaxation of the differentiation-
integration tradeoff is facilitated by the organizing principle of hierarchical participation restriction,
which is characterized by diversity and overlap in the task scope of organizational members, and by
the experiential learning of organizational members from their task experiences.
Knowledge-Based View of the Firm: Efficiency and Flexibility of Knowledge Hierarchy
The knowledge-based view of the firm asserts that the firm exists because it provides organizing
mechanisms to facilitate knowledge generation, sharing, and integration which cannot be achieved in
the market (Foss, 2007; Ghoshal & Moran, 1996; Grant, 1996a; Kogut & Zander, 1996; Spender,
1996). To the extent that it focuses on knowledge as the most strategically important firm resource,
the knowledge-based view might be considered as an outgrowth of the resource-based view of
competitive advantage. However, it goes beyond the resource-based view by also addressing some
“fundamental concerns of the theory of the firm, notably the nature of coordination within the firm,
organizational structure, the role of management and the allocation of decision-making rights,
determinants of firm boundaries, and the theory of innovation” (Grant, 1996a, p.110).
An important theoretical challenge that remains largely unaddressed is to understand how the
elements of the organizational context such as structures, rules and routines shape the organizational
learning process and vice versa (Argote & Miron-Spektor, 2011; Levitt & March, 1988; Nickerson &
Zenger, 2004). An organization’s tasks interact with its organizational context, provide its members
with experiences, and create their knowledge. Knowledge acquired by learning is embedded in the
organization’s context and thereby changes the context as well – i.e., the interaction of the internal
and external environments.
22
This study contributes to the knowledge-based view of the firm by unpacking the interdependent
knowledge processes and by demonstrating when and why hierarchical participation restriction
which can be effectively achieved within organizational boundaries, but not in the market – enables
efficient and adaptive knowledge accumulation and application. This is essential in coping with
environmental uncertainty and heavy, turbulent knowledge demand. The mechanisms discussed above
provide us with insights into how ‘organizational knowledge’ is created, distributed, and applied
through the interactions of individuals and tasks, and how managers can influence theses knowledge
processes by revising the formal aspects of organization such as the structure of task allocation or
project participation.
The fit between the organizing principle of hierarchical participation restriction and the behavioral
assumptions of bounded rationality, experiential learning, and limited knowledge capacity suggests
that the same principle could be applied to macro-level organizations such as industries and
economies. The economy consisting of firms and markets might be structured in the same manner
along the principle of hierarchical participation restriction and knowledge distribution. If knowledge
is indeed the important source of competencies and advantages, this form may have been favored by
selection forces as it promotes and balances specialization for efficient knowledge accumulation at the
firm level and coordination for effective knowledge application at the industry or economy level.
Participation Rights vs. Participation Restrictions
In this study, we utilize the concept of participation restriction rather than participation rights. This is
to be consistent with the assumption of bounded rationality in identifying, classifying and allocating
individuals with various scope and depth of knowledge (Epstein et al., 2010; Foss, 2007; Kogut &
Zander, 1992; Miller, 2008; Simon, 1955, 1991a). Given such boundedness, participation restrictions
allow organizational members to improvise on their project participation when there are multiple
accessible projects. In contrast, the principle of participation rights does not allow for any
participation which has not been planned ex ante. The difference might be little in our simple model,
but it would not be small when we come to larger and more complex organizations in the reality such
as business firms, industries or economies.
Implications for Empirical Studies
Our theory suggests several testable empirical hypotheses: the correlation between hierarchic
restriction of participation, hierarchic distribution of knowledge or task scope, and organizational
performance. This inquiry assumes that both the degree of specialization and the distribution of
heterogeneous knowledge among organizational members matter. The empirical issue here would be
23
to measure the degree of hierarchical participation restriction and knowledge distribution. Decision
structures, role descriptions, task specifications, or work histories may be used as a useful source of
information. To enhance reliability of measurement, it would be a good way to focus on service
industries where firms are usually engaged in human-resource intensive activities based on
specialized knowledge – e.g., industries of legal, medical, financial, consulting, or education services.
Such focus is also expected to reveal more significant empirical findings.
Another application of the theory could be made to examine the implication of changes in a firm’s
boundary for the collective performance of the firms in its value chain or local industry as well as
itself. This inquiry is based on the conjecture that a value chain or industry, as a macro-level
organization, may also benefit from the principle of hierarchical participation restriction (e.g.,
hierarchically overlapped, diverse business scopes or firm boundaries). For example, we can compare
car-manufacturing companies whose value chains are different in the degree and distribution of
overlaps in production and sales activities of member firms. One value chain might consist of firms
operating on the basis of distinct scopes of specialization while another one might have firms which
collaborate in developing or integrating parts. Our theory suggests that such differences in the
organization of firm boundaries may lead to different capabilities in dealing with changing demands
for different parts/cars or innovations.
Our results may also help explain why different industries (e.g., high- vs. low-tech) exhibit more or
less variety of organizational structures. For instance, with highly difficult problems in our model, the
organizational structures make little differences in problem-solving efficiency (e.g., Figure 2 and 3).
This is because more difficult problems take more time to be resolved, and the associated projects last
longer. This limits the number of members or the amount of efforts available to the new problems or
projects arising today, and things get worse. This effect of problem difficulty is common to all
structures, and thus reduces the performance gap between different structures.
Limitations and Future Research
In our model and analysis, we deliberately abstracted from the effects of complexity or
interdependence between different kinds of knowledge or problems – we explored an aspect of
interdependence (in projects) in the sensitivity analysis, though. This was to better understand in a
simpler setting how the coincidence of knowledge accumulation and application shapes the relation
between organizational structure and problem-solving efficiency. A broad range of literatures on
organizational learning and design, however, point out that such knowledge processes may also
depend on the complexity (e.g., architecture of interdependencies) of organizational or product
24
knowledge and structure (Baldwin & Clark, 2000; Ethiraj & Levinthal, 2004; Sanchez & Mahoney,
1996; Simon, 1962). In addition, our model assumes that participants prefer to solve easy projects
close to accomplishment, an assumption that represents a form of exploitation bias. Recent studies,
however, found that incentives may impact the knowledge processes and organizational learning and
problem-solving (Baumann & Stieglitz, 2014; Lee & Meyer-Doyle, 2017). Future research on the
implications of those aspects of real-world organizations would be an important and promising way to
go.
25
REFERENCES
Argote, L. 1999. Organizational learning: Creating, retaining and transferring knowledge. Bücher:
Springer.
Argote, L., & Miron-Spektor, E. 2011. Organizational learning: From experience to knowledge.
Organization Science, 22(5): 1123–1137.
Arrow, K. J. 1962. The Economic Implications of Learning by Doing. The Review of Economic
Studies, 29(3): 155–173.
Baldwin, C. Y., & Clark, K. B. 2000. Design Rules: The Power of Modularity. Cambridge, MA: MIT
Press.
Barney, J. B. 1986. Strategic factor markets: Expectations, luck, and business strategy. Management
science, 32(10): 1231–1241.
Barney, J. B. 1996. The resource-based theory of the firm. Organization Science, 7(5): 469–469.
Baumann, O., & Stieglitz, N. 2014. Rewarding value-creating ideas in organizations: The power of
low-powered incentives. Strategic Management Journal, 35(3): 358–375.
Becker, G. S., & Murphy, K. M. 1992. The Division of Labor, Coordination Costs, and Knowledge.
Quarterly Journal of Economics, 107(4): 1137–1160.
Bendor, J., Moe, T. M., & Shotts, K. W. 2001. Recycling the garbage can: An assessment of the
research program. American Political Science Review, 95(1): 169–190.
Bunderson, J. S., & Boumgarden, P. 2010. Structure and Learning in Self-Managed Teams: Why
“BureaucraticTeams Can Be Better Learners. Organization Science, 21(3): 609–624.
Carley, K. M. 1992. Organizational Learning and Personnel Turnover. Organization Science, 3(1):
20–46.
Christensen, M., & Knudsen, T. 2010. Design of Decision-Making Organizations. Management
Science, 56(1): 71–89.
Cohen, M. D., March, J. G., & Olsen, J. P. 1972. A garbage can model of organizational choice.
Administrative Science Quarterly, 17(1): 1–25.
Cohen, M. D., March, J. G., & Olsen, J. P. 2012. “A Garbage Modelat Forty: A Solution That Still
Attracts Problems. In A. Lomi & R. Harrison (Eds.), The Garbage Can Model of
Organizational Choice: Looking Forward at Forty (Research i.). UK: Emerald Group
Publishing.
Conner, K. R., & Prahalad, C. K. 1996. A resource-based theory of the firm: Knowledge versus
opportunism. Organization Science, 7(5): 477–501.
Demsetz, H. 1973. Industry Structure, Market Rivalry, and Public Policy. Journal of Law and
Economics, 16(1): 1–9.
Demsetz, H. 1988. The Theory of the Firm Revisited. Journal of Law, Economics, & Organization,
4(1): 141–161.
Demsetz, H. 1997. The Firm in Economic Theory: A Quiet Revolution. Journal of Economic Review,
87(2): 426–429.
Denrell, J., & March, J. G. 2001. Adaptation as information restriction: The hot stove effect.
Organization Science, 12(5): 523–538.
Dierickx, I., & Cool, K. 1989. Asset Stock Accumulation and Sustainability of Competitive
Advantage. Management Science, 35(12): 1504–1511.
26
Drazin, R. 1995. Differentiation. In N. Nicholson, R. Schuler, & A. H. Van de Ven (Eds.), The
Blackwell Encyclopedic Dictionary of Organizational Behavior (137th ed.). Blackwell,
Cambridge, MA.
Epstein, A. J., Ketcham, J. D., & Nicholson, S. 2010. Specialization and matching in professional
services firms. Rand Journal of Economics, 41(4): 811–834.
Ethiraj, S. K., & Levinthal, D. A. 2004. Modularity and Innovation in Complex Systems.
Management Science, 50(2): 159–173.
Fang, C., Lee, J., & Schilling, M. A. 2010. Balancing Exploration and Exploitation through Structural
Design: The Isolation of Subgroups and Organizational Learning. Organization Science, 21(3):
625–642.
Fioretti, G., & Lomi, A. 2008. An agent-based representation of the garbage can model of
organizational choice. Journal of Artificial Societies and Social Simulation, 11(1): 1.
Foss, N. J. 2007. The Emerging Knowledge Governance Approach: Challenges and Characteristics.
Organization, 14(1): 29–52.
Ghoshal, S., & Moran, P. 1996. Bad for practice: A critique of the transaction cost theory. Academy of
management Review, 21: 13–47.
Grant, R. M. 1996a. Toward a knowledge-based theory of the firm. Strategic Management Journal,
17(winter): 109–122.
Grant, R. M. 1996b. Prospering in dynamically-competitive environments: Organizational capability
as knowledge integration. Organization Science, 7(4): 375–387.
Jacobides, M. G., & Winter, S. G. 2005. The co-evolution of capabilities and transaction costs:
explaining the institutional structure of production. Strategic Management Journal, 26(5): 395–
413.
Knudsen, T., & Levinthal, D. A. 2007. Two faces of search: Alternative generation and alternative
evaluation. Organization Science, 18(1): 39–54.
Kogut, B., & Zander, U. 1992. Knowledge of the firm, combinative capabilities, and the replication of
technology. Organization Science, 3(3): 383–397.
Kogut, B., & Zander, U. 1996. What Firms Do? Coordination, Identity, and Learning. Organization
Science, 7(5): 502–518.
Lawrence, P. R., & Lorsch, J. . 1967. Differentiation and integration in complex organizations.
Administrative Science Quarterly, 12(1): 1–47.
Lazer, D., & Friedman, A. 2007. The Network Structure of Exploration and Exploitation.
Administrative Science Quarterly, 52(4): 667–694.
Lee, S., & Meyer-Doyle, P. 2017. How Performance Incentives Shape Individual Exploration and
Exploitation: Evidence from Microdata. Organization Science, 28(1): 19–38.
Levinthal, D. A., & March, J. G. 1993. The myopia of learning. Strategic Management Journal,
14(S2): 95–112.
Levinthal, D., & March, J. G. 1981. A Model of Adaptive Organizational Search. Journal of
Economic Behavior and Organization, 2(4): 307–333.
Levitt, B., & March, J. G. 1988. Organizational Learning. Annual Review of Sociology, 14: 319–340.
Lippman, S. A., & Rumelt, R. P. 1982. Uncertain imitability: An analysis of interfirm differences in
efficiency under competition. Bell Journal of Economics, 13(2): 418–438.
Lomi, A., & Harrison, J. R. 2012. The Garbage Can Model of Organizational Choice: Looking
Forward at Forty. Emerald Group Publishing.
27
Lundberg, E. 1961. Produktivitet och rdntabilitet, (Productivity and Profitability. Stockholm,
Sweden: P.A.Norstedtand S¨ oner.
March, J. G. 1991. Exploration and exploitation in organizational learning. Organization Science,
2(1): 71–87.
March, J. G., & Simon, H. A. 1958. Organizations. New York: John Wiley & Sons.
Masuch, M., & LaPotin, P. 1989. Beyond Garbage Cans: An Al Model of Organizational Choice.
Administrative Science Quarterly, 34(1): 38.
Miller, K. D. 2008. Simon and Polanyi on Rationality and Knowledge. Organization Studies, 29(7):
933–955.
Miller, K. D., Zhao, M., & Calantone, R. J. 2006. Adding interpersonal learning and tacit knowledge
to March’s exploration-exploitation model. Academy of Management Journal, 49(4): 709–722.
Mintzberg, H. 1979. The Structuring of Organizations. Englewood Cliffs, NJ: Prentice Hall.
Mintzberg, H. 1980. Structure in 5’S: A Synthesis of the Research on Organization design.
Management Science, 26(3): 322–341.
Nickerson, J. A., & Zenger, T. R. 2004. A Knowledge-Based Theory of the Firm? The Problem-
Solving Perspective. Organization Science, 15(6): 617–632.
Osterloh, M., & Frey, B. S. 2000. Motivation, knowledge transfer, and organizational forms.
Organization science, 11(5): 538–550.
Padgett, J. F. 1980. Managing garbage can hierarchies. Administrative Science Quarterly, 25: 583–
604.
Padgett, J. F. 2013. Alessandro Lomi and J. Richard Harrison, eds.: Research in the Sociology of
Organizations, vol. 36: The Garbage Can Model of Organizational Choice: Looking Forward at
Forty. Administrative Science Quarterly, 58(3): 472–482.
Polanyi, M. 1966. The Tacit Dimension. Garden City, NY: Doubleday.
Puranama, P., Stieglitz, N., Osmanc, M., Pillutlad, M. M., Puranam, P., Osman, M., et al. 2015.
Modelling Bounded Rationality in Organizations: Progress and Prospects. The Academy of
Management Annals, 9(1): 337–392.
Rivkin, J. W., & Siggelkow, N. 2003. Balancing search and stability: Interdependencies among
elements of organizational design. Management Science, 49(3): 290–311.
Sanchez, R., & Mahoney, J. T. 1996. Modularity, flexibility and knowledge management in product
and organization design. Strategic Management Journal, 17(S2): 63–76.
Siggelkow, N., & Levinthal, D. A. 2003. Temporarily Divide to Conquer: Centralized, Decentralized,
and Reintegrated Organizational Approaches to Exploration and Adaptation. Organization
Science, 14(6): 650–669.
Siggelkow, N., & Rivkin, J. W. 2005. Speed and search: Designing organizations for turbulence and
complexity. Organization Science, 16(2): 101–122.
Simon, H. A. 1955. A Behavioral Model of Rational Choice. Quarterly Journal of Economics, 69(1):
99.
Simon, H. A. 1962. The architecture of complexity. Proceedings of the American philosophical
society, 106(6): 467–482.
Simon, H. A. 1991a. Bounded Rationality and Organizational Learning. Organization Science, 2(1):
125–134.
Simon, H. A. 1991b. Organizations and markets. Journal of Economic Perspectives, 5(2): 25–44.
28
Spender, J.-C. 1996. Making knowledge the basis of a dynamic theory of the firm. Strategic
Management Journal, 17(Winter Special Issue): 45–62.
Sutton, R. S., & Barto, A. G. 1998. Reinforcement Learning: An Introduction. MIT Press.
Thompson, J. D. 1967. Organizations in Action. New York: McGraw-Hill.
Tushman, M. L., & O’Reilly III, C. A. 1996. Ambidextrous organizations: Managing evolutionary and
revolutionary change. California Management Review, 38(4): 8–30.
Williamson, O. E. 1975. Markets and Hierarchies. New York, NY: Free Press.
Winter, S. G. 2000. The satisficing principle in capability learning. Strategic Management Journal,
21(10): 981–996.
29
Figure 1. Problem-Solving Efficiency: Homogenous Knowledge
Figure 2. Problem-Solving Efficiency: Heterogeneous Knowledge without Learning
0%
20%
40%
60%
80%
100%
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Solved Problems
Problem Difficulty
Unsegmented
Hiearchic
Specialized
0%
20%
40%
60%
80%
100%
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Solved Problems
Problem Difficulty
Unsegmented
Hiearchic
Specialized
30
Figure 3. Problem-Solving Efficiency: Heterogeneous Knowledge with Learning
(a) Slow learning (r = 0.2)
(b) Fast learning (r = 0.8)
0%
20%
40%
60%
80%
100%
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Solved Problems
Problem Difficulty
Unsegmented
Hiearchic
Specialized
31
Figure 4: Effects of Learning Rate on Problem-Solving Efficiency
(a) Problem difficulty = 0.2
(b) Problem difficulty = 0.8
0%
20%
40%
60%
80%
100%
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Problems Solved
Learning Rate
Unsegmented
Hierarchic
Specialized
0%
10%
20%
30%
40%
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Problems Solved
Learning Rate
Unsegmented
Hierarchic
Specialized
32
Figure 5: Problem-solving Efficiency in the Presence of Interdependence in Projects
(a) Slow learning (r = 0.2)
(b) Fast learning (r = 0.8)
0%
20%
40%
60%
80%
100%
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Problems Solved
Problem Difficulty
Unsegmented
Hierarchic
Specialized
0%
20%
40%
60%
80%
100%
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Problems Solved
Problem Difficulty
Unsegmented
Hierarchic
Specialized
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
The garbage can theory of organizational choice is one of the best-known innovations in modern organization theory. It also has significantly shaped a major branch of the new institutionalism. Yet, the theory has not received the systematic assessment that it both deserves and needs. We evaluate the early verbal theory and argue that it fails to create an adequate foundation for scientific progress. We then analyze and rerun Cohen, March, and Olsen's computer model and discover that its agents move in lockstep patterns that are strikingly different from the spirit of the theory. Indeed, the simulation and the theory are incompatible. Next, we examine how the authors have built upon these incompatible formulations in developing the theory further. We assess this larger program, which includes the March-Olsen version of the new institutionalism, and find that many of the problems that attended the original article have intensified over time. We conclude that a fundamental overhaul is required if the theory is to realize its early promise.
Article
Full-text available
Much of the formal modelling work in the organizational sciences relies on Herbert Simon's conception of bounded rationality, and it stakes a claim to drawing on behaviorally plausible assumptions about human behavior and action in organizations. The objectives of our review are threefold. First, we summarize the formal literature by “model families”—classes of models sharing the same analytical structure—to highlight sharply the behavioral assumptions being made. Specifically, we discuss model families involving (a) adaptation through search and learning by individual agents, (b) mutual adjustment of interacting agents, and (c) information aggregation in organizational decision-making. Second, we examine to what extent these models of bounded rationality in organizations are in fact consistent with the behavioral evidence in psychology and other related fields. Finally, we discuss opportunities for further research that strengthens the links between formal modelling in organizations research, and its behavioral foundations. In particular, we highlight the promise of experimental methods that translate organizational models to multiple-subject experiments in the behavioral laboratory.
Chapter
Full-text available
The papers collected in this volume celebrate the 40th anniversary of “A Garbage Can Model of Organizational Choice” – one of the most influential and sustained attempts to represent organizational decision-making processes in a way that accounts for generally recognized but hard to accept features of organizational life. In our overview of the volume we emphasize ways in which the garbage can model (GCM) differs from more generally accepted models of organizational decision making. We suggest that future progress in linking the GCM to specific empirical settings might be facilitated by attempts to model explicitly the interdependencies connecting participants, problems, solutions, and decision opportunities in organizations. We discuss examples of current work in which this strategy is followed in a way that is consistent with the original spirit of the model. We present the overall organization of the volume and discuss how the various chapters contribute to the further development of organizational research inspired by ideas contained in the original GCM and in some of its more recent variants and critiques.
Article
The problem of designing, coordinating and managing complex systems is central to the management and organizations literature. Recent writings have emphasized the important role of modularity in enhancing the adaptability of such complex systems. However, little attention has been paid to the problem of identifying what constitutes an appropriate modularization of a complex system. We develop a formal simulation model that allows us to carefully examine the dynamics of innovation and performance in complex systems. The model points to the trade-off between the virtues of parallelism that modularity offers and the destabilizing effects of overly refined modularization. In addition, high levels of integration can lead to modest levels of search and a premature fixation on inferior designs. The model captures some key aspects of technological evolution as a joint process of autonomous firm level innovation and the interaction of systems and modules in the marketplace. We discuss the implications of these arguments for product and organization design.
Article
The existing literature has highlighted the role of individuals in exploration and exploitation, yet our understanding of what shapes those activities at the individual level remains limited. We integrate the literatures on exploration/exploitation and incentives to examine how incentives impact individual behavior to explore new ideas or exploit existing ideas. Using novel microdata on the commercial projects of sales employees at a South Korean e-commerce firm, we find that individuals engage in relatively more exploration when performance-based incentives are weakened; yet, interestingly, this increase in exploration is driven mainly by high-performing individuals in our setting. Weakening performance-based incentives also lead to higher exploration performance via experiential learning, especially for individuals who work in complex task environments. Overall, this study contributes to the literature on exploration/exploitation by examining how individual exploration and exploitation behavior and performance is shaped by incentives. We also contribute to the incentives literature by investigating the implications of incentives in an important domain, namely, that of knowledge exploration behavior. This study also provides insights into important individual-level microfoundations of firm capabilities and performance by highlighting the important role of individuals in exploration/exploitation activities and how this role is contextualized by the incentives a firm deploys.
Article
In his classic article on exploration and exploitation, March (1991) presented an agent-based model simulating learning within an organization. His model includes mutual learning between individuals and an organizational code. We extend March's model by (1) allowing for direct interpersonal learning, (2) locating individuals in a space, which makes the distinction between local and distant search relevant, and (3) recognizing that knowledge has a tacit dimension that cannot be transmitted through codification. We also consider the implications of personnel turnover for learning over time. Our results qualify and extend March's (1991) conclusions.