INNOVATION BY COLLABORATION AMONG FIRMS. A NEW

METHODOLOGY

Building Theory from Case Study Research and Simulation Models

Paola Pisano

University of Torino, C.so Svizzera 185, Torino, Italy

Keywords: Multi Agent System, Case Study, Collaboration.

Abstract: The purpose of this work is to introduce a methodology known as “multi agent systems” (MAS) and show

how it could serve not only a similar purpose when compared to cases by using virtual artefacts instead of

real world observations, but also overcome the limitation of case study process. Besides being used for

creating new theory, this approach is also effective for teaching and transmitting knowledge in a

“maieutical” way, through experimentation on the model and cause-effect analysis of results. After

describing the paradigm itself, and how it, along with simulation, can be used in social domains, it will be

shown how it could interface the conceptual flow typical of case study analysis. Particular attention will be

devoted to the interactivity deriving from the way in which this methodology is conceived, allowing

researchers to perform “scenario analysis”, i.e.: a process of analyzing future occurrences by considering

alternative possible outcomes, after the baseline experiments obtained through the model have given

positive results. In the second part of the paper a case study is presented, obtained by employing MAS

methodology. It aims to study enterprise collaboration formation and modification: the goal is to study how

innovation management and sharing could bring to non-equity link formation among them. The model is

introduced in detail and qualitative results are analyzed, by deriving general concepts from them. Last, some

points of strength and weaknesses of this methodology are briefly underlined.

1 USING CASES TO CREATE

NEW THEORY:

INTRODUCTION

Case data represent one of many possible form of

inquiry for inductive theory building; other forms of

data include participant observation, document

analysis, in-depth interviews, field notes, etc. Many

authors have described the process for creating

theory using case approach. Glaser and Strauss

(1967) described the process giving importance to

the ideas of theoretical sampling, theoretical

saturation, overlapped coding, data collection, and

analysis. Yin (1984) structured the process analysing

the notions of case study design, replication logic,

and concern for internal validity. Miles and

Huberman (1984) concentrate their research on the

tools of tabular display of evidence particularly

helpful in the discussion of building evidence for

constructs. Miles (1979), Miles & Huberman (1984),

Kirk & Miller (1986), centred their work on topics

such as qualitative data analysis; Yin (1981, 1984)

and Mc- Clintock et al. (1979), on case study design

and Van Maanen (1988) on ethnography. The

creation of process for building theory from case

study research has been developed from Kathleen

M. Eisenhardt (1989) in contrast with Strauss (1987)

and Van Maanen (1988), more concerned on a rich,

complex description of the specific cases under

study evolve and less on generalizable theory.

According to Eisenhardt (1989, p.532-549),the steps

for building theory from case studies are

summarized in next lines.

2 BUILDING A THEORY FROM

CASE STUDY:

A TRADITIONAL APPROACH

First of all it is important to fix the research question

since it is easy to get overwhelmed by the volume of

data. The research question can shift or modify

106

Pisano P. (2010).

INNOVATION BY COLLABORATION AMONG FIRMS. A NEW METHODOLOGY - Building Theory from Case Study Research and Simulation Models.

In Proceedings of the 2nd International Conference on Computer Supported Education, pages 106-113

 SciTePress

Figure1: Building theory from case study: a traditional approach.

during the research thanks to the observation and

data analysis. Another element for building theory

from case studies is the selection of cases. For the

data collection it can be useful to employ multiple

investigators (e.g., Pettigrew, 1988). This allows the

case to be viewed from the different perspectives of

multiple observers. Analyzing data is the heart of

building theory from case studies, but it is both the

most difficult and the least codified part of the

process. However, there is no standard format for

such analysis. Quinn (1980) developed teaching

cases for each of the firms in his study of strategic

decision making in six major corporations as a

prelude to his theoretical work. Mintzberg and

McHugh (1985) compiled a 383 page case history of

the National Film Board of Canada. These authors

coupled narrative description with extensive use of

longitudinal graphs tracking revenue, film

sponsorship, staffing, film subjects, and so on.

Gersick (1988) prepared transcripts of team

meetings. Leonard-Barton (1988) used tabular

displays and graphs of information about each case.

Abbott (1988) suggested using sequence analysis to

organize longitudinal data. Next step for building a

theory is searching for case patterns. The tactic is

driven by the reality that people are notoriously poor

processors of information. They leap to conclusions

based on limited data (Kahneman & Tversky, 1973),

they are overly influenced by the vividness (Nisbett

& Ross, 1980) or by more elite respondents (Miles

& Huberman, 1984), they ignore basic statistical

properties (Kahneman & Tversky, 1973), or they

sometimes inadvertently drop disconfirming

evidence (Nisbett & Ross, 1980). The risk is that

investigators reach premature and even false

conclusions as a result of these information-

processing biases. Thus, the key to good cross-case

comparison is counteracting these tendencies by

looking at the data in many divergent ways. Shaping

hypotheses consists in systematically comparing the

emergent frame with the evidence from each case in

order to assess how well or poorly it fits with case

data. This is a two-part process involving refining

the definition of the construct and building evidence

which measures the construct in each case. This

occurs through constant comparison between data

and constructs so that accumulating evidence from

diverse sources converges on a single, well defined

construct. The central idea is that researchers

constantly compare theory and data-iterating toward

a theory which closely fits the data. A close fit is

important for building a strong theory since it takes

advantage from the new insights made possible by

data and yields to an empirically valid theory. One

step in shaping hypotheses is the sharpening of

constructs. An essential feature of theory building is

a comparison of the emergent concepts, theory, or

hypotheses with the literature. This involves asking

what is this similar to, what does it contradict, and

why. A key to this process is to consider a broad

range of literature. Examining literature which

conflicts with the emergent theory is important for

two reasons. First, if researchers ignore conflicting

findings, then confidence in the findings is reduced.

Second and perhaps more importantly, conflicting

literature represents an opportunity. The

juxtaposition of conflicting results forces researchers

into a more creative, frame breaking mode of

thinking than they might otherwise be able to

achieve. The last step for building a theory from

case study is reaching closure. Two issues are

important in reaching closure: when to stop adding

cases, and when to stop iterating between theory and

data. All the process of building theory from case

study is described in figure number 1.

From 1995 to 2007, the MAS was use for

developing different practice case in different

subjects: Tomlin, Pappas and Sastry in (1995) use

multi-agent hybrid system for analyzing a conflict

resolution for Air Traffic Management; lygeros,

Godbol and Sastry (1996) used Hybrid system for

automated vehicles; Vassileva at al. (1999)

presented a multi agent approach to design of

adaptive distributed collaborative and peer help

environments; Bonabeau (2001) used agent-based

for simulating human systems; Bonabeau (2002)

applied Agent Based Modeling for understanding

the Business Complexity; Joanna J Bryson, Yasushi

Ando, and Hagen Lehmann (2007) used Agent base

modeling to develop a case study on social

behaviors, Vagnani (2009) used MAS for studing

financial market. The construction of theory from

MAS approach usually undergoes several phases,

ranging from analytical to more practical and

technical ones. MAS „s phases are listed as follow

(fig. n 2): 0- Real world analysis and reduction; 1-

Logical framework design (interactions at a micro

level and general environmental rules. Metaphors

INNOVATION BY COLLABORATION AMONG FIRMS. A NEW METHODOLOGY - Building Theory from Case

Study Research and Simulation Models

107

Figure 2: Building theory from case study: a “MAS” approach.

Figure 3: Building theory from case study: logical steps of a “MAS” approach.

building); 2- Parameters definition (qualitative and

quantitative); 3- Simulation run and scenarios

selection and 4- Analysis of results.

The step “zero” is about studying the real

problem that is going to be represented and modeled.

Since a model is a scaled down representation of the

real world, it‟s important to identify which parts of

the system could be scarified without losing the

general behavior of the system at an aggregate level.

The first phase of modeling regards the design of

the logical framework, that will constitute the “

formal project” to be used as the basis for the next

steps. In this phase the agents must be defined, along

with the interaction rules at the micro level (agent

level) and the general environmental rules, i.e.: those

which all the agents must face. This phase is crucial,

since MAS modeling is “bottom up”, meaning that

the aggregate level is an emergent feature and

derives strictly from the premises, or better, from the

interaction among the agents, and with the

environment. Phases two is about formally defining

which parameters will be considered in the model,

and is a derivation of the previous step. After the

model is implemented, through repeated executions

and confrontation of the behavior with the real

modeled system, the numerical parameters are tuned

so that the “baseline scenario” maps exactly the

reference one. The third step is actually the most

operative and practical one; the model, in this phase,

is to be considered an artifact, in the meaning given

by H. Simon. This allows users to employ it as a tool

for social experiments, exactly how a laboratory

would be for natural sciences. In this phase the

different scenarios are defined, and simulations are

run. The last phase is analytical: the results, be them

quantitative or qualitative, are gathered and linked

with the premises through cause-effect relations. The

importance and influence of individual parameters is

tracked by means of “multi-run analysis”, i.e.: a

ceteris paribus approach consisting of changing one

parameter at a time, by leaving the others

unchanged. Besides, the aggregate behavior is

considered and analyzed at different time steps,

allowing the researcher to understand the evolution

of the system over time. The comparison of the two

methodology is represented in the figure number 3.

3 A PRACTICAL CASE: “THE

ANALYSIS OF FIRMS

COLLABORATION THROUGH

A MAS APPROACH” FOR

CREATING A THEORY

3.1 Introduction to the Practice Case

and Structure

A model in the field of enterprise management is

described in this work. Its main goal is to represent

and analyze the dynamics and interrelations among

the complex phenomenon of innovation diffusion

and firms clusters formation and modifications.

After formal description of the model, along that of

its main parameters a qualitative results are

described. The conclusion is a comparison between

case methodology and MAS methodology underline

the weakness and strength of each methodology.

CSEDU 2010 - 2nd International Conference on Computer Supported Education

108

3.2 Research Question

The research question is to demonstrate how the

innovation diffusion impacts on the collaboration

among business enterprises, by creating and

changing –qualitatively and quantitatively – ties and

partners. For analyzing this issue, it is useful, first of

all, to define some concepts used in case

developing.

3.3 Collaboration Among Firms:

Nodes and Ties

The literature on collaboration clearly demonstrates

that whilst firms collaborate for many different

reasons the most common reason to do so is to gain

access to new or complementary competencies and

technologies.The nodes can be similar or different

depending on the organizations are competitors or

works in different position the value chain. The

types of partner‟s firms engaged in collaboration

appears to be related to the type of innovation

occurring: incremental innovators rely more

frequently on their customers as innovation partners

whereas firms that have products new to a market

are more likely to collaborate with suppliers and

consultants. Advanced innovators and the

development of radical innovations tends to demand

more interaction with universities. The ties

represents the type of relationship among the actors‟;

ties could be different in structure, type and number.

The type and the number of ties could affect

collaboration‟s efficiency: for example, a

collaboration composed of relationships with

partners comprising few ties among them would

enable control for the principle partner. A

collaboration of many non-overlapping ties would

provide information benefits: in “Interfirm

cooperation and startup innovation in the

biotechnology industry” (1994), the authors Shan,

Walker, and Kogut, suggest that the number of

collaborative relationships a firm is involved in, is

positively related to innovation output while,

conversely, closed networks have been found to

foster innovation more than open ones (Coleman,

1988). Numerous other measures of strength have

also been used or proposed. These include frequency

of contact (Granovetter, Lin et al.) with strong ties

assumed to be more frequent ones; and (in research

conducted in closed populations, where perceived

relationships are studied from both sides) by mutual

acknowledgement of contact (Friedkin), with strong

ties assumed to be those acknowledged by both

parties. Other plausible indicators of tie strength

include the extent of multiplexity within a tie (noted

by Granovetter,1973), the duration of the contact,

the provision of emotional support and aid within

the relationship (Wellman), the social homogeneity

of those joined by a tie (indexed in terms of

occupational status by Lin et al.), the overlap of

memberships in organizations between the parties to

a tie, and (for closed populations) the overlap of

social circles (Kadushin). When the innovation start

to circulate, it can affect the collaboration efficiency:

firms can decide to cooperate inside the network by

developing an external exploration behavior,

meaning that a firm decides to be related to other

organizations in order to exchange competences and

innovations. Otherwise if the firm considers its

internal capability to create innovation as a point of

strength, or if the cost of external exploration is

perceived as higher than that of internal research,

then it could prefer to assume an internally

explorative behavior in which it tries to create new

competences (and possibly innovations) inside the

organization itself. During the process of innovation

diffusion the collaborations can change in the

number of actors (exit and entry), and in numbers

and patterns of link information (Steinke, 2006). The

collaborations can expand, churn, strengthen or

shrink. Each collaboration‟s change is brought about

by specific combination of changes in tie creation,

tie deletion, and by changes in an actor's portfolio

size (number of link) and portfolio range (numbers

of partners) (Steinke, 2006). Also the propensity to

collaborate affects innovation diffusion. When firms

has a highly collaborative attitude, the innovation

tends to diffuse more quickly, if the ties are dense,

non redundant, strong and reciprocal. If the firms are

collaborative, but the ties are weak or unidirectional,

the innovation spreads slowly and could not reach

all the nodes in the collaboration. To explore and

analyze these complex social dynamics, an agent

based model is described in the following

paragraphs, that keeps into account most network

and enterprise variables.

3.4 Logical Framework Design

Agent based simulation is an effective paradigm for

studying complex systems. It allows the creation of

virtual societies, in which each agent can interact

with others basing on certain rules. The agents are

basic entities, endowed with the capacity of

performing certain actions, and with certain

variables defining their state. In the model presented

here, the agents are reactive, meaning that they

simply react to the stimuli coming from the

INNOVATION BY COLLABORATION AMONG FIRMS. A NEW METHODOLOGY - Building Theory from Case

Study Research and Simulation Models

109

environment and from other agents, without

elaborating their own strategies. Agents have

traditionally been categorized as one of the

following types (Grandori, 1997): Reactive;

Cognitive/Deliberative; Hybrid. The agents used in

this paper are reactive, but organized in the form of a

MAS (Multi Agent System), which can be thought

of as a group of interacting agents working together

or communicating among each other. To maximize

the efficiency of the system, each agent must be able

to reason about other agents' actions in addition to its

own. A dynamic and unpredictable environment

creates a need for an agent to employ flexible

strategies. Many simulation paradigms exist; agent-

based simulation is probably the one that best

captures the human factor behind decisions. This is

because the model is not organized with explicit

equations, but is made up of many different entities

with their own behavior. The macro results emerge

naturally through the interaction of these micro

behaviors and are often more than the algebraic sum

of them. This is why this paradigm is optimal for the

purposes of modeling complex systems and of

capturing the human factor. The model presented in

this paper strictly follows the agent based paradigm

and employs reactive agents, as detailed in the

following paragraph.

3.5 The Model

The model is built in Java, thus following the Object

Oriented philosophy and has been engineered and

built at the e-business L@B, University of Turin. All

the numerical parameters can be decided at the

beginning of each simulation (e.g.: number of

enterprises, and so on). Everything in the model is

seen as an agent; thus we have three kinds of agents:

Environment, Enterprises and Emissaries (E³). This

is done since each of them, even the environment, is

endowed with some actions to perform.

3.6 Heat Metaphor and the Agents

In order to represent the advantage of an enterprise

in owning different competences, the “heat”

metaphor is introduced. In agent based models for

Economics, the metaphor based approach

(Remondino, 2003) is an established way of

representing real phenomena through computational

and physical metaphors. In this case, a quantum of

heat is assigned for each competence at each

simulation turn. If the competence is internal (i.e.:

developed by the enterprise) this value is higher. If

the competence is external (i.e.: borrowed from

another enterprise) this value is lower. Heat is also

expendable in the process of creating new internal

competences (exploitation) and of looking for

partner with whom to share them in exchange of

external competences (exploration). At each time-

step, a part of the heat is scattered (this can be

regarded as a set of costs for the enterprise). If the

individual heat gets under a threshold, the enterprise

ceases its activity and disappears from the

environment. At an aggregate level, average

environmental heat is a good and synthetic measure

to monitor the state of the system. In order to

formally describe the model, a set of equations is

described in the following. The multi agent system

at time is defined as:

MAS

=< E



, e, ε



, link



> (1)

Where E



represents the environment and is

formed by a grid n m, and a set k







=< , k



n, m > 0



. (2)

Where the set k



definines the heat for each cell, e

is the set of enterprises with cohordinates on the

grid, and ε



is the set of the emissaries, also scattered

on the grid:













=< k

i,j

e=< e

′



=< ε

′′

0 < , i

′

, i

′′

n

0 < , j

′

, j

′′

m



. (3)

Each enterprise is composed by a vector c, and

an emissary (ε

). The vector c defines the owned

competences, with a length L and competences C

represented by a boolean variable (where 1 means

that the l

competence is owned, while 0 means that

it‟s lacking):







i,j

c, ε

c=



L, C



0 l L

= Boolean



(4)

In T = t > 0 , k

i,j

that‟s the heat of each cell on

the grid, depends on the heat produced by the

enterprises (K

) and the dispersion effect ( d). The

heat of each enterprise is function of the

competences it possesses and of the behavior it

carried on in the last turns (b

CSEDU 2010 - 2nd International Conference on Computer Supported Education

110







i,j

= f(K

, d)

= f



c

, b



b b



=<   >



(5)

In particular, a certain behavior can be

successful, meaning that at the end of a phase of

internal or external exploration, a new competence

(internal or outsourced, respectively) will be

possessed. Otherwise, a it‟s unsuccessful when, after

some steps of research and development (internal

exploration) or external market research to find a

partner, nothing new is found, and thus the l

competence remains zero.







b = success



then C

= 1

else C

= 0



b b





(6)

At each time-step the set of links (connecting

two enterprises together) is updated basing on the

competences of the enterprises.



link



=< e

i,j

, e

′

>

link e

i,j

, e

′

= f(c

i,j





, c

′





)



(7)

Specifically, when an enterprise does external

exploration, it looks for a good partner, i.e.: an

enterprise with a number of competences to share.

So, if an enterprise with a vector like

1 0 0 0 1 meets one with a vector like

0 1 1 1 0 then there is a perfect match and the

two enterprises will create a link among them, to

share the reciprocally missing competences. This is

the perfect situation, but not the only one in which

two enterprise can create a link; in fact, it‟s enough

that there is at least one competence to reciprocally

share. The strength of the link is directly

proportional to the exchanged competences. This set

of equations and rules is enough to explore the

effects on the network of the behaviors of the

enterprises, namely the way in which the firms are

managed (externally or internally focused). Though

the model allows also to explore the effects on

innovation (i.e.: a competence that‟s possessed only

by one enterprise). In T = t

′

>  a radical innovation

can be metaphorically introduced in the system (this

is called “shock mode”, since this is decided by the

user, at an arbitrary step) by means of increasing the

length of the vector of competences of a specific

enterprise:



L L + 1

l+1





= 1

l+1



ee



= 0



. (8)

Meaning that the competence C

l+1

will be

possessed by only one enterprise, at that time, while

the same competence will be lacking to all the

others; though, all the enterprises‟ vectors will

increase in length, meaning that potentially all of

them will be able to internally develop that new

competence through R&D, from then on. The vector

length metaphorically represents the complexity of

the sector (industry) in which the enterprises

operate; an highly technological sector has many

more potential competences than a non-

technological one. So, another kind of “shock effect”

to the system is that of increasing the length of the

vector by more than one component, and by leaving

all the new components to zero for all the

enterprises. In this way, they‟ll have to develop

themselves the new competences by means of

internal exploration. The analysis phase is carried on

after several steps after t

′

, in order to see how the

introduction of the innovation impacted the network

and the enterprise in which the innovation was first

introduced. So we have an analysis phase in

T = t

′′

> t

′

defined as:



MAS

t′

vs MAS

t′′

I dθ link; dθ e; dθ k



(9)

Namely, the comparison among the system at

time t

′

and the same system at time t

′′

, since the

innovation has differential effects on the number

(and nature) of the links, on the number of

enterprises and the heat of the cells composing the

environment, always depending on the managerial

behavior of the involved enterprises. At the

beginning of a simulation, the user can change the

core parameters, in order to create a particular

scenario to study and analyze.

At the beginning of a simulation, the user can

change the core parameters, in order to create a

particular scenario to study. Some of the parameters

are constituted by a scalar value, others are in

percentage, others are used to define stochastic

(normal) distributions, given their mean value and

their variance.

3.7 Analysis of Results

The impact of innovation diffusion on the network

depends on the collaboration degree of the system. If

the network is collaborative the diffusion of

INNOVATION BY COLLABORATION AMONG FIRMS. A NEW METHODOLOGY - Building Theory from Case

Study Research and Simulation Models

111

innovation strengthens the ties and increases the

number of the links among organizations (figure 4).

Figure 4: A sample of quantitative output E³ simulation

model.

4 CONCLUSIONS

As for every methodology, creating and developing

a theory by using case study features points of

strength and weaknesses. One point of strength of

theory building from cases is its likelihood of

generating novel theory. The likelihood of valid

theory could be high because the theory-building

process is so intimately tied with evidence that it is

very likely that the resultant theory will be

consistent with empirical observation. Besides it's

possible to create more theories starting from the

same case, and a case can be analyzed under

different viewpoints. Obviously this is made

possible thanks to the many acquired underlying

data. Regarding the points of weakness, an

important one is that the intensive use of empirical

evidence can yield to theory which is overly

complex. The result can be a theory which is very

rich in details, but lacks the simplicity of an overall

perspective. Another weakness of cases is the

impossibility of repeating the case if the first

analysis is not coherent with the research question.

Another point of weakness is that building theory

from cases may result in narrow and idiosyncratic

theory. Case study theory building is a bottom up

approach such that the specifics of data produce the

generalizations of theory. The risks are that the

theory describes a very idiosyncratic phenomenon or

that the theorist is unable to raise the level of

generality of the theory. For example, Gersick

(1988) presented a model of group development for

teams with project deadlines, Eisenhardt and

Bourgeois (1988) developed a mid-range theory of

politics in high velocity environments, and

Burgelman (1983) proposed a model of new product

ventures in large corporations. Such theories are

likely to be testable, novel, and empirically valid,

but they do lack the sweep of theories like resource

dependence, population ecology, and transaction

cost. Perhaps "grand" theory requires multiple

studies-an accumulation of both theory-building and

theory-testing empirical studies. Finally the case

study will be linked always to the practice case that

develop, soak in the context and scenario that

describe. The use of MAS as a methodology for

analyzing real world situations and creating new

theory, thus moving from the particular scenario to

the general case. First of all, since the model has to

represent a scaled down situation, and not the whole

reference system, it‟s quite easy to track down the

data necessary to build the reference scenario. This

reflects also on the fact that a limited range of real

world data is used, thus preventing misleading

aggregate results deriving from too many data,

contributing to create white noise during the

traditional analysis. Since the agent based

methodology relies on metaphors, it‟s potentially

possible to represent any social situation, if the

proper computational transitional function is found

and implemented. For example, in the model

presented in this work, the heat metaphor is used to

evaluate the general health of the system, and is the

unity of measurement, the payoff and the cost that

the enterprises must face during their own business.

This is easily translatable into formal programming

language, since it‟s based on physics and thus on

mathematical functions. So it‟s up to the

creativeness of the designer and any case study

could be potentially recreated in a dynamic and

interactive way. The most important feature of a

model based on agent is the possibility of repeating

the experiment several time, by changing one or few

variables at a time, by leaving the other ones

unchanged. This is referred to as “what if” analysis

or “ceteris paribus” methodology. This has a double

worthiness: on the one side, this can be used to track

the cause-effect relationships among variables and

results. On the other, it can be used to fine tune the

results in order to make it as reliable as possible,

when compared to real world ones. In social

observations, this kind of approach would be

impossible. Human factor and changing context

would simply make it unworthy to replicate

experiments or measurements, unless the confidence

interval is kept at a very large range. This leads to

the fact that from one base scenario, other scenarios

can be created, if the right parameters are changed.

This allows different case study, by starting from the

same model. Another interesting possibility, when

using MAS to represent social systems, is to create

qualitative situations that have not been directly

studied in the real world, through data harvesting.

This means that, when observing a trend in a real

CSEDU 2010 - 2nd International Conference on Computer Supported Education

112

scenario, this could be represented in the

computational model; after that, data can be

collected from the model itself and studied, as if

they came from the real situation, if the overall trend

has been respected. This is the case of such

situations heavily dependent from randomness or,

simply, from too many variables to be tracked in the

real world case. Last but not least, models based on

MAS have an important educational power; ranging

from simple models, that could be perceived as

games (e.g. business games) to be used into schools

and universities, all the way up to complex models

to be used for implicit knowledge formalization,

knowledge transfer and management within

enterprises. The “maieutilcal” approach allowed by a

model of this kind is evident when dealing with

organizational theories about Management and

Economics: students can “learn by doing” using the

model as an artifact on which carrying on their own

experiments, thus directly discovering theories,

without simply studying them by heart, and taking

them as “dogmas” coming from books. In this way,

the model becomes a virtual laboratory and the

experiments can be done in a supervised (by

teachers) or unsupervised way by the learners.

This approach doesn‟t want to substitute the

practice case but just to integrate and overcome the

limitation of practice case approach for supporting

the creation of new economic theory.

REFERENCES

Abbott, A. 1988. Workshop on sequence methods.

National Science Foundation Conference on

Longitudinal Research Methods in Organizations,

Austin.

Bonabeau E. ,2001. Agent-based modeling: Methods and

techniques for simulating human systems. Icosystem

Corporation, 545 Concord Avenue, Cambridge, MA

02138.

Bettenhausen, K., & Murnighan, J. K.,1986. The

emergence of norms in competitive decision-making

groups. Administrative Science Quarterly, 30, p.350-

372.

Coleman, J., 1988. Social capital in the creation of human

capital. American Journal of Sociology, 94, 95.

Evolution, Yin, Robert K., 1998: Case Study Research:

Design and Methods, 2

ed., Sage Publications,

Beverly Hills, CA.

Gersick, C. ,1988.Time and transition in work teams:

Toward a new model of group development. Academy

of Management Journal, 31, p.9-41.

Grandori, A. ,1997. An organizational assessment of

interfirm coordination modes. Organization Studies,

18 (6), p.897.Grandori, A., Soda, G. 1995. Inter-firm

networks: Antecedents, mechanisms and forms.

Organization Studies, 16 (2), p.183.

Granovetter, Mark S. 1973.The Strength of Weak Ties.

American Journal of Sociology, 78(6) 1360-1380.

Huberman, A. M., 1984.Qualitative data analysis. Beverly

Hills, CA: Sage.

Jick, T. ,1979.Mixing qualitative and quantitative

methods: Triangulation in action. Administrative

Science Quarterly, 24, pp. 602-611.

Kadushin, C,Delmos J.,1992. Social Networks and Urban

Neighboorhoods in New York City.City & Society

6(1):58-75.

Miles, M.,1979.Qualitative data as an attractive nuisance:

The problem of analysis. Administrative Science

Quarterly, 24, pp.590-601.

Mintzberg, H.,1979.An emerging strategy of "direct"

research. Administrative Science Quarterly, 24,

pp.580-589.

Nisbett, R., & Ross, L.,1980.Human inference: Strategies

and shortcomings of social judgment. Englewood

Cliffs, NJ: Prentice-Hall.

Pettigrew, A.,1988.Longitudinal field research on change:

Theory and practice. National Science Foundation

Conference on Longitudinal Research Methods in

Organizations, Austin.

Phlippen, S., Riccaboni M., 2007. Radical Innovation and

Network evolution, the effect of genomic revolution of

the evolution of pharmaceutical R&D network.

Podolny, J.M. 2001. Networks as the pipes and prisms of

the market.American Journal of Sociology, 107:

pp.33-60.

Quinn, J. B.,1980. Strategies for change. Homewood. IL:

Dow-Jones Irwin.

Remondino M., Correndo G., 2006. MABS Validation

Through Repeated Execution and Data Mining

Analysis, International Journal of SIMULATION:

Systems, Science & Technology (IJS3T) 7,

Shan, W., Walker, G. and Kogut, B.,1994. Interfirm

cooperation and startup innovation in the

biotechnology industry. Strategic Management

Journal, 15, 387.

Vagnani, M, 2009. The Black-Scholes model as a

determinant of the implied volatility smile: A

simulation study, Journal of Economic Behavior and

Organization, Volume 72, issue 1, pp. 103-118.

Woolridge, M., & Jennings, N.,1995. Intelligent agents:

Theory and practice. Knowledge. Engineering Review

,10(2) p.115-152.

Van Maanen, J.,1988.Tales of the field: On writing

ethnography. Chicago: University of Chicago Press.

Kahneman, D., & Tversky, A.,1973.On the psychology of

prediction. Psychological Review, 80, p.237-251.

Yin, R.,1984,.Case study research. Beverly Hills, CA:

Sage Publications.

INNOVATION BY COLLABORATION AMONG FIRMS. A NEW METHODOLOGY - Building Theory from Case

Study Research and Simulation Models

113