CONDITIONS FOR LONG LASTING SUSTAINABLE

INNOVATION IN AN AGENT-BASED MODEL

Luca Ansaloni

, Marco Villani

1,2

, Roberto Serra

1,2

and David Lane

Modena and Reggio Emilia University, via Allegri 9, I-42100 Reggio Emilia, Italy

European Centre for Living Technology, Calle del Clero, 2940, 30124 Venice, Italy

Keywords: Innovation model, Novelty creation, Agent-based model, Economic agent-based models.

Abstract: During the last decades, innovation has become a hot topic in a variety of socio-technological contexts: in

particular, a key problem is that of understanding its origins. Moreover, scientists are not able to evaluate

the sustainability of innovation processes, and it is difficult to discover what sort of conditions might lead to

their crisis and even collapse. In this paper we present a model where agents are able to create new artifacts

and can develop and enact strategies able to sustain innovation for very long periods. We discuss some

results and make observations useful for understanding the processes and the strategies that sustain the

growth of diversity in social and technological organizations.

1 INTRODUCTION

During the last several decades, innovation has

become a hot topic in a variety of social and

technological contexts, including technology itself,

commerce, social systems, economic development,

and policy construction. There is therefore a wide

range of approaches to innovation in the literature

(Fagerberg et al. 2004). In this paper we focus our

attention on socio-technological systems where the

changes are deliberately introduced by agents, which

design artifacts on the basis of specific goals. In the

systems we consider innovation is typically

understood as the successful introduction of

something new, as for example new objects,

methods, techniques, or practices or new or modified

products and services, whose functionality is

determined endogenously, that is, within the system

itself.

At this level of abstraction, the agents could be

software agents interacting in artificial environments

or (groups of) human beings or organizations in the

real world. What is important is that the agents have

the capacity, supported by their internal

sophisticated cognitive and communication

structures, of creating and modifying artifacts. Aim

of this work is that of identifying the minimal

structures and the strategies (if any) that the agents

need in order to achieve a long lasting sustainable

growth of the system.

Modeling such innovation processes is a difficult

challenge, involving many non-linearly interacting

elements. Indeed, human societies consist of large

numbers of agents (human beings or organizations

composed of human beings) involved in distributed

sparse interactions, mediated by the presence of

artifacts (tangible, as chairs and cars, or intangible,

as languages and services). These interactions give

rise to macroscopic regularities such as trading

relationships, protocols, widely accepted duties or

technological innovations, which in turn feed back

into the structure of agents

interactions. The result is

a complex dynamical system composed of recurrent

causal chains connecting agent behaviors,

interaction networks, and collective outcomes.

Similar patterns of interaction may emerge also in

artificial worlds, in which sophisticated software

agents engage in autonomous interaction streams,

through which they seek to invent new kinds of

artifact.

In order to integrate a new kind of artifact into

the already existing patterns of interaction there

must be a certain degree of convergence of agents’

attributions about the new artifact’s identity (that is,

about its properties and functionalities). Several

agents have to align themselves around its use, by

building or modifying other artifacts in order to

combine with it and in such a way support the new

functionalities. If this happens, the invention

becomes an effective innovation (that is, an object

410

Ansaloni L., Villani M., Serra R. and Lane D..

CONDITIONS FOR LONG LASTING SUSTAINABLE INNOVATION IN AN AGENT-BASED MODEL.

DOI: 10.5220/0003175104100417

In Proceedings of the 3rd International Conference on Agents and Artiﬁcial Intelligence (ICAART-2011), pages 410-417

ISBN: 978-989-8425-41-6

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

embedded in patterns of use, potentially able to

foster the growth of new active zones in artifact

space). The agents can use artifacts created by other

agents: this fact allows the occurrence of so-called

exaptation processes (Gould and Verba, 1982), very

often observed in socio-technological systems

(Villani et al. 2007) (Villani et al. 2009) (Villani et

al. 2010).

The reciprocal feedbacks between microstructure

(agents and artifacts) and macrostructure

(organizations) has long been explicitly recognized

as of fundamental importance for social sciences

(Hayek, 1948) (Olsen, 1975) (Schelling, 1978)

(Smith, 1937), but they are relatively new topics in

artificial agents research area. For long time

scientists have lacked the tools to quantitatively

model these feedbacks, nor could they deal with

their complexity. The most salient characteristic of

traditional quantitative models on these topics,

derived from economic or physical researches, is

their top-down construction: frameworks such as

fixed decision rules, common knowledge, mean field

and equilibrium assumptions occupied the greatest

part of the researches. Face-to-face interactions

among heterogeneous economic agents typically

play no role, with the only exception of the highly

stylized game tournaments (Fudenberg and Tirole,

1991) (Dutta 1999).

A major advance was the introduction of agent-

based models (Lane 1993a) (Lane 1993b) (Epstein

and Axtell, 1996) (Gilbert and Terna, 2000)

(C.Cioffi-Revilla 2002) (Ormerod et al. 2002)

(Axelrod and Tesfatsion, 2006). These models deal

with the topic of coordination and cooperation

among heterogeneous agents, often lacking a

complete knowledge of the whole system; the

models aim to bridge the gap between micro-level

interactions and emerging patterns at the macro-

level, avoiding the misleading “representative agent”

micro-macro link.

Agent-based models are very useful tools, but

many of them in the innovation context

underestimate the specific role of agents and, even

more, the attributions that agents make about artifact

identity, as noted above (Lane and Maxfield 1997)

(Lane et al. 2005). Agents and artifacts interact in

complex ways, giving rise to the so called socio-

technological systems.

In fact, one of the most intriguing observations

on these kinds of system is the growth of the

quantity and diversity of artifacts that agents use:

over time, not only the quantity and the diversity of

artifacts has grown, but also the number and kinds of

organizations has increased. These two phenomena

are in reciprocal relationship (van der Leeuw et al.

2009); both phenomena contribute in important

ways to the system’s information coordination and

processing capabilities. In particular, in the actual

world, a high rate of innovation seems to be a

peculiar and fundamental key to sustain the systems

itself.

But can the current explosion in number of

artifacts and organizational forms continue

indefinitely? How can agents lacking a global vision

of the whole system coordinate their actions, in

order to cooperate in building a coherent system?

Are there agents’ strategies that favor a sustainable

growth, and others that lead to system collapse? In

order to address this question, we need to understand

the dynamics of innovation processes.

In this paper we make use of an agent-based

model, where the relationships among agents are

mediated by the presence of artifacts. Agents

endowed with a suitable internal structure survey the

opportunities offered by their social and material

environment to create new (kinds of) artifacts, which

in turn change and shape the present pattern of

interactions heavily influencing the emergence of

new agents-artifacts (sub)systems. This kind of

approach has already provided some interesting

results highlighting the importance of relationships

among agents, which can influence the information

flows through the system (Lane et al. 2005) (Serra et

al. 2009) (Villani et al. 2007) (Villani et al. 2008). In

this paper, we describe four scenarios, which taken

together indicate that the conditions enabling a long

lasting sustainable growth are neither simple nor

widespread. The paper is organized as follows. The

second section provides a detailed introduction to

the basic innovation model. The third section

describes results obtained by exploring four different

innovation theoretic scenarios; the fourth section

presents some conclusions derived from simulations

based upon these scenarios.

2 THE MODEL

2.1 Agents and Artifacts

There are numerous approaches to studying

innovation dynamics, but few of them attempt to

construct models in which the reciprocal causality

between transformations in the space of artifacts and

organization in the space of agents plays an essential

role. Rather, most models assume that only artifacts

matter (for example, theories of technological

trajectories), whereas others are agent-centric, based

CONDITIONS FOR LONG LASTING SUSTAINABLE INNOVATION IN AN AGENT-BASED MODEL

411

on the idea that creativity or knowledge is the key

factor underlying innovation dynamics (Dosi 1982)

(Schumpeter 1934) (March 1991).

The model with which this paper is concerned is

based on a theory of innovation developed in (Lane

and Maxfield 1997) (Lane and Maxfield 2005). It

represents a simplified world, inhabited by highly

abstract representations of real world agents,

artifacts, and attributions. The aim of the model

therefore is not to describe in detail a real innovation

context providing quantitative predictions: rather, its

purpose is that of identifying the feedbacks and the

causal connections implicit in the theory and useful

in describing certain kinds of qualitative behaviours

of real innovation contexts.

A claim of the theory is that agents and artifacts

are both important for innovation, because artifacts

mediate interactions between agents, who in turn

actively produce and manipulate the knowledge

needed to make effective the artifacts’

functionalities.

A key point is the representation of artifacts and

their combinations. For modelling purposes, we

have considered different alternatives: binary coding

as in classifier systems; -calculus as in the

Alchemy model (Fontana, 1992) (Fontana and Buss

1994); or simply numbers, either natural or real,

with functions to describe interactions. What is

required is that the space has an algebraic structure,

and that suitable constructors can be defined to build

new artifacts by combining existing ones. We

concentrated on the integer number representation

and the use of arithmetic or other simple functions as

operators, because it is more compact than the

binary representation and simpler than the -

calculus.

Despite this very simple representation, the real

meaning of an artifact is not trivial, since it is

determined not by the thing “in itself,” but by which

agents use it, and for what. For that reason, in the

model the same entity is representing:

 a type of artifact, i.e., the “idea” – or archetype –

of the article the producer is making (for example,

the platonic idea of a chair – or of the number “12”

in a particular model run): the artifact “name” in the

following;

 the artifact(s) a particular producer is making

(the article a particular producer is making and

offering to other agents): “article” in the following;

 a single artifact token (a single chair present in

the stock of an article): “item” in the following.

The entities manipulated by our algorithm are the

articles, which in the model have a unique identifier

and a stock.

The intelligent part of the system is embedded in

the agents, endowed with sophisticated cognitive

and communication capabilities. In particular, agents

1. can explore their environment (composed of

articles and other agents);

2. can manipulate articles (in order to build other

articles);

3. can choose their goal (a particular name);

4. can use their knowledge in order to reach their

goals.

Agents’ capabilities are finite; therefore, they are

not manipulating all the articles present in their

world, nor know the goals of other agents. The role

of agents is defined by what they do, and by the

other agents with whom they interact. Agents have

not a complete information, and this situation

heavily influences their behaviour. Agents have to

identify useful goals and pursue them; in so doing,

they may or may not collectively build a sustainable

world.

Note that at this abstract level this description

applies both to living systems and to totally artificial

systems. The topic with which we are dealing

therefore embraces the more general theme of

coordinating many different agents that can

manipulate and interact with their environment, also

by introducing new objects. The new objects could

be tangible or intangible; the model we present here

however explores worlds where these objects (the

artifacts) are countable - the simplest and most

common situation.

Now we can describe the agents’ internal

structure. In this model we aim to identify the

simplest set of structures and strategies needed to

assure the agents’ functionality, so an agent:

1. can detect the presence of (a subset of the)

already existing articles and agents;

2. can manipulate some article by means of

“recipes” (ordered sequences of article identifiers

and production operators), producing other articles;

3. can identify goals, derived by its world

knowledge;

4. can manipulate its recipes in order to build new

recipes, producing the article that match their goals.

The most complicated structure owned by an

agent is the recipe, the tool it uses to process the

items it obtain from other agents in order to produce

the items of its output articles. In the experiments

reported here, recipes employ the arithmetic

operators “+” and “-”.

ICAART 2011 - 3rd International Conference on Agents and Artificial Intelligence

412

An agent can possess more than one recipe at the

same time. If the stock of a particular recipe is lower

than a desired level (10 items in the following

simulations) the agent repeats the production action

a finite number of times (whose maximum level here

is set to 10). If the stock exceeds this level, the agent

decreases production; finally, in each step all the

recipes owned by one agent have to be produced at

least once, if inputs are available (the agents need

artifacts in order to survive). A recipe that during the

last 15 time steps is not produced (because the

needed input names are not available in the system)

or not used (because no other agents have taken

items from its stock) is removed from the

simulation: in such a way “useless” articles

disappear.

In order to achieve their goals, the agents can

create new recipes. Once the goal is set, with

probability p

inn

the agents try to build a recipe able

to produce an article having a name close (in some

metric) to the goal. Several optimization strategies

can be applied to this aim: in the implementation

used here, the agents combine their existing recipes

by means of genetic algorithms, whose fitness

function is the inverse of the distance between the

goal and the realized name. If the genetic algorithm

is not able to create a recipe building an article

whose name is within a given distance from the

goal, the invention process fails and the agent

doesn’t reach the goal (see (Serra et al. 2009) for

further details).

The goal is the name the agent is trying to

realize; it can be maintained until the building of an

article with the same name succeeds, or it can be

changed at each step with a given probability Pgoal.

The agents could use several strategies in order to

set their goal: in the following simulations the agents

randomly keep the name of one of the already

existing articles and occasionally mutate it (by

multiplying it by the value C_jump the 30% of the

times a new goal is set – an action that correspond to

a “jump” in the artifacts’ space). Note that in such a

way the choice of the goal is influenced by the

number of articles that have the same name: replicas

therefore are not negligible.

The systems’ environment is very simple, and is

constituted by a set of articles (the “raw materials”)

whose stock is unlimited, the interesting study of

systems in which raw materials have a production

limit being postponed to further works.

2.2 Dynamics and Novelty Generation

A typical run (see also (Serra et al. 2009) and (Vil-

lani et al 2007)):

1. creates of a set of initial conditions (N agents

having 2 recipes each)

2. repeats until n_passimax_passi

a. sets Count=0

b. repeats until CountN

i. the random choice of an agent (add 1 to

Count)

ii. the determination of a new goal (with

probability P

inn

) and its realization

iii. the production of the actual agent’s

recipes

iv. the increase of Count by one unit

3. final visualizations

It is possible that some stocks become empty,

since very often several recipes make us of the same

articles; in this case these recipes have to change

provider, by finding a new one producing an article

with the same name and a non-empty stock. This

process has several interesting consequences:

 new articles, just built, have the possibility of

being used (so allowing their inclusion on the

already existing patterns of interaction);

 articles having the same name could be realized

by different recipes, combining different set of

articles; a frequent change of providers allows

therefore the existence of a highly heterogeneous

mixture of artifacts, favouring high diversity in the

systems;

 cycles composed of articles and agents can be

formed, and can become the source of long-lasting

patterns of interaction (each article of value for the

next one).

The continuous creation of new recipes making

use of the already existing articles, combined with

the change of providers, allows the formation of new

(groups of) cycles, stabilizing in this way these new

parts of the system. As a first conclusion, the change

of provider seems the key feature enabling (directly

or indirectly) the stabilization of the innovations, by

means of the consequent formation of cycles.

2.3 Typical Behaviors

The model provides the basic elements for a suitable

description of the creation and stabilization of

innovations. Table 1 and fig.1 show the parameters

and a portrait of a typical scenario, where the

number of artifacts (fig.1a), the diversity (the

number of different names present at a given step -

fig.1b), the typical recipes’ production level (fig.1c)

CONDITIONS FOR LONG LASTING SUSTAINABLE INNOVATION IN AN AGENT-BASED MODEL

413

and the diameter (the difference between the

maximum and the minimum names - fig.1d) reach a

stable situation.

Table 1: The main model parameters and their values in

the standard case.

(a) (b)

Figure 1: Behaviour of 10 systems initially composed of

40 agents, each agent having 2 recipes. Each plot reports

the smallest, average, median and biggest value of each

variable vs. time. (a) Number of articles; (b) diversity; (c)

production levels; (d) diameter in the artifact space.

3 RESULTS

Fig.1 shows that the systems’ diversity can undergo

a considerable growth. However, there are strategies

that can be adopted in order to significantly increase

the number of articles and their diversity. We

propose here four different scenarios, able to support

different diversity levels. Besides the standard one

(P_stand scenario, with p

inn

=0.2), we can

significantly increase the agents’ innovation

probability (p

inn

=1.0, P1_0 scenario), enable a

feedback between a measure of the agents’ size (as

for example the number of their recipes) and the

innovation probability (P_chang scenario); or finally

we can compare these scenarios with a situation

where the goal setting is random (p

inn

=0.2, P_rand

scenario).

(a) (b)

Figure 2: Median on 10 different runs of variables

characterizing the 4 scenarios described in the text. (a)

Number of articles; (b) diversity - the number of different

names existing in the system; (c) the production levels; (d)

the diameter in the artifact space (note the use of

logarithmic scale in this plot).

As we can see in fig.2, the different scenarios:

 produce significantly different quantities of

articles;

 support very different diversity levels;

 are able to maintain different recipes’ production

levels;

 explore different portion of the artifacts’ space.

To these differences there correspond very different

structures of the artifact space, as it is shown in fig.3

for typical runs.

(a) (b)

Figure 3: Artifact space of a typical run of the 4 scenarios

described in the text. (a) random goals (b) standard

situation; (c) p

inn

=1.0 scenario; (d) p

inn

dependent on the

number of recipes owned by each agent. Note the use of

different scales in (a) and (c) cases.

ICAART 2011 - 3rd International Conference on Agents and Artificial Intelligence

414

From data in fig.2 and fig.3 we can draw some

significant observations:

 the random system

o has high diameter size,

o uniformly covers the explored area,

o support high diversity levels;

o has low recipe production levels;



the introduction of an imitative goal setting, with

respect to the random system

o strongly reduces the explored area;

o reduces the diversity level;

o allows high recipe production levels,

o allows high number of articles;

 a very strong innovation

o restores the previous diameter, diversity and

recipes production levels

o by maintaining a small covered area and

shrinking it to some restricted “levels”;

 the feedback between agents’ size and innovation

probability allows

o long lasting periods of diversity growth;

o long lasting periods of growth of the number of

articles;

o high numbers of articles

o very high levels of article production.

Further clues can be inferred from table 2, which

shows the efficiency of the scenarios in integrating

the novelties into the already existing patterns of

interactions. As we can see, in systems with high

diameters, the innovation processes have high

percentages of failure and produce a great number of

articles that are not useful: artifacts too dispersed

likely lead to the creation of other “outlier” artifacts

not easily fitting with others. The systems where

agents set the goal randomly have a very low article

production, almost to the point that each article is a

unique exemplar; the diversity is high, but so is the

article turnover, because of the great difficulty in

replacing inputs that have disappeared. The system

wanders through artifact space. The goal setting

strategy of P1_0 systems leads them to reduce the

covered area despite the high diameter, but once the

structure of artifact space reaches a kind of

stationary condition, lots of agents’ innovation trials

fail, because of the increased difficulty in realizing

suitable recipes starting from a very sparse situation

(note the emergence of levels, as partial reaction to

this situation).

The combination of the diameter and the covered

area in the artifact space seems therefore to play a

significant role in determining the main system

behaviours. The strategies the agents adopt can

influence the diameter: agents able to tune the

diameter in the artifacts’ space are able to force the

systems’ global behaviour, driving it toward high

diversity or high production levels.

Table 2: Some results about the four scenarios presented

in the text (averages on 10 runs). The “failures” column

show the percentages of the unsuccessful innovation trials

(there are no failures in P_rand scenario, because of in

absence of definite goals each single innovation is

produced). The “unsuccessful successes” are articles that

the agents were able to build, but that once made are not

used by other agents and therefore disappear; the

“successful successes” are articles that once made never

disappears; the “medium successes” covers the remaining

cases (the values in these last columns are percentages on

the total number of articles made during the simulation).

The table shows also the average final diversity of the

scenarios, and the corresponding fraction on the total

number of articles.

Failures Medium

successes

Unsuccessful

successes

Successful

successes

Diversity Diversity/

#artifacts

P_rand ‐‐‐ 39% 60% 2% 320 0.86

P_stan

26% 47% 40% 14% 240 0.06

P1_0 72% 27% 59% 14% 350 0.05

P_chan

26% 58% 34% 8% 1050 0.16

The only strategy supporting both a high number

of articles and elevated production and diversity

levels is the P_chang scenario, where there is a

feedback between agent size and probability of

innovating, leading to the presence of heterogeneous

agents (in fact, agents having different numbers of

recipes innovate at different rates, increasing the

already existing gap). We have simulated systems

with each of these characteristics separately, but

without obtaining any evident increase in the

number of articles or the diversity. In particular, we

simulated (a) worlds having from the beginning high

heterogeneity in agents’ innovation probabilities

(two groups of agents having p

inn

=0.2 and p

inn

=1.0),

(b) worlds where the innovation probability

increases in time and (c) worlds where the

innovation probability is p

inn

=1.0 from the

beginning. None of these variants sustain long

lasting growth, whereas agents that develop high

innovation activity after a soft growth phase are able

to do so.

These remarks show the presence in the model of

path dependent processes: in fact, agents having

from the beginning p

inn

=1.0 are not able to sustain an

endless growth, despite the high number of produced

innovations, whereas agents whose innovation

probability is linked to their recipes’ number (but

CONDITIONS FOR LONG LASTING SUSTAINABLE INNOVATION IN AN AGENT-BASED MODEL

415

only during the first part of the simulation) are able

to.

P_chang worlds are able to efficiently recruit the

new inventions, making them effective innovations

(8%+58%=66% of the new articles are useful) –

note that the superiority of the P_stand systems in

making “successful successes” (14%) is only

apparent: in fact the total number of artifacts in the

P_chang worlds is overwhelming, as is their

absolute number of completely successful

innovations.

4 CONCLUSIONS

In this work we present an agent-based model of

innovation, where agents and artifacts coevolve

giving rise to a system where innovations take place

and become integrated into the already existing

pattern of interactions. The model captures the basic

features of innovation processes, and allows the

search for strategies able to support the system

expansion, both in term of artifact space exploration

and of the artifacts’ quantity and diversity.

Quite unexpectedly, strategies implying high

innovation rates are not able to support long lasting

increases in system diversity. A key aspect is that of

the diameter of the explored area in the artifacts’

space: if too large, the artifacts match poorly,

leading to poor worlds with very low production

levels.

Systems having at the same time high diversity

levels and high production levels require a kind of

balance between exploration and exploitation

processes. A too strong expansion in artifact space

can lead to the building of artifacts that are not

integrated with one another, whereas an intense

propensity to build artifacts not so dissimilar to each

other can limit the exploration range.

Another unexpected feature that strongly

influences the integration of novelties into the

already existing patterns of interaction is the change

of providers, which allows:

 the existence of many agents with similar

specializations;

 the simultaneous presence of several ways to

build the same kind of objects (supporting in such a

way high diversity levels);

 the stabilization of complete chains of integrated

artifacts.

Systems without this peculiar feature cannot

sustain or diffuse innovations.

A strategy able to provide a long lasting

sustainable enrichment is that of varying the agents’

propensity to innovate, as for example by coupling it

with a measure of the agents’ size. This strategy

allows a complex interplay among each single agent

and its environment (the other agents and the

artifacts), bringing to growth path dependent

processes.

We identify therefore several processes able to

sustain diversity, and make some observations about

the behaviour of the number and diversity of

artifacts produced by groups of agents. Further work

is needed to study the influence of artifact coding on

the feedbacks discussed here and to analyze the

formation of the structures in agent-artifact space

that differentiate the proposed scenarios, in order to

develop new strategies able to support long lasting

and sustainable cascades of innovations.

ACKNOWLEDGEMENTS

We would like to thank Andrea Ginzburg, Giovanni

Bonifati, Sander van der Leeuw and Margherita

Russo for valuable discussions and suggestions.

REFERENCES

Axelrod R., Tesfatsion L., 2006 Handbook of

Computational Economics Vol.2: Agent-Based

Computational Economics Ed. L. Kenneth, K. L. Judd,

L. Tesfatsion North-Holland, Amsterdam.

Cioffi-Revilla, C., 2002. Invariance and Universality in

Social Agent-based Model Simulations, Proceedings

of the National Academy of Sciences of the U.S.A., 99,

Suppl. 3.

Dosi G., 1982 Technological paradigms and technological

trajectories. Research Policy, Vol. 11, No. 3.

Dutta P. K., 1999, Strategies and games: theory and

practice, MIT Press.

Epstein J.M., Axtell R., 1996 Growing Artificial Societies:

Social Science from the Bottom Up MIT Press,

Cambridge.

Fagerberg, J., David C. M., Richard R. N., 2004

Innovation: A Guide to the literature. The Oxford

Handbook of Innovations. Oxford University Press

Fontana, W., 1992. Algorithmic chemistry, in Langton CG

et al (eds): Artificial Life II. Redwood City (CA),

Addison Wesley.

Fontana W., Buss L. W, 1994. The arrival of the fittest:

towards a theory of biological organization, Bull.

Mathem. Biol. 56: 1-64.

Fudenberg D., Tirole, J., 1991. Game theory. MIT Press

Gilbert N., Terna, P., 2000. How to build and use agent-

based models in social science. Mind & Society, no. 1

ICAART 2011 - 3rd International Conference on Agents and Artificial Intelligence

416

Gould S. J., Verba E., 1982. Exaptation: a missing term in

the science of form Paleobiology, 8(1).

Hayek, F. A., 1948. Individualism and economic order.

Chicago: University of Chicago Press.

Lane D, Serra, R., Villani, M, Ansaloni L., 2005. A

theory-based dynamical model of innovation processes

Complexus, Vol. 2, No. 3-4.

March J. G., 1991. Exploration and Exploitation in

Organizational Learning, Organization Science, Vol.

2, No. 1.

Lane D., 1993a. Artificial Worlds and Economics, part I,

Journal of Evolutionary Economics, v.3.

Lane D., 1993b. Artificial Worlds and Economics, part II,

Journal of Evolutionary Economics v.3. Reprinted

also in N. Gilbert, ed., Computational Social Sciences

(Sage: Sage Benchmark in Social Research Methods

Series), 2010.

Lane D., R. Maxfield 1997. Foresight Complexity and

Strategy in Economy as an evolving complex system II

Ed. WB Arthur, S.Durlauf, D. Lane, Addison Wesley,

Redwood city.

Lane D., R. Maxfield 2005. Ontological unsecurity and

innovation Journal of Evolutionary Economics 15: 3-

50.

Olsen, M., 1965. The logic of collective action.

Cambridge, MA: Harvard University Press.

Ormerod P., Rosewall B., 2002. On the methodology of

assessing agent-based evolutionary models in the

social sciences. in Proceeding of the Second Brisbane

workshop on complex adaptive systems, Brisbane.

Schelling, T. C., 1978. Micromotives and macrobehavior.

New York, NY: W. W. Norton and Co.

Schumpeter J., 1934. The Theory of Economic

Development, Harvard University Press, Cambridge,

Massachusetts.

Serra, R., Villani, M. & Lane, D., 2009. Modelling

Innovation. In Complexity perspectives on innovation

and social change D. Lane, D. Pumain & S. van der

Leeuw (eds):. Berlin: Springer.

Smith, A., 1937. An inquiry into the nature and causes of

the wealth of nations, New York, NY: Cannan Edition,

American Modern Library Series.

van der Leeuw S. E., Lane D. and Read D., 2009. The

long-term evolution of social organization. In

Complexity Perspectives in Innovation and Social

Change, Ed. D.Lane, D.Pumain, S.E. van der Leeuw

and G.West Springer, Netherlands.

Villani, M., Bonacini, S., Ferrari, D., Serra, R., Lane, D.,

2007. An agent-based model of exaptive processes.

European Management Review 4

Villani, M., Serra R., Ansaloni L. & Lane D., 2008.

Global and local processes in a model of innovation In

Cellular Automata. 8th International Conference on

Cellular Automata for Research and Industry H.

Umeo et al (eds): Springer Lecture Notes in Computer

Science 5191.

Villani M., Bonacini S., Ferrari D., Serra R., 2009

Exaptive processes: an agent-based model in

Complexity Perspectives in Innovation and Social

Change, Ed. D.Lane, D.Pumain, S.E. van der Leeuw

and G.West Springer, Netherlands.

Villani M., Ansaloni L., 2010. A theory-based dinamical

model of exaptive innovation processes in P.

Ahrweiler (ed.) Innovation in Complex Social Systems

Routledge.

CONDITIONS FOR LONG LASTING SUSTAINABLE INNOVATION IN AN AGENT-BASED MODEL

417