Cui.unige.ch

Self-Organising Mechanisms from Social and Business/EconomicsApproaches HASSAS SalimaLIRIS-CNRS, University of Lyon [email protected]://www710.univ-lyon1.fr/˜hassas DI MARZO-SERUGENDO GiovannaUniversity of [email protected]://cui.unige.ch/˜dimarzo/ KARAGEORGOS AnthonyUniversity of Thessaly, [email protected]://inf-server.inf.uth.gr/˜karageorgos/ CASTELFRANCHI CristianoUnit of AI, Cognitive Modelling and [email protected]://www.istc.cnr.it/ Keywords: social inspired computing, self-organising business, economics, self-organising social mod-els This paper reviews self-organising socially inspired approaches and their use in the con-text of computing systems. It presents different mechanisms such as stigmergy fromsocial insects behaviours, epidimec spreading, gossiping, trust and reputations inspiredby human social behaviours, as well as other approaches from social science related toBusiness and Economics. It also reviews issues related to social network dynamics, so-cial network patterns, social networks analysis, and their relation to self-organisation.
Socially inspired computing approaches are then illustrated through applications in thecontext of the WWW, computer networks and B2B.
Socially inspired computing gathers comput- Nowadays computing systems are open systems ing techniques that make use of metaphors in- evolving in a dynamic complex environment.
spired by natural social behaviors, exhibiting self- They are designed as sets of interacting compo- organisation, self-adaptation and self-maintaining nents, highly distributed both conceptually and physically. The evolving complexity of these sys- haviours range from those observed in biological tems requires computing techniques exhibiting ro- entities such as bacteria, cells and social insects bustness, high scalability and decentralised auto- to animals and human societies. One important nomic control. Social activities provide a great characteristic of these societies is their emergence source of inspiration for developing such highly as patterns developed from relatively simple in- teractions in a network of individuals.
underlying network as highly connected nodes.
These systems are well known to exhibit inter- Tools of social networks analysis, make use of esting characteristics such as robustness, capacity three popular metrics: 1) Activity: expressesd by of self-adaptation and survivability in uncertain the position of one particular node as a ”con- and dynamic environments, tolerance to random- nector” if it has a high number of direct links to other nodes; 2) Betweenness: expressed by a From the plethora of existing mechanisms of node which could have a small number of con- particular interest are those originating from so- nections but holds a powerful position as the sole cial insects, human social networks and in partic- ”boundary spanner” between different groups; 3) ular those related to business organisations and Closeness: expressed by nodes having the short- est paths to all others. These tools and metrics mechanisms ultimately derived from business and can be applied at the level of individuals, organi- economic domains could allow a new generation sations and industries. They can also be used to of IT systems that would have some of the prop- analyse computer networks and information sys- erties of human societies such as: adaptability, tems to develop adequate tools for their manage- robustness, emergent specialisation, redundancy, distribution of decision making and being inten-tionally interpretable.
In this paper we start with a discussion of so- cial network dynamics and we elaborate on the Recent studies have shown that networks result- impact that self-organising behaviours have on ing from social activities exhibit a self-organising their topology. Subsequently we focus on differ- characters similar to those networks generated ent types of social behaviours, such as those ob- by some biological metabolic or natural systems.
served in insect and human societies as well as This self organising character is observed through business and economic organisations, which orig- inate from social network dynamics and can be [38] ”Many empirical evidences have prompted used as source of inspiration for developing dy- that most of the times the resulting networks namically adaptable computer systems.
topology exhibits emergent phenomena which can-not be explained by merely extrapolating the lo-cal properties of their constituents”. Recent re- search [18][6] states that ”complex man-made net-works , serving as support for social activities and Social networks are networks where vertices rep- dynamics, such as Internet and the world wide resent individuals and edges represent social con- web, share the same large-scale topology as many nections, contacts, interactions, etc. Social be- metabolic and protein networks of various organ- haviour both shapes and is shaped by such social isms.” and that ”the emergence of these net- works is driven by self-organizing processes thatare governed by simple but generic laws”. Topol- ogy of such networks is characterised as scale free [1]. A scale free network is a class of an nonhomo-geneously wired network. The topology of these Social network analysis reveals densely connected networks is extremely sparse with a few randomly clusters or communities of practice. Topology of distributed highly connected nodes (hubs) and social networks is the direct expression of social many nodes with few connections. This implies behaviours. Communities for example, represent that the probability P (k) that a vertex has de- entities that have something in common, such as gree k, is characterised by a power-law behaviour web pages with similar content and individuals P (k) = k−γ , where 2 < γ ≤ 3 is a character- sharing common knowledge. They are observable istic exponent [38]. These networks also exhibit on their associated social network as a cluster of the so-called small world phenomenon [50], which nodes with high connectivity. Positions of partic- means that they are characterised by a short av- ular nodes as authority nodes are expressed on the erage path length, implying that each node of the Self-Organising Mechanisms from Social and Business/Economics ApproachesInformatica 30 page xxx–yyy network could reach any other node through a rel- a particular amount of time before it evaporates.
atively small number of hops. These charcateriscs When other ants encounter a trail of pheromone, make these networks surprisingly robust in facing while exploring their environment, they are in- random attacks and failures and in counterpart fluenced to follow the trail until the food source, extremely vulnerable to strategic attacks directed and while coming back to the nest they enforce to nodes with high connectivity [2][37].
the initial trail by depositing additional amountsof pheromone. The more the trail is followed, themore it is enforced and has a chance of being fol- lowed by other ants in the future. Collective sort- ing is a collective behaviour through which somesocial insects sort eggs, larvae and cocoons. As Social insects societies such as ants, bees, wasps mentioned in [8], an ordering phenomenon is ob- and termites exhibit many interesting complex served in some species of ants when bodies are behaviours such as emergent properties from lo- collected and are dropped later in some area. The cal interactions between elementary behaviours probability of picking up an item is correlated with the density of items in the region where tive behaviour is the outcome of a process of the operation occurs. This behaviour has been self-organisation, in which insects are engaged studied in robotics through simulations and real through their repeated actions and interactions implementations [28]. Robots with primitive be- with their evolving environment [28].
haviour are able to achieve a spatial environment organisation in social insects relies on an underly- structuring by forming clusters of similar objects ing mechanism : Stigmergy, originally introduced via the mechanism of stigmergy described above.
Moreover, these kind of social insect behaviours haviour of a kind of termites during the construc- have inspired many mechanisms for building arti- tion of their nests and noticed that the behav- ior of workers during the construction process isinfluenced by the structure of the constructionsthemselves. This mechanism is a powerful prin- ciple of cooperation in insect societies.
been observed within many insect societies suchas wasps, bees and ants. It is based on the use of the environment as a medium of inscriptionof past behaviours effects, to influence future be- Human collective behaviour occurs without cen- haviours. This mechanisms defines what is called tral control, and through self-organisation. In this auto-catalytic process, that is the more a process case, intimately linked with the notion of self- occurs, the more it has a chance to occur in the organisation is the notion of ”emergence” in the sense that ”social functions” arise out from (self- how simple systems can produce a wide range interested) human collective behaviour. In social of more complex coordinated behaviors, simply sciences different interpretations of the notion of by exploiting the influence of the environment.
social functions have been expressed, essentially Many behaviours in social insects, such as forag- considering that even if social functions are not ing or collective sorting are rooted on the stig- intentional and possibly unknown they constitute mergy mechanism. Foraging is the collective be- the ultimate end of the society and explain its haviour through which ants collect food by ex- existence. This concept has also been given an ploiting their environment. During the foraging explanation as the ”invisible hand” which would process, ants leave their nest and explore their manage forms of unplanned coordination (like environment following a random path. When an market) in which human interest increases [27] ant finds a source of food, it carries a piece of food through the apparently ”spontaneous emergence and returns back to the nest by laying a trail of a of an unintentional social order and institutions”.
hormone called pheromone along its route. This As pointed out by [14], the problem with this chemical substance persists in the environment for view is: ”how an unintentional effect can be an end” for the society; and ”how is it possible that tems for open distributed environment have also we pursue something that is not an intention of been realised, for instance [34] proposes a delega- ours”. An alternative could be avoiding the con- tion logic including negative evidence, and dele- cept of social functions because of the problems gation depth, as well as a proof of compliance for and questions that they provoke. However, this is both parties involved in an interaction. The Pol- not satisfactory too, because nevertheless social icyMaker system is a decentralised trust manage- emergence happens and has the form of a goal- ment systems [5] based on proof checking of cre- dentials allowing entities to locally decide whether Therefore, it is important to distinguish two or not to accept credentials (without relying to a kinds of social emergence: 1. the emergent phe- centralised certifying authority). Eigentrust [32] nomenon is perceived by an observer, but has no is a trust calculation algotrithm that allows to cal- effect on the society; 2. the emergent phenomenon culate a global emergent reputation from locally has an effect on the society by self-reproducing maintained trust values. Recently, more dynamic and enforcing the social phenomenon.
and adaptive schemas have been defined, which Given the considerations above, Castelfranchi allow trust to evolve with time as a result of evi- considers that ”in order to have a function, a be- dence, and allows to adapt the behaviour of prin- haviour or trait or entity must be replicated and cipals consequently. We report here the results of the European funded SECURE [12] project, The principal argument is that ”the invisible which has established an operational model for hand” is not necessarily a good thing for society trust-based access control. Systems considered by (especially in the case of self-interested agents).
the SECURE project are composed of a set of au- The optimum order for the society can actually be tonomous components, called principals, able to bad for individuals or for everybody. For instance, take decisions and initiatives, and are meaning- prisons generate criminals that in turn feed pris- ful to trust or distrust. Principals maintain local ons. This is a function not a social objective.
trust values about other principals. A principal The important thing is that ”re-organisation that receives a request for collaboration from an- simply maintains the system, but not necessariy other principal decides to actually interact with that principal or not on the basis of the currenttrust value it has on that principal for that par- ticular action, and on the risk it may imply forperforming it. If the trust value is too low, or Trust-based systems or reputation systems take the associated risk too high, a principal may re- their inspiration from human behaviour. Uncer- ject the request. After each interaction, partic- tainty and partial knowledge are a key charac- ipants update the trust value they have in the teristic of the natural world. Despite this uncer- partner, based on the evaluated outcome (good tainty human beings make choices, take decisions, or bad) of the interaction. A principal may also learn by experience, and adapt their behaviour.
ask or receive recommendations (in the form of trust values) about other principals. These rec- rity policies, credentials and trust relationships, ommendations are evaluated (they depend on the trust in the recommender), and serve for updat- based management systems combine higher-order ing current trust values. Artificial systems built logic with a proof brought by a requester that is on the human notion trust as exposed above have the particularity to exhibit a self-organising be- tially based on delegation, and serve to authenti- haviour [17], as identified by Nobel prize Ilya Pri- cate and give access control to a requester [51].
gogine and his colleagues [20]. Additional trust Usually the requester brings the proof that a and reputation systems are surveyed in [21], and trusted third entity asserts that it is trustable for the particular case of multi-agent systems they have been designed for static systems, where anuntrusted client performs some access control re-quest to some trusted server [3, 7]. Similar sys- Self-Organising Mechanisms from Social and Business/Economics ApproachesInformatica 30 page xxx–yyy which occurs by this way. This mechanism pro- vides a powerful abstraction metaphor for infor-mation spreading, knowledge exchange and group In social science, it is now established that so- organisation in large scale distributed systems. In cial interactions play a fundamental role in learn- P2P systems, a class of protocols categorised as ing dynamics, and lead to cognitive development.
epidemic protocol has been proposed [47]. These This phenomenon is known as ”Zone of Proxi- protocols are characterised by their high robust- mal Development” which Vygotsky describes it as ”the distance between the actual development level been also used for routing in sensor networks. For as determined by independent problem solving and example in [9], a rumour routing algorithm for the level of potential development as determined sensors networks is proposed. This algorithm is through problem solving under adult guidance or based on the idea of creating paths leading to each in collaboration with more capable peers” [48] [16].
event and spreading events in the wide-network The effect of socialisation has also been proven through the creation of an event flooding gradi- to benefit to the propagation of knowledge inside ent field. A random walk exploration permits to an interconnected population. In [15] the authors considered social learning in a population of my-opic, memoryless agents. They have made someexperiments to study how technology diffuses in a population based on individual or collective eval- uation of the technology. The authors have shownthat under a learning rule where an agent changes his technology only if he has had a failure (abad outcome), the society converges with prob- Market-based mechanisms are built along the ability 1 to the better technology. In contrast, lines of economic markets. In this approach, sys- when agents switch on the basis of the neighbour- tems are modelled along the lines of some eco- hood averages, convergence occurs if the better nomic model in which participating entities act towards increasing their personal profit or utility.
ments show how a better technology spreads in a System wide parameters are modelled in a manner population through a mechanism of imitation and similar to macroeconomic variables such as eco- thanks to neighbourhood connections. In another nomic growth. The parameters of the individual work [4], the authors develop a general framework entities correspond to microeconomic parameters.
to study the relationship between the structure The key point in such systems is to select suitable of these neighborhoods and the process of social micro level parameter values and market interac- learning. They show that, in a connected soci- tion rules so that desired system goals, both local ety, local learning ensures that all agents obtain the same payoffs in the long run. Thus, if actions Market-based approaches contrast the tradi- have different payoffs, then all agents choose the tional way of modelling self-organisation and emergence in economic systems, which is primar-ily based on analytic general equilibrium mod- els, for example as is done in [19].
problem with analytic approaches is that theycannot represent all possible situations due to As cited in [30] Gossip is one of the most usual the non-linearity of economic phenomena [11], social activities. This mechanism allows for the which is due to the fact that economies are com- aggregation of a global information inside a pop- plex dynamic systems [45]. Instead, market-based ulation, through a periodic exchange and update approaches view macroeconomic phenomena as of individual information among members of a emergent results of local interactions of the eco- group. The neighbourhood as well as the level nomic entities [11, 29, 45]. An example is eco- of precision of the exchanged information play an nomic growth which can be described at the macro important role on the nature of social learning level but it can never be explained at that level [13]. The reason is that economic growth results existing money from lenders, credit economy al- from the interaction of a variety of economic ac- lows producers to obtain credit up to a certain tors, who create and use technology, and demand- level from creditors in order to pay for production of new products. In this way, producers can more There are numerous variations of market-based easily force their way into the market but the dan- self-organisation mechanisms. An exemplar such ger of becoming bankrupt is increased. To explain mechanism which is based on the creative destruc- economic development in this framework one only tion principle is described in the following section.
needs to explain why entrepreneurs would wantto introduce new products to the market.
fective entrepreneurs survive the battle and in-crease their profit. Failed entrepreneurs cannot Creative destruction is a term coined by Schum- repay their debt and therefore they go bankrupt peter [40] to denote a ”process of industrial muta- and they are eliminated. As initially stated by tion that incessantly revolutionizes the economic Schumpeter [40] and later evaluated experimen- structure from within, incessantly destroying the tally, for example [10], economic growth in this old one, incessantly creating a new one.” In other model is generated in cycles that emerge from the words, creative destruction occurs when a new disturbance caused by entrepreneurs entering the setting eliminates an old one leading to economic development. According to this view an economic In such a model there is particular interest from system must destroy less efficient firms in order both the global, macro economic perspective and to make room for new, possibly more efficient en- the local microeconomic one. Individual produc- trants. A representative example of creative de- ers can decide on their entrepreneur policy so struction is the evolution of personal computer that to increase their profit and avoid the risk industry which under the lead of Microsoft and of getting bankrupt. On the other hand the eco- Intel destroyed many mainframe computer com- nomic system regulators can decide on the self- panies; however, at the same time one of the most organisation rules so that to increase overall sys- important technological achievements of this cen- The main roles that economic actors play in a market-based economy are those of producer,worker and consumer. Producers produce goods or provide services that consumers demand. Con- business models and theories which use self- sumers consume the goods and use the services organisation. In an increasingly complex global in exchange of some monetary or utility value.
economy, businesses are faced with unpredictable When there is high demand producers tend to behaviours and fast pace of change. As a result, hire workers to assist them in goods production the emphasis in contemporary business models or service provision in exchange of some wage.
has shifted from efficiency to flexibility and the Since producers cannot sell their production be- forehand, they must hold enough money to pay More recent approaches, for example the one the workers in order to start up production and described in [43], increasingly introduce business they can only get the necessary money by enter- models originating from the study of complex ing debt. According to the creative destruction principle, if producers are not able to pay the tions are guided and tied together by ideas, by worker wages then they go bankrupt and they are their knowledge of themselves, and by what they removed from the system, for example they are do and can accomplish. Therefore, the focus in reduced to simple workers, opening the way to such models is on the complex relationships be- other economic entities to try to become success- tween different business components and the ef- ful producers and satisfy the consumer demand.
fects that a change into some part of the system The creative destruction process is better illus- or its environment, however distant, might have trated in a credit economy. In contrast to a mon- on the behaviour of the entire system.
etary economy where producers can only borrow As examples of self-organising business models Self-Organising Mechanisms from Social and Business/Economics ApproachesInformatica 30 page xxx–yyy we discuss personalised marketing and activity- cally and horizontally via ”round table meetings”, which are organised along the lines of assessmentmeetings normally held in companies to assess results and handle exceptions. In these virtualround tables suitable participants soon emerge as Personalised marketing refers to following a per- de facto leaders due to their knowledge and expe- sonalised market strategy for each individual cus- rience. Subsequently, leaders tend to participate tomer which is evolving according to customer in each newly formed ”round table”. The view expressed in [46] is that to model the interactions proach is the one-to-one variable pricing model of participants in a ”round table”, it is necessary [25], which refers to providing an individual of- to simulate the whole activity of each of them in- fer to each customer using Internet technologies.
cluding their reasoning and communication.
The model uses self-organisation in the market-ing policies by changing customers targeted andthe prices quoted based on market dynamics, cus- tomer characteristics and the business goals.
A shift towards to personalised marketing mod- els is viewed as being driven by syndication [52].
Syndication involves the sale of the same good to many customers, who then integrate it with otherofferings and redistribute it, as is the case in redis- Based on the SECURE trust and risk security tributing popular TV programs. An example of framework, an anti-spam tool has been developed a company using syndication is FedEx which syn- which allows collaboration among e-mail users dicates its tracking system in several ways [52].
by exchanging recommendations about e-mail’s The company allows customers to access com- senders. An authentication scheme has been com- puter systems via its Web site and monitor the bined to the SECURE framework in order to in- status of their packages. For corporate customers crease the level of sender authentication [41].
FedEx provide software tools that enable the or- On the WWW, a plethora of systems have been ganisation to automate shipping and track pack- developed for content retrieval, filtering or organ- ages using their own computing resource. Each isation using socially inspired computing. As an customer is offered different prices depending on illustration, we present here a pioneering work a variety of parameters. Many websites, such as [36], in information retrieval field which combined eBay, also apply variable pricing for their offers.
inspiration of social human behaviours, and eco-nomic markets to propose an interesting systemfor information retrieval on the web. In this work, documents are represented by keyword vectors, Another example from the area of management representing individuals (agents) of an artificial is the theory of activity described in [46]. In this ecosystem. This population evolves through an view a company consists of networks of working evolutionary process of natural selection using a groups that can change their structure, links and genetic algorithm to find documents which best fit behaviour in response to business requirements.
the user request. The user feedback is used to re- The aim is to capture the self-organisation deci- ward (resp. to punish) the fittest individual (the sions that need to be taken during the business less fitting individual) by giving it a credit value.
operations both by managers and by interactions These credits are then used by agents in a market between employees. The emphasis is on solving based metaphor to estimate the cost of inhabiting potential conflicts of interests in both the inner the artificial ecosystem. The fittest agents have and the external co-operative activity of the com- enough credits to continue living in the ecosystem and the less fitting agents will die. Another sys- In this approach the structure of the company tem called WACO has been proposed in [26]. The is virtual. There is no clear hierarchy and control; instead control effects can be initiated both verti- agents deployed on the web to form clusters of se- mantically similar documents and dynamically or- In such a network, agents maintain indices to ganise the web content. These agent behaviours, actual documents and to other agents as well, take inspiration of social insect behaviours. They treating both in a similar manner - based on the combine foraging ant behaviour and the collective semantics of their content. The key feature in this approach is content dependent query redirection,based on semantic indexing. If an agent is unable find a document on a given topic, it re-directs thereceived query to the agents which believes are T-Man is a generic protocol based on a gossip most likely to find it. The connections between communication model and serves to solve the the agents adapt themselves based on the his- topology management problem [31]. Each node tory of successfully served queries, forming a dis- of the network maintains its local (logical) view tributed self-organising search engine which is ca- of neighbours. A ranking function (e.g. a dis- pable of executing on heterogeneous servers over tance function between nodes) serves to reorgan- the internet and dynamically indexing all avail- able documents. The important aspect of such a tance). Through local gossip messages, neighbour search engine is that each node, though possess- nodes exchange or combine their respective views.
ing only limited amount of local information, can Gradually, in a bottom-up way, through gossiping and ranking, nodes adapt their list of neighbours, Each piece of information received from an and consequently change and re-organise the net- agent corrects the coordinates of its representa- work topology. The T-Man protocol is particu- tion in the semantic index of the recipient. Fur- larly suited for building robust overlay networks thermore, each link to an agent has also its own supporting P2P systems, especially in the pres- utility based rating. Those ratings are used for ence of a high proportion of nodes joining and the selction of the right candidates for redirecting The SLAC (Selfish Link and behaviour Adap- Rating adaptation is done using a free mar- tation to produce Cooperation) algorithm [24] ket approach. According to this approach agents favours self-organisation of P2P network’s nodes provide chargeable search services to each other.
into tribes (i.e. into specialised groups of nodes).
Each query has some limited amount of network The SLAC algorithm is a selfish re-wiring proto- currency, termed neuro, which dissipates in the col, where by updating its links with other nodes course of query processing in the network. Neu- in order to increase its utility function, a specific ros circulating through the network are used by node leaves its current tribe, and joins a new one.
the agents to update their connections with the In addition to P2P systems, the SLAC algo- other agents, based on their utility, in a similar rithm has many potential applications, for in- manner that money flow in a real economy deter- stance to organise collaborative spam / virus fil- mines the structure of business relationships.
tering in which tribes of trusted peers share meta- The semantic network economy is based on the information such as virus and spam signatures.
This would elimite the need for trusted third par-ties with central servers.
– The cost of each delegated query processing – The cost of each document (query) transac- – Agents aim to minimize their expenditures.
organisation mechanisms can be found in the According to these rules each agent keeps track domains of business community networks [33]. An of the balance of transactions of all other agents example of such approach is the self-organising semantic network of document indexing agents nomically rational and aiming to maximise their profit they tend to delegate queries to experts in Self-Organising Mechanisms from Social and Business/Economics ApproachesInformatica 30 page xxx–yyy the query topic, thus minimizing effective cost of [2] R. Albert, H. Jeong, and A. Barabasi. Error and attack tolerance of complex networks.
Similar market-based techniques are applied in trade networks where the aim is to select tradepartners based on continually updated expected [3] Andrew W. Appel and Edward W. Felten.
Proof-carrying authentication. In ACM Con-ference on Computer and CommunicationsSecurity, pages 52–62, New York, NY, USA,1999. ACM Press.
[4] Venkatesh Bala and Sanjeev Goyal. Learning In this paper we have surveyed some self- from neighbours. Review of Economic Stud- organising social approaches and presented their use as metaphors for distributed computing sys- [5] M. Balze, J. Feigenbaum, and J. Lacy. De- centralized trust management. In IEEE Sym- niques have shown their effectiveness for sys- posium on Security and Privacy, pages 164– tems and applications evolving in distributed and highly dynamic environments, such like current from thoseobserved in biological entities such as bacteria, cells and social insects, to animals and human societies, are rooted in the dynamics of their underlying social network. Social behaviourboth shapes and is shaped by such social net- [7] L. Bauer, M. A. Schneider, and E. W. Fel- works. One important characteristic of societies ten. A proof-carrying authorization system.
is their emergence as patterns developed from rel- Technical report, Princeton University Com- atively simple interactions in a network of in- forced through dynamics underlying the so ob- and J-L. Deneubourg. The phase-ordering ki- known to exhibit interesting characteristics such netics of cemetery organization in ants. Tech- as robustness, capacity of sel-adaptation and sur- nical Report 98-01-008, Santa Fe Institute, vivability in uncertain and dynamic environment sented different mechanisms of social behaviours [9] D. Braginsky and D. Estrin. Rumour routing and showed their use in computing environments algorithm for sensor networks. In Proceed- through some illustrative applications. Socially ings of the Fisrt Workshop on Sensor Net- inspired computing metaphors, provide a real new paradigm for programming highly distributed and dynamic computing systems. However, proposedapproaches are still developped in an ad hoc man- [10] Charlotte Bruun. The economy as an agent- ner, and a real theory for socially inspired com- namics. Industry and Innovation, 10(4):475–491, 2003.
mechanics of complex networks. Reviews of Modern Physics, 74(1):47–97, 2001.
http://www.socsci.auc.dk/˜cbruun/aceintro.pdf.
[12] V. Cahill and al. Using trust for secure col- laboration in uncertain environments. IEEE Pervasive Computing Magazine, special issueDealing with Uncertainty, 2(3):52–61, 2003.
[23] Sergei Guriev, Igor Pospelov, and Margarita [13] Bo Carlsson and Eliasson Gunnar.
works in an economy with imperfect infras- tructure. In Second International Conference Industry and Innovation, pages 1–25. Taylor volume 22. Society for Computational Eco-nomics, Geneva, Switzerland, 1996.
[14] C. Castelfranchi. The theory of social func- science and multi-agent learning. Journal of Cognitive Systems Research, 2(1):5–38, 2001.
work. In S. Brueckner, G. Di Marzo Seru-gendo, D. Hales, and F. Zambonelli, ed- [15] Kalyan Chatterjee and Susan H. Xu. Tech- itors, Engineering Self-Organising Applica- nology diffusion by learning from neighbours.
tions (ESOA’05), Utrecht, The Netherlands, Advances in Applied Probability, 36(2):355– [25] Glenn Hardaker and Gary Graham. Energiz- ing your e-commerce through self-organising collaborative marketing networks. Technical http://www.balancedreading.com/vygotsky.html.
report, School of Business, University of Sal-ford, UK, 2002.
[17] G. Di Marzo Serugendo. Trust as an interac- tion mechanism for self-organising systems.
[26] S. Hassas. Using swarm intelligence for dy- namic web content organization. In Proceed- plex Systems (ICCS’04), 2004. Available at ings of the IEEE Swarm Intelligence Sympo- http://cui.unige.ch/˜dimarzo/papers/iccs04.pdf.
sium, pages 19–25, Los Alamitos, CA, USA,2003. IEEE Computer Society.
[18] M. Faloutsos, P. Faloutsos, and C. Faloutsos.
On power-law relationships of the internet [27] F. A. Hayek. Studies in philosophy, politics topology. In Proceedings of the Special Inter- and economics. Routledge & Kegan, London, SIGCOMM’99), pages 251–262, New York,NY, USA, 1999. ACM Press.
self-organization and sorting in collective [19] Sergio Focardi, Silvano Cincotti, and Michele robotics. Artificial Life, 5(2):173–202, 1999.
crashes. Journal of Economic Behavior and [29] Peter Howitt and Robert Clower.
Organization, 49(2):241–267, 2002.
emergence of economic organization. Jour-nal of Economic Behavior and Organization, [20] P. Glansdorff and I. Prigogine. Thermody- namic study of structure, stability and fluc-tuations. Wiley, 1971.
[30] M. Jelasity. Engineering emergence through [21] T. Grandison and M. Sloman. A survey of Steven Gustafson, David Hales, and Natalio trust in internet applications. IEEE Commu- Krasnogor, editors, Proceedings of the Joint nications Surveys and Tutorials, 3(4), 2000.
Symposium on Socially-Inspired Computing,pages 123–126, Hatfield, UK, 2005. Univer- interactions inter-individuelles chez les belli-coitermes natalenis et cubitermes, la th´ Gossip-based overlay topology management.
Self-Organising Mechanisms from Social and Business/Economics ApproachesInformatica 30 page xxx–yyy itors, Engineering Self-Organising Applica- tions (ESOA’05), Utrecht, The Netherlands, worthy Email Addresses. In Proceedings of the Second Annual Conference on Privacy,Security and Trust (PST’04), pages 47–58, Fredericton, New Brunswick, Canada, Octo- H. Garcia-Molina. The eigentrust algorithm for reputation management in p2p networks.
In 12th International World Wide Web Con- [42] Sergey Shumsky. Self-organizing internet se- [33] Ulrike Lechner and Beat F. Schmid. Com- [43] Max Stewart. The Coevolving Organization.
munities - business models and system archi- Decomplexity Associates LtD, Rutland, UK, tectures: The blueprint of mp3.com, napster and gnutella revisited. In E. Sprague, editor,the 34th Hawaii International Conference on [44] Leigh Tesfatsion. A trade network game with [34] N. Li, J. Feigenbaum, and B. N. Grosof. A logic-based knowledge representation for au- editors, Computational Approaches to Eco- thorization with delegation. In 12th IEEE [35] J.-P. Mano, C. Bourjot, G. Lopardo, and tional economics: A constructive approach P. Glize. Bio-inspired mechanisms for arti- to economic theory. In K. L. Judd and Leigh ficial self-organised systems. Informatica, In Tesfatsion, editors, Handbook of Computa- [36] A. Moukas. Amalthaea: Information discov- ery and filtering using a multiagent evolv- Economics Series. North-Holland, 2005.
ing ecosystem. Applied Artificial Intelligence, [37] S.T. Park, A. Khrabrov, D.M. Pennock, self-organization and cooperation processes.
S. Lawrence, C.L. Giles, and L.H. Ungar.
In XIII Intern. Conference on the Applica- Static and dynamic analysis of the internet’s tion of Artificial Intelligence in Engineering, susceptibility to faults and attacks.
pages 91–96, Galway, Ireland, 2002.
Conference on Computer Communications(INFOCOM), 2003.
[47] S. Voulgaris, M. Jelasity, and M. van Steen.
A robust and scalable peer-to-peer gossip- [38] R. Pastor-Satorras and A. Vespignani. Epi- ing protocol. In G. Moro, C. Sartori, and demic spreading in scale-free networks. Phys- M.P. Singh, editors, Proceedings of Agents ical Review Letters, 86(14):3200–3203, 2001.
and Peer-to-Peer Computing: Second Inter- [39] S. D. Ramchurn, D. Huynh, and N. Jennings.
Trust in multi-agent systems. Knowledge En- of Lecture Notes in Artificial Intelligence, gineering Review, 19(1):1–25, 2004.
pages 47–58, Berlin, 2003. Springer-Verlag.
[40] J. A. Schumpeter. The economy as a whole - [48] L.S. Vygotsky. Mind and society: The devel- seventh chapter of the theory of economic de- opment of higher mental processes. Harvard velopment. Industry and Innovation, 9(1/2), [49] S.C. Wang, K.Q. Yan, and C.H. Wei. Mo- bile target advertising by combining self-organization map and decision tree. In Pro-ceedings of the IEEE International Confer-ence on e-Technology, e-Commerce and e-Service (EEE´ı04), pages 249–252, 2004.
[50] D. J. Watts and S. H. Strogatz. Collective dynamics of small-world networks. Nature,393:440–442, 1998.
[51] S. Weeks. Understanding trust management model for business in the internet era. Har-vard Business Review, 85:85–93, 2000.

Source: http://cui.unige.ch/~dimarzo/papers/InformaticaJournal1.pdf