ACONFESIONALIDAD Y NEUTRALIDAD
VI. POSTURAS OPUESTAS: COMPARATIVA ENTRE ANDALUCÍA Y CATALUÑA 1 Legislación autonómica
3. Aspectos relevantes: horario, currículo y docencia
While this review makes clear that there are readily available techniques for achieving cooperative behaviour among agents in a multi-agent system and for selecting interaction partners, in this thesis we focus on a different problem that has not generally been addressed. In particular, we are concerned with agents operating in open cooperative systems. We assume that such agents are self-interested, in the sense that they might not always be willing to cooperate with others due to individual interests or resource limitations and, additionally, that services are not always priced and may be provided free of charge.
Effective cooperation in this context requires cooperative behaviour to be motivated, so that agents have the incentive to cooperate even in the absence of any economic gain, and agents requesting services do not have to rely on altruism to guarantee service provision. Moreover, since agents in open systems are highly heterogeneous, with different skills
and preferences, it is not desirable for all possible cooperations to take place. Therefore, agents need some means for choosing between alternative cooperations.
Although there are various approaches to achieving cooperative behaviour among self- interested agents without economic compensation, they have some limitations. Evolu- tionary approaches do not consider the real motivations of agents to cooperate with others, nor the properties of alternative partners (since in most approaches partners are
selected at random (Feldman et al, 2004) or are selected among agents from the same
group (Riolo et al, 2001)). Therefore, they do not provide any grounds for an agent to
make autonomous decisions over interactions.
Most existing approaches using social incentives and reciprocity provide explicit incen- tives for agents to cooperative with others, but they do not account for the differences in the properties of alternative partners. In particular, in the brownie point approach (Glass and Grosz, 2000), although there is an incentive for group cooperation, there is no concrete benefit for the agent in gaining a brownie point when it depends on others to ex-
ecute itsown tasks. This is because brownie points represent an agent’s self-valorisation
(the agent rewards itself for cooperating in a team) and not the valorisation an agent receives from others in response to the provided service, which could then be used to receive a service in reciprocation in the future.
In utility-based decision-making, which considers expectations of future interactions (Sen
et al, 2003), agents that depend on each other have the incentive to cooperate to improve expectations of future interactions. However, this approach does not consider the success of the cooperative relations in terms of the counterbalance between provided and received services, which is important if the environment has agents with different preferences and perspectives, or even agents that reciprocate but by providing low quality services. Finally, regarding normative and organisation-based approaches (Dignum, 1999; Dignum and Dignum, 2003), we argue that instead of being used to motivate cooperations, they
are more appropriate to enforce reciprocation by ensuring that agents do reciprocate
through norms, contracts, and so on.
Regarding the problem of forming cooperations, current mechanisms consider either information on the properties of provided services (such as service evaluation and sim-
ilarity) (Casati et al, 2004; Caverlee et al, 2004) or reciprocal relationships between
providers and requesters (Sichmanet al, 1994; Davidet al, 2001). However, in an open
cooperative system, both types of information are relevant when choosing a cooperation partner, and there is no attempt to balance this information in a single partner selection mechanism.
A key challenge in open systems with free services, therefore, is to enable cooperative behaviour of self-interested providers and requesters to result from their autonomous
decision-making. This requires a means not only to provide non-monetary incentives
of the quality of provided and received services. There is also a need for selection mech-
anisms that balance aspects of reciprocation, service quality, and resource limitations,
each of which is relevant to cooperation in the kinds of systems we consider. In the rest of this thesis, we develop mechanisms and models to meet this challenge.
Problem Scenario
3.1
Introduction
Computational systems in which participants share personal tools and data, are becom- ing very popular, both in commercial and academic communities. Computer programs with different purposes and functionalities are constantly being produced, as well as data of different kinds (including scientific publications, newspapers, experimental results in physics, biology, computer science, and so on). Instead of being just for personal use,
these tools and data can be made accessible to others as services in a distributed sys-
tem, so that a participant in such a system can make requests from a remote service and receive results after the service has completed. With such systems, participants can gain access to services they would not otherwise be able to access if they were in an isolated and closed system, and can request services at the time they are needed, without being connected to just one service all the time. Such characteristics are specially desirable for application domains that are constantly changing, since newly discovered data and newly developed tools can be made available to others.
One such domain is bioinformatics, which has seen an explosion in the number of devel- oped services since the start of genome sequencing projects all over the world. Indeed, it continues to see an increase in the number of services being developed for more specific areas like proteomics and drug discovery. In addition, in many bioinformatics laborato- ries, unique data sets are being created that are not published in public databases, but could usefully be shared with the global community. Conversely, the number of services available in closed bioinformatics systems is limited when compared with the variety of services that are available in the global open community. In particular, access to a
wider range of services, including private tools and databases, can facilitate the search
for, and use of, more suitable services, of better quality, in order to improve the results of bioinformatics experiments more generally.
Although there are benefits for participants to have the opportunity to interact with a 30
large number of providers and requesters, it is not desirable for all possible interactions to take place. First, because services with different characteristics are available, not all of them will have the same quality or will take the same time to return results. With some services being better than others, it is important that participants are able to choose those with better properties, like higher quality or smaller response time. Thus, when a participant requesting a service can find several alternatives with similar functionalities, some mechanism is needed to select the service most suitable for its needs, for example in terms of performance, quality, and speed. In addition, when participants in a distributed application are self-interested, service providers need some compensation to be given in return for their effort and investment in performing a service, and requesters need to be sensitive to this need in order to be able to find a service provider willing to accept its request in a timely fashion.
Since participants providing services may receive many requests from others, and per- forming such requests may be computationally costly (as in the case of bioinformatics services which usually involve processing large amounts of data), from a provider’s per- spective it is necessary to limit service provision to avoid being overwhelmed with services to provide for others. In addition, providers must have the autonomy to decide whether to accept requests at all. Thus, when a participant providing a service receives more requests than its available computational power or wants to select which interactions to engage in, some mechanism is necessary to choose between incoming requests.
In this chapter we investigate open cooperative systems in the context of bioinformatics, which we use as a problem scenario, since it offers characteristics of dynamism, service variety, and resource constraints, giving rise to the problems in requesting and providing services described above.
The chapter is organised as follows. We first introduce the bioinformatics domain and discuss the particular area of proteomics in Section 3.2. Next, we present an overview of existing computational tools used in bioinformatics research in Section 3.3, and the key distributed applications that have been proposed to integrate bioinformatics tools in Section 3.4. Our bioinformatics application scenario is presented in Section 3.5. We then identify the requirements for the effective operation of the application scenario in Section 3.6, and discuss the key problems to be addressed in the target domain. Finally, we conclude in Section 3.7.