Research Issues

There are two technical research issues, i) collective decision making, and ii) collective knowledge structuring. See the following figure. We will also develop a participatory platform for collective intelligence that will be used to develop actual social information systems:

Figure : Research Issues

A collection of intelligent agents, each of which represents a particular decision making thread, is called a multiagent system. Agent modeling is based on the observation of human behavior and reacts according to the protocol of collective decision making.

a) Research on Agent Modeling

A key objective of agent modeling is to create software that can substitute for human decision making. This enables us to not only simulate collective intelligence formation but also reduce the burden of humans in collective intelligence formation by automating such task as incentive matching. For observing human behavior, ethnography and quantitative measurements are well known approaches. However, discerning the human decision making is difficult if only observations are used. Thus, we will develop participatory methods such as gaming and participatory simulations, and establish a methodology to let humans observe the behavior of agents, and refine the model through discussion.

b) Research on Collective Decision Making Protocol

We will develop a collective decision making protocol that well matches human society. In order to form a group of people into a team, it is necessary to precisely match their incentives. Although many algorithms have been proposed for this matching problem, it is not straightforward to apply the proposed algorithms to real-world problems such as collective intelligence formation. Even if the performance of an algorithm is theoretically efficient, it may not work well, if humans misunderstand its rules. In order to resolve this issue, we need to develop participatory technologies to confirm whether or not the developed algorithms are well understood and utilized. This proposal aims at exploring protocols for collective decision making that is acceptable to society.

An atomic service is defined to allow access to and the use of accumulated knowledge, and a composite service is defined as a workflow to coordinate atomic services.

a) Research on Service Ontology

To realize service composition, services should be classified and standardized in advance. Since services range from simple to complex, a service ontology with hierarchical concept structure is useful for the classification of various services. Thus, we aim at developing a common upper ontology that enables step-by-step construction of service ontology based on the needs of each user field within specific application domains.

b) Research on Service Supervision

We propose service supervision as a framework to manage the execution of composite services consisting of multiple atomic services. Service supervision is the meta-level architecture for service execution. So far, similar ideas have been proposed for software reuse. Service supervision, however, requires far more complicated processing than program supervision because rights management and incentive comparisons are required when dealing with services.

We will develop infrastructure software for collective intelligence formation. First, we will realize participatory simulations with large numbers of agents, each of which represents a model of human behavior. Next, collective decision making will be simulated. Agent models and the protocol are to be refined with the participation of the stakeholders associated with the simulation. This platform provides service supervision functions such as digital rights management and will be used to develop actual social information systems.