The Web is pervasive, increasingly populated with inter-connected data, services, people and things. Empowering people through sophisticated social machines,11 the Web has to enlarge its scope to address hybrid communities10 in which intelligent agents (e.g., crawlers, bots, recommenders) collaborate with people as peers to help them cope with the ever growing number of available resources. As the Web continues to grow, it will continue to unlock practical uses that require increasingly autonomous, cooperative and long-lived agents operating on it.6 In this context, hypermedia, the engine of application state on the Web, will drive interaction among agents and people fueled by their autonomous and goal-directed behaviours. To ensure social order and global coherent behaviour, it is however important to complement this engine with regulation mechanisms to enforce and constrain actions in desired directions.
Regulation has been addressed by introducing policies targeting more the individual level in the Web domain and by norms, kind-of agreed policies targeting the social level in multi-agent systems.5 Policies on the Web have been discussed for a long time now. As stated by Daga,8 they have different meanings in relation to security and privacy,9 13 access and control,7 14 adaptable and context-aware systems to control the behaviour16 and expression of legal knowledge (terms and conditions licenses)3 (e.g., ODRL12). In the multi-agent domain, regulation has been studied to large extent in research on normative approaches,1 where regulative norms15 or prescriptions2 are generally used to specify who does what, in what context and as subject to what deontic modality (e.g., obligation, prohibition, permission). Regulative norms affect agent behaviour in an indirect manner so that agents can autonomously decide to conform or not to the prescriptive norms, for instance in order to balance internal motivation versus external consequences. To our knowledge, there does not exist yet a model that satisfies both domains.
Focusing on the organisation meta-model of the JaCaMo multi-agent framework,4 the aim of the project is to transform it in a Web readable format with ontologies complying to the Semantic Web standards, and to reuse already existing ontologies.
Based on the work initiated in 2012,17 we want to enrich the ontology for organisation specifications with the ontology to describe the organisation entities, that is, a description of the running organisation where agents are playing roles, executing missions, interacting altogether. Such descriptions will be provided as Web data as well, such that agents can uniformly query and reason about agent organisations, data and ontologies. Based on these representations, reasoning mechanisms can be leveraged in order to allow agents to check the consistency of the organisation structure, to infer new information from the existing state of the organisation. Ontologies will be used to provide a metamodel of the organisation as well as the organisation specifications, shared and distributed over the Web.
This internship takes place in the context of the ANR project HyperAgents.
To apply, send a CV, a motivation letter, grade transcripts, links to public professional profile, references, etc. to: Olivier Boissier (email@example.com), Andrei Ciortea (firstname.lastname@example.org), and Antoine Zimmermann (email@example.com).