TRIAS - Logic of trust and reliability for information agents in science
- Collaboration between MINE (Schommer) and ICR (van der Torre, Weydert)
- Administrator (full project) : Christoph Schommer
- Subproject TRIAS-MINING : see MINE (Christoph Schommer)
Subproject TRIAS-LOGIC
Researchers
- Emil Weydert (Principal investigator)
- Leon van der Torre (Professor)
- Guillaume Aucher (Postdoc)
- Mathijs de Boer (PhD student)
Description
Science is characterized by a large network of evolving heterogeneous information sources and knowledge processors of partly/indirectly known, variable quality. Major tasks are therefore the discovery, evaluation, and integration of relevant information. To estimate its value and usability, the science agents have to estimate the trustworthiness of the sources by exploiting different forms of incomplete and uncertain world knowledge. TRIAS is the initial project of a longterm program whose goals are to
- investigate and formally model the factors influencing reliability and trust in science
- build a suitable agent logic of trust and belief with a reasonable semantics
- develop nonmonotonic inference methods for estimating trust from complex, incomplete and uncertain information
- explore the potential of textmining for collecting relevant scientific meta-information
Activities in 2007 (TRIAS)
- Agent prioritization in trust networks based on comparing support trees (principles).
- General framework for information merging exploiting trust networks and local prioritized merging.
- Critical assessment and generalization of the ranking system theory from the perspective of trust.
- Use of many-valued logics for handling contradictory information sources.
- Proposal of a simple numerical algorithm for measuring trust in the presence of cycles.
- Analysis of how to connect qualitative and quantitative representations and techniques (for trust formalisms).
- Initial investigation of a new source-semantic-based framework for resolving conflicting conditional information.
- Identification of the weaknesses of trust logics and exploration of different extensions with more realistic assumptions.
- Investigation of the pros and cons of dynamic epistemic logic for trust modeling.
- Investigation of communicative acts in the context of defeasible reasoning to handle epistemic trust.
- Textual fingerprinting as a tool for classifying and evaluating papers (MINE).
- Attitude mining for the recognition of modalities (MINE).