Merging with unknown reliability
This work tackles a foundational problem in belief merging for AI systems where source reliability is often unknown, providing a theoretical motivation for existing merging methods.
This paper addresses the problem of merging beliefs when the reliability of their sources is unknown. It proposes two approaches: either considering all possible reliability profiles and accepting only what holds universally, or assuming one source is completely reliable but its identity is unknown. These approaches are shown to motivate maxcons-based merging and arbitration, respectively.
Merging beliefs depends on the relative reliability of their sources. When unknown, assuming equal reliability is unwarranted. The solution proposed in this article is that every reliability profile is possible, and only what holds according to all is accepted. Alternatively, one source is completely reliable, but which one is unknown. These two cases motivate two existing forms of merging: maxcons-based merging and arbitration.