Information Evaluation

Chapter 5

Multidimensional Approach to Reliability Evaluation of Information Sources

5.1. Introduction

Knowing the reliability of one’s information sources is useful for a number of reasons: it helps to order them, sort them and also to exploit the information which they transmit. In particular, if a source provides testimony in the form ω ∈ A, where ω denotes, e.g., the type of weapon used in the 31 May bombing in the capital of Ektimostan¹, and A is a subset of the set Ω of the different possible types of weapons, and if that source is reliable to a degree α, between 0 and 1, then the operation of discounting introduced by Shafer [SHA 76] attributes the coefficient α to the fact of being able to categorically state ω ∈ A, and the coefficient 1-α to ω ∈ Ω. This transformation of the information emitted by a source is based on the following model of the notion of reliability: if a source (a sensor, an agent) is considered to be reliable, we agree to enrich our knowledge base with the information which that source gives us, and if he is not reliable, we do not take the information transmitted into account [SME 93]. This commonly-accepted view of reliability is similar, though different, to other notions such as sincerity of sources or trust (see [DES 11] for a comparison of these notions). Note that information provided by an unreliable source may very well be true. For instance, a broken clock will show the right time twice a day, but since there is no way of knowing whether the ...