Belief calculation : introduction

When using Factlink to evaluate whether a fact is true, we want to inquire for each fact what the majority opinion is (firstly), and why (secondly). In this document we will introduce the terminology and try to sketch how they are related while avoiding to mention the algorithms to calculate values, because we have different options for those.

Basic calculations

To be able to express majority opinion we first need a representation. We can of course do this by showing only that the majority believes a fact, but we choose a more granular approach. For each fact a user can retrieve the majority [opinion](Terminology - Opinion), a tuple consisting of three values:

• The share of believers
• The share of doubters
• The share of disbelievers

The values in the tuple add up to 1.

Of course this still does not provide a lot of information. Now we know what the majority opinion is, but we still don't know how many research has gone into this opinion. Is this just the opinion of a single mad-man? Or is it the opinion of all the global-warming researchers in the world. To measure this we introduce [credibility](Terminology - Credibility). The credibility is a measure for the amount of brain cycles that has gone into the opinion. You could for instance think of this as the number of people who expressed their opinion, but we can think of more advanced ways to measure credibility.

The credibility together with the opinion provides us with a [weighted opinion](Terminology - Weighted Opinion).

Expressing opinions

In the previous section, we have introduced weighted opinions. Now we will provide a way to actually get such weighted opinions. A user can directly express his belief on a fact. A belief is either "belief", "disbelief", or "doubt". The weighted opinion can then be based on the three lists of believers, disbelievers, and doubters.

Introducing relations

We have now covered the first part (what is the majority opinion), but not so much the second part (why is this the majority opinion). To allow users to express their "why" we introduce relations between facts: factrelations. A factrelation is a directed relation from one fact to another fact which expresses that a fact either supports, or weakens another fact. This means we have two types of factrelations: weakening factrelations and supporting factrelations.

A factrelation is itself also a fact. That means that users can also express their opinion on a factrelation. To enable users to influence the weighted opinion through the factrelations they're adding, while not requiring them to express their belief on the fact itself we also use the weighted opinion on factrelations to calculate the weigthed opinion on the fact they are supporting or weakening.

So if we have a Fact A, which is supported through factrelation R by fact B, we use the weigthed opinions on B and R to calculate the weighted opinion on A.