Definitions of `confidenceMethod` in this spec and VCDM are not consistent

[iherman]

The definition of confidenceMethod

• in the VC Spec:

  A property used for specifying one or more methods that a verifier might use to increase their confidence that the value of a property in or of a verifiable credential or verifiable presentation is accurate.

• in the Conf. method draft:

  Each confidence method is bound to one or more subjects in the verifiable credential, and provides enough information for a verifier to determine whether the holder can generate a verifiable presentation to increase the verifier's confidence that they are the same entity referenced by the confidence method.

The two definitions are inconsistent. The VC spec seems to be fairly open ended; the way I read it (although the sentence is difficult to parse) is that a confidence method could be used for any statement in a credential. On the other hand the definition in the draft strictly refers to the credential subject(s).

I presume the plan is that the definition of confidence method will be removed from VCDM 2.1 in favor of a reference to the new spec, but I just wanted to flag the inconsistency. Ie, that what we will specify does not correspond what we described in VCDM 2.0, and we are ok with it...

[TallTed]

YES!

I have always subscribed to the VC Spec's rather open definition, as I have tried to express in recent calls.

I am not OK with the current Conf. method draft definition, which radically shrinks the utility of the confidenceMethod, as it ties the confidence to a subject, rather than to an attribute of that subject.

[denkeni]

A property used for specifying one or more methods that a verifier might use to increase their confidence that the value of a property in or of a verifiable credential or verifiable presentation is accurate.

This means every claim about the subject could come with one or more confidenceMethod, right? I’m curious what kind of use case would it be like?

[jandrieu]

I am not OK with the current Conf. method draft definition, which radically shrinks the utility of the confidenceMethod, as it ties the confidence to a subject, rather than to an attribute of that subject.

That would be a very different sort of thing.

The confidence method in the current spec is, as you point out, about confidence that the data subject in different context is the same subject. Specifically, providing mechanisms for the verifier to check themselves.

The confidence in a given attribute is a matter of trusting the issuer's judgment. I can see value in an issuer expressing some form of confidence, but that would be a different property than the one we've been advancing in the Conf. Method Spec.

[TallTed]

This means every claim about the subject could come with one or more confidenceMethod, right? I’m curious what kind of use case would it be like?

In the terminal case, yes, there could be one (or conceivably more) confidenceMethod per claim. I expect that would be a rather rare case, as I expect that any given confidenceMethod would support multiple claims/attributes regarding a given subject.