PC> The PREMIS Editorial Committee received a request to include an
PC> optional "certainty attribute" in the Data Dictionary to indicate the
PC> degree of certainty that the value provided for a particular semantic
PC> unit is correct. Although the requester thought it might be of use
PC> for all semantic units, the specific use case was in reference to
I'm the one who made this request.
Yes, file format is, for me, the element which would most benefit from
some kind of certainty information, although it could also be applied to
some 'significantProperties' and other 'objectCharacteristics'.
Anyone could devise their own system, but I think that certainty is too
important to be "local business information" that can't easily be shared
with or understood by others. It would be nice not to have to generate
invalid PREMIS documents, just to preserve certainty information.
As you said, there are many ways that certainty can be expressed - a
binary yes/no, a proportion or percentage for cases of conflict, or
special cases like incomplete or sparse information.
A 'certainty' element could have a type attribute, indicating the way that
the uncertainty is being expressed:
<!-- not certain that value is correct for some reason -->
<!-- 3 out of 4 tools have picked this value -->
<!-- only one tool could provide a value -->
<certainty type="nominal">single result</certainty>
Implementers could decide which types to support in their applications,
but something like the above is reasonably legible to humans.
As Rob has pointed out, uncertainty is everywhere, and certainty is more
of a current lack of uncertainty, which should be regularly tested. But I
think it is worth trying to preserve this important information, if only
for the future when we have developed some way to act on it.
Have other people tried to solve this problem in other ways?
The University of St Andrews is a charity registered in Scotland: SC013532