confidence

[Mr Fortuner connection modem]

About the two aspects in the assignment of "confidence" proposed by Peter
Rauch:

First situation described by Peter Rauch:
1. An agency/organization
which must make management/development planning decisions submits some
specimens. You, the taxonomist, reply with the determinations, appropriately
weighted by the mother of all "confidence" measures. Now, the "customer" says
to you  --"This is a life and death (or a million-dollar) matter. Are you
_really_ sure about this determination." You say, "Well, wait a minute. Let me
take another look at it."

First, what was the confidence measure (or my
endorsement factor) attached to the original identification? Let's examine two
cases : very low confidence and very high confidence.
If the confidence was
very low, obviously you are not "really sure" about the identification. Now
that you know it's a life and death matter, you should try to increase this
confidence level: send the specimens to someone with higher expertise, improve
the observation set-up, look for more characters, ask for more/better
specimens, i.e., act on the taxonomist- or taxon-centered confidence scale.
If
the confidence level was already very high, "taking another look at it" means
double-checking your observations, see how confident you are about each
individual piece of data, in other words, check what I call "PIF". Here again,
you act on the taxonomist- or taxon-centered confidence scale.

Second
situation described by Peter Rauch:
2. You provide the agency/org with a
determination, X_us y_us, that you've given a "confidence" measure of 8 on a
scale of 1-10 (10 = perfect knowledge); another series of similar specimens,
X_us z_us, with a 5; and yet another, different but closely similar series,
X_us ?b_us, with a 4. The agency comes back and says "But can you tell me if
any of these might be the rare species X_us g_us?"
You might say, "Well, no,
not a chance." Or, you might have to say, "Possibly, but if it is, then I'd
have to call it X_us g_us with a score of 2."

Here, the question is not
about your confidence that species y, z, and b are present, but that species g
is absent. You could double-check your identifications as in 1), but you also
should compare the 3 series of specimens with species g and come up with a
measure of dissimilarity between each series of specimens and g (if you have
access to the proper identification tool), together with the corresponding
taxonomist- or taxon-centered confidence rating (e.g., series 1 is 95.9%
dissimilar with species g, with a confidence scale of 9).

The situations
described by Peter Rauch are quite real, but I believe that the answers to the
questions raised by the agency/organization can be found only in taxonomist-
or taxon-centered confidence scales.

Renaud
Fortuner
fortuner at math.u-bordeaux.fr