Data Gathering Bias

Hi,
I work in the area of fingerprint identification for a
correctional facility. We use a methodology called ACE-V in which
fingerprints are Analyzed, then Compared and an Evaluation (match, no
match, can't tell) is made and then the work passed off to anoter
individual who Verifies the work. This is a modified form of
hypothesis testing and is subject to type I and II errors as well.

This work however is not related to latent (crime scene) fingerprints,
so the fingerprints are rolled (digitially) in a controlled
environment and their quality (dependant on the individual) is usually
excellent. The fingerprints are then coded using an algorithm and
matched against a database containing all the fingerprints collected
previously to determine if the person is new to the system or has
previously established identity.

There is a feature in our database matching by where if the name that
is associated with the card exists exactly as it appears on the
demographics associated with the fingerprint images, then the
fingerprints associated with that name from a previous booking come up
for comparison. If a no match decision is registered, then the
fingerprints go on to be searched in the way outlined below (non name
match)

In a non name matching scenario, fingerprints for the top five closest
match to the fingerprints in question come up. These candidates for
matching come up based on their similarity to the fingerprints scanned
in, as determined by the algorithm.

There are scenarios where people who have been previously arrested
have produced very poor quality fingerprints, for whatever reason
(lifestyle, age, profession, poorly rolled, etc). As a result, when
the name matching function is enacted, it pulls up fingerprints that
would not have matched otherwise because the algorithm wouldn't have
seen the same features and scored the fingerprints in the top 5
candidate list, which is all we are presented.

My question (finally) is this: Does the name match feature of the
database create a data-gathering bias? If fingerprint comparison is
really hypothesis testing applied to a specific discipline
(fingerprints), then can you be guilty of restricting your sample
population based upon non scientific reasoning and does this affect
the veracity of the decision produced? In this example the similarity
of two names is not a scientific criteria in my opinion. Many people
come in using false names, and many of those false names are people
who have fingerprints on file.

.