Double Blind Verification

[Dan Perkins]

Hi,

I'm aware that the definition of double blind verification specifies that the verifier would not know they are being tested but how might this specifically relate to comparing friction ridge detail impressions?

Is blind testing where the verifier would not know the conclusion of the previous examiner(s) and double blind where they would not even know they were a verifier?

thanks

Dan

[David Fairhurst]

In a double-blind experiment, neither the participants nor the researchers know which participants belong to the control group, as opposed to the test group.

The term cannot really be applied to latent print verification (nor is there any need to try) as we are conducting "tests" not "research".

"Blind verification" means that the verifier does not know the results of the previous examination (and may also not know that there was a previous examination.)

[Dan Perkins]

Thanks David

Dan

[L.J.Steele]

There's been some discussion about using double-blind verification for close non-matches so that the verifier (if they are aware of their role) can't assume they are looking at something someon else thought was a match.

[ER]

That would be considered a blind verification in some offices. There is still no clear distinction as to what a 'double blind' verification is.

.... and there is still no research that demonstrates that blind verification produces improved results over non-blind verification. I would suggest that once that research is completed, it will demonstrate just the opposite. That a rigorous non-blind verification where the verifier has been trained to assume that the initial conclusion is an error will catch more errors than the mere repetition of a blind verification.

[josher89]

I agree with ER...good science relies on that principle: state a hypothesis, from that hypothesis, create a null hypothesis and attempt to prove the null hypothesis. If you can't, than you must accept the original hypothesis.

Science (latent print exam)
Hypothesis: Gravity pulls things towards that of greater mass--objects fall down when dropped (Joe Schmo's left index fingerprint made this latent print collected as L15)
Null hypothesis: Gravity has no effect on objects of smaller mass--objects float or move in unpredictable ways when dropped (Joe Schmo's left index fingerprint did not make this latent print collected as L15)
Testing: Drop a golf ball and see what happens (A-C)
Results and conclusions: Objects do fall downwards when dropped; earth's mass impacts the mass of the golf ball more than the mass of the golf ball affects the earth (E)
Peer review: Publish (V)

In trying to prove the null hypothesis (it couldn't be proven), we are forced to accept the original hypothesis. In the case of science, we have experimented to see if we could disprove the original hypothesis. In latent exams, we are trained (or should be trained) to try and disprove the original conclusion (or hypothesis--ID or no ID) but we need to know that in order to disprove it.

It's like Mankevich's article I mentioned on another thread; there is a different between independently-performed experimentation and independently-replicated conclusions. Search the threads for his article I linked (I don't want to regurgitate it here).

I think that a research project that entails the distinction between blind verification and non-blind verification will show that blind verification isn't going to improve the reliability of a conclusion or catch more errors.

Eric, you and me? Think about it...I know we're both busy but perhaps I can see something on the horizon.