Can very smilar nonmatch invalidate weaker identification?

[minu]

First of all, I really enjoy the excellent explanation and defense for latent examination by David Fairhurst, Gerald Clough and others. I agree with you on several points: erroneous identification under high standard is extremely rare; crisp conclusion is necessary in criminal justice system; latent matching process in human's brain is complex to study; and we have not yet found other better ways to report conclusion.

But I am still confused on several questions: whether quantized conclusion (namely, if suspect cannot be excluded, RMP is reported instead of identification or inconclusive) deserves pursue in a long term, whether the process (efforts to clarification of ACE procedure) towards this goal can lead to return in a short term, whether the final goal is just to make defense lawyers happy.

In DNA the maths and the data are there to produce the numbers and, rightly, allow the jury to make that decision (some have called it a "leap of faith", others have said "it's no great leap"). In latent prints the data and maths are not there. We have to help the jury and make the leap for them. That is the role of an expert witness.

If quality and quantity are sufficient (DNA profile with 10 markers, a latent with 20 L2D), there is no problem for either the jury to make the leap or the expert witness to help the jury make the leap. depends on the number of available marker.

But let's assume a hard situation (I don't know whether this is common or not). If you only have partial DNA profile (say, 3 markers), the RMP may be 1 in one thousand. The DNA examiner will present this RMP number in the court and let the jury to make the leap by considering other evidence. But if you have a latent with strength similar to this 3-marker DNA, what should a LPE do? Should the LPE "help" the jury make the leap by combing latent and other evidence, since the latent is not sufficient for making the leap? Do the jury know that the RMP of the latent is actually 1 in one thousand? Or should the LPE report it as inconclusive?

I'd back a professional Baseball team against a load of robots every time, but as a word of caution "Even the best catcher occasionally drops a sitter!"

This is a good analogy. I'd also back a latent examiner if there is a game of examiner vs. robot (AFIS) in terms of accuracy. But more importantly, outside the game, this robot is the cooperator of examiners, rather than an opponent. Helping improve this robot is helping examiners. I believe that clarification of ACE will provide good ideas for improving AFIS. But, in order to coding new tricks of matching latents into computer, the tricks have to be very clearly and quantitatively defined.

Respectfully
Jianjiang Feng
MSU

[David Fairhurst]

Finally I understand the proposition.

Latent A is individualised by an examiner.
Latent B is inconclusive.

Someone develops a computer system which can work out a random match probability based on certain features that can be observed in latent prints. This robot can be used to present evidence to the jury, consider it alongside all the rest and decide whether to make the leap or not. I'm in favour of this; as long as the model is robust enough.

Each of the latents above is entered into the system and latent B comes out with a lower RMP than latent A.

Does this invalidate the individualisation of latent A?

No.

The data set used by the RMP system is limited. The examiner is able to process much more information than the computer.
The true RMP of friction ridges is zero. A statistical model cannot give a probability of zero. The better the model is the closer it will get to zero but because it is a model it will never give the true probability.

The higher RMP of latent A over latent B can easily be explained by the limitations of the model and the restricted data that it works with.