Case Specific AFIS implementation

[hilmo]

Has anyone implemented Case Specific AFIS software into their workflow? Has it improved your workflow?
Do you work in a large city, state or smaller jurisdiction?
Have you tried the CSIpix CAFIS software? What do you think? What would you change, add, improve on?

Thanks!

Hillary Moses Daluz

[Boyd Baumgartner]

Our AFIS (morphotrak) has this ability built into its functionality (minus ad hoc scanned cards). You can narrow down the pool of subjects to one if you'd like by entering their AFIS numbers or to just the subjects in a case as long as they have AFIS numbers. In practice it's used in more of a 'hail mary' capacity or as a quality assurance measure. We have the ability to declare an impression Inconclusive due to ambiguity (e.g. an impression whose orientation is questionable), so we don't run the same statistical risk of marking a false exclusion because we aren't forced into that conclusion. The problem it poses I would say, is diluting the standard for AFIS entry and by extension greatly increasing the work in a case. In theory, you can put any print into AFIS, and if you do it for a limited subject pool, why not for an extended one? And if you do it for a local AFIS why not regional and national ones?

[hilmo]

Boyd,
Thank you for your response. Case-specific AFIS has been proposed as a time-saving, triage tool. Yet I have the same questions you do. The agencies who are experimenting with this workflow are also entering elimination prints in order to save time by weeding out prints that are not critical to the investigation. It is an interesting paradox.

[josher89]

I think this may be helpful in situations where you have decent detail available in a latent (let's say suitable for ID) but you just can't find it with the impressions you were given (maybe you can't ascertain orientation, location on a palm, etc.) so you just want to give the 'machine' a crack at it. If you still can't find it (or CS-AFIS can't either), you can still say Incomplete or Inconclusive or whatever you can say but you've at least tried something else that was available to you (if it was).

[g.]

Hilmo:

1. I have a research paper on this if you would like a copy... glenn@eliteforensicservices.com

2. I agree in part, with you and Boyd. Particularly as Case Specific AFIS applies to using the Big Network AFIS database and doing closed searches (as the vendors call them). We (Carey Hall and I) were against this approach from the start for the same reasons you described. What we wanted was a separate stand alone AFIS that allowed the fingerprint examiner in certain cases to enter the knowns of the victims/elims/possibly suggested suspects in the case, etc. with no entry restrictions. We wanted to be able to search fingers, tips, sides, mcps, joints, morgue prints, officer field elimination print kits, palms, feet, toes---essentially any friction ridge skin exemplars you have that could be encoded. We didn't want the restrictions of BIG AFIS telling us it had to be in ANSI NIST standard format on these cards and possibly even get network/CJIS permission to enter them. So we worked with a vendor to develop a prototype for that purpose.

3. We explored the use of this in 3 ways: A) a manual search replacement, time saver; B) as QC/verification on exclusion decisions (or maybe "no identification effected inconclusives"); C) as a case audit tool/QA tool to search for erroneous exclusions/missed opportunities for IDs after the fact (perhaps useful for cold cases).

I am not convinced it will be a big time saver in every case, but it will be useful in those large cases with lots of latents and lots of suspects/elims requiring lots of searches over a very long time, with new elims/suspects being added to the case.
Where I was most impressed was using it as a check on the non-IDs (excl. or inconcl.). The Army Crime Lab is doing something similar to this in their case work flow--using AFIS as a QC check. I strongly believe this will reduce the number of erroneous exclusions. The errors we found in research and in anecdotal cases were mostly due to incorrect assumptions about the latent print (orientation, location, etc.).

4. It wasn't the marginal, thin, possibly inconclusive latent prints that we were worried about, where an examiner will now push the envelope on searching smaller crappy LP fragments because the database shrunk. Many of the misses we were seeing, and the CAFIS tool was finding, were relatively clear latent prints that were oriented incorrectly, mistaken for palms v. fingers, or the person was just looking in the wrong spot. CAFIS was easily detecting these misses in the elimination prints or suspect cards. See research paper for some examples.

Good luck with this approach. I think it has value if thoughtfully used and included as a tool for the right job.

There are some vendors out there further developing this tool. I'm getting involved with some new research using this tool. John Maloney in Australia has been developing a version of this as well (Fingerprint Comparison Software-FCS).

It's an incredibly useful tool if the examiner has control over the kinds of exemplars that can be entered, particularly problematic elimination prints. But also, needs to be the right tool for the right job. If latent and exemplar entry and encoding are fairly quick and accurate, the time spent on the front end to load a case quickly outweighs the time to do a lot of manual searches.

Another thing we thought was that agencies that are reluctant/unable to go to a verification scheme for exclusions/inconclusives may want to consider this as an alternative QC measure. The great thing about it is, as demonstrated in the paper, you could have non-latent examiners do the entry/searching/checks. It could be done by a processing tech, AFIS tech, 10-print tech, etc. Suspected "hits" (missed IDs) would then be forwarded to the case examiner to review. It's certainly better than "no check at all".

g.

[g.]

Josher89,

Exactly!