CLPEX Chat Board

In another post, Pat asked for a further explanation of circular. Here's my opinion (kind of long so I started a new thread).

Pat,

Circular reasoning is different from a circular process (recursive).

Logical reasoning is when people find sufficient evidence that leads them to a conclusion. It’s not really this simple; there are other criteria as well (like considering supporting and contradictory information, or having an appropriate link between the premise(s) and the conclusion). There are many types of illogical reasoning; circular reasoning is just one form. Logical reasoning can sound linear (evidence leads to conclusions) but there can be many steps prior to arriving at a conclusion. These steps can be circular in nature (a recursive process but not necessarily circular reasoning).

Circular reasoning is the formal term used when people use the conclusion (or part of it) as support for their premise(s). Some examples of circular reasoning are simple and easily seen while other are not so easy to notice. Here are a few examples:
-He is stupid because he’s an idiot. (The reason is a form of the conclusion)
-The building is tall because it is so high. (The reason is a form of the conclusion)
-It is a discrepancy because I’ve determined the prints were not made by the same person.
-It is a dissimilarity because I’ve determined the prints were made by the same person.
-How do you know if ACE-V was used correctly? Because the conclusion will be correct.
The illogical part of the last statement is that we cannot determine if ACE-V lead to a correct conclusion without knowing whether or not the conclusion is correct.

If something is analyzed-tested-reanalyzed-tested-reanalyzed-and testing and then a conclusion is arrived at, then you have a circular process (re-occurring process) but not necessarily circular reasoning. It is only circular reasoning if assumptions used to support the conclusion are part of the conclusion.

The circular process is beneficial in that it allows for additional information to be considered. Re-checking data, either by you or others, is encouraged in scientific disciplines. It is thought that it leads to better results because you may see something you didn’t originally see, and consider things you didn’t originally consider. It’s possible that the circular process can introduce bias by allowing someone to work backwards (working from the conclusion to the premises). Again, this is not ‘circular reasoning’, it is reasoning in a backward motion. The biasing factor is acknowledged and accounted for. One way the bias is accounted for is by trying to falsify your conclusion and allow others to try to falsify your conclusion (not simply having another person reproduce the results). Making the process linear is one way to eliminate bias and circular reasoning but the linear process doesn’t allow for re-analysis, re-testing, and gathering more information. It also doesn’t help practitioners understand different methods that are valued in science (it limits valuable tools available to them and limits their ability to explain the process to others).

A more common form of illogical thinking is using supporting data or premises that seems related but really aren’t (the conclusion is not caused by the premise). An example of this could be stating that a latent is fresh because powder adhered to it quickly, easily, or appeared dark. This is illogical because how the powder adheres to a latent has nothing to do with the freshness of the latent. Or, the latent was left in blood because it reacted with Amido Black. Again, the conclusion is not caused by the premise. Or, the Ninhyrin didn’t work because there was no reaction on the paper. There could be other reasons for ‘no reaction’.

Using supporting data that seems related but really isn’t related is common in statistics and something that needs to be watched for. Here’s an example, Honda’s are 90% more likely than Hummer’s to be involved in an accident. This may be a true statement but these two things may not be equitable since we haven’t considered the percentage of Honda’s and Hummers on the road.

To bring this back to your question, I’m guessing Ashbaugh talked about ACE-V in terms of a circular process and not in terms of circular reasoning (or any type of poor logic).

I haven’t reviewed any of the documents about the Mayfield print but it appears as though the OIG asked the FBI to eliminate circular reasoning and bias, and the FBI eliminated them by having a linear process. I have heard that the FBI does allow for re-analysis but this may not be obvious from this OIG report. Does anyone know if the FBI’s SOP’s on this are available for others to view? If people want to debate the FBI’s procedures, it would be valuable to know what their procedures are.

Hi Michele,

Okay, let me see if I understand this. Consider the following scenario: I have a latent print. There is no core or delta, but there are maybe six relatively clear points with an area of lightly smudged ridges off to one side of the clear ridges where the six points are easily seen. I put the latent up on the screen and put red dots over the six points, but I do not try to run the ridges or define what is present inside the slightly smudged area. Next, I start my comparison with inked prints of a suspect. I find the an area with points consistent with the six I originally marked and I put this inked print up on the screen next to the latent print. At this time, I notice that one of the six points in the inked print is a ridge ending, whereas in the latent print it looks more like a bifurcation.

Now, if my understanding of "recursive" is correct, the path between recursive and truly linear diverges. If I am committed to a truly linear process, I have to simply say there is a difference and stop right there.

But if I am following a recursive process, then I can reanalyze the point in the latent with reference to the inked print and decide that it can be accounted for by heavier matrix on the latent and only very light ink on the inked prints. So I can accept that the difference in appearance is due to a combination of matrix and pressure and was caused by distortion, not an actual difference in the ridge detail itself. Then, if I am following a recursive process, I can work on the slightly smudged area. Finding a few more features in the smudge and matching them to features in the inked print, I declare an identification.

If I were following a strictly linear process, I could not try and reconcile the difference between the apparent ridge ending and the apparent bifurcation, and I could not go into the smudge area to try and sort things out there unless I had completed my analysis of that area prior to looking at the inked print.

But by doing what I did and reanalyzing the ridge ending/bifurcation situation, and by then going into the smudge to figure out if it supported an identification, I have used a recursive process.

Am I on the right track here?

Pat,

Yes, that’s right. Some may claim the examiner was bias but the weight of the ‘additional points’ and ‘the ending ridge appearing to be a bifurcation’ is whether or not you have enough visual information to satisfy others (I intentionally didn’t say other competent examiners). I’m sure people are asking themselves, ‘how do I know if I have enough to satisfy other people?’ The answer is to test it. Send it around to others to see what they think (this can be done blindly or not-blindly). If they don’t see the additional characteristics then the original examiner either has better eye sight or bias is creeping in. One important aspect of this is to ask the most critical people, not the ones you know will agree with you. If your premises hold up to the utmost scrutiny, then you should feel comfortable that you’re not being biased.

Here’s one problem I have with the linear approach. Suppose you have 10 latent prints with 30-40 minutia clearly from the thenar area. You plot the minutia and then look at the thenar and find that the exemplar is not usable. You’ve just wasted quite a bit of time needlessly. The recursive approach is more efficient.

Another problem I have with the linear approach is that it puts weight in ‘when’ the characteristics are seen. Characteristics seen prior to viewing the exemplars seems to get more weight than if something is seen after the comparison has begun. I don’t believe ‘when’ the characteristics are seen should be given weight. The weight should be in whether or not others can corroborate the characteristics.

Pat, I think we have to distinguish between true recursion and that which merely appears recursive. One aspect of true recursion is that there is a core value that is not reached by recursion. In Mathematics, Fib(1)=1 is the core. The recursive function is conditioned, then, on n>1, for which the function Fib(n) must use recursion. Can I express some aspect of latent print examination as recursion? I don't think so. Recursion is a nested set of values. You have to chase the nesting down to the core value. Latent print examination process may invoke the same function repeatedly, but the values are gathered piecemeal and combined linearly to reach a conclusion.

But as to your examination example, it is worth looking deeply into why it is circular process. Perception is not monolithic. We do not see all that there is to see upon first look. The problem is that, upon subsequent look, everything we note may not be perceived just as it would have been had we noted it on first look. When something like your ending ridge/bifurcation situation is encountered, we are remarking, "Oh. Now that I see the other mark, I see what that might be. I just wasn't looking at it like that the first time." We will be making a choice. The ultimate choice may not even be made from what we have observed just up to that point. The issue may be revisited, again and again. We might, for instance, go on to other points and accumulate what we interpret to be some specific kind of distortion. Now, when we go back to that ending ridge/bifurcation point, we perhaps believe we understand the distortion process that gave it the appearance of a bifurcation. And we are perhaps accumulating a weight of consistent points of such number that the apparent discrepancy is discounted.

But a question may be raised as to whether this circularity is a bad thing. I would hold that it is not. We analyze to a purpose. When an opinion of highly probably ultimate truth could have been produced but was not, the analysis fails its purpose. If I were to hold that the entire analysis of the latent be finalized upon first look and no reanalysis allowed, I doom many potentially successful examinations to failure, because human perception is not such a mechanical and mechanically precise thing as that. This kind of circularity is not faulty logic, because the observation of the record print did not prompt an immediate blind assumption that, no matter what I must have seen during first look at the latent, it was not as I interpreted it. Instead, it sends me back to observe again. It is still open. I might reobserve and then declare, "No. They are different structures, and not of the same source."

It is not circular reasoning, because one cannot point out that the conclusion incorrectly derives from a correct observation. There is, of course, a partial reversal of strict linear process. And it may indeed become truly circular, the examiner going back and forth making observations. But the key is that one may present reasonable alternative analyses of the observable data and therefore an alternate conclusion. The issue is argued on the basis of interpretation of observable data and, perhaps, a question of threshold sufficiency overall. Neither an expert who concluded from the same examples by strictly linear process nor the expert who concluded in the usual circular way can, with a wave of the hand, show the other's conclusion irrational. Nor can you say that two experts who render the same conclusion from the same examples must have both worked linear or both worked circular processes. One may interpret the point as an ending ridge on first look, which the other may have decided it was an ending ridge after much back and forth reobservation.

What's important about all this is that circular reasoning is indeed an evil. But I do not see that it can be eliminated in all its subtleties, because the entirely proper circular process cannot be immunized against circular reasoning. Reasoning is not a conscious process. Reasoning takes place in the non-conscious. Our perception of conscious reasoning is an illusion, as any number of sequences of brain activity have found. One cannot be certain that tentative convictions and other knowledge have not joined with other factors to produce an interpretation, especially in the midst of circular reanalysis. One can only immunize the examiner from extraneous knowledge, and that is very often impossible. Thus, the inherent propriety of the FBI decision to report conflicts and thereby render them likely Brady material. (Which may well cause a reexamination by another expert subject to quite different circular reasoning influences.)

Gerald,
I'd have to disagree about your characterization of recursion as not applying to latent print examination. Research methodology cites recursion and iteration quite frequently. I also fail to see a demarcation between 'true recursion' and apparent recursion. Recursive just means self referential repitition of an operation. So the reapplication of a previous step, let's say Analysis in latent print examination for hypothetical purposes, is by definition a self referential repitition of an operation if you go back after exposure to the known.

Here's a few references of the recursive nature in research methodology/hypothesis testing for anyone interested.

http://www.materials.drexel.edu/Program ... Method.pdf
http://urv.academia.edu/DavidJLobina/Pa ... ve_Science

And for those who don't like technical reading, here's a book review that I just found on Friday from a book titled "The Recursive Mind: The origins of human language, thought, and civilization"

http://www.newscientist.com/blogs/cultu ... human.html

Which states:

It allowed us to engage in mental time travel, says Corballis, the recursive operation whereby we recall past episodes into present consciousness and imagine future ones, and sometimes even insert fictions into reality.

and

Instead, he suggests (recursion) came of progressive increases in short-term memory and capacity for hierarchical organisation

From the first link above, you can see that in research methodology, it's the hierarchical nature of the research that makes use of recursion.

This makes sense when we talk about friction ridge examination due to the fact that Evaluations are higher order hypothesis that are consistent with and more complete than lower level hypothesis about the relation of comparative elements used in the Comparison. That is to say, there is a hierarchy of hypotheses implicit in the process. Higher order hypotheses are reducible and coherent with lower level hypotheses. In this way, an Evaluation is both iterative and recursive.

I'm not hard core on the recursion thing. I just think it's too much to bring the examination process into the recursion fold. I think that in attempting to analyse the process, we have stepped off into presuming the nature of a largely non-conscious process. But then, my formal education is in mathematics and computer science, where sharp distinctions are more easily made. Language is another thing. The recursive nature of language is far from universally accepted, again, largely because it's not a process we can observe with much precision.

I would agree that my feeling is that if one were making an attempt to write a software emulation of human latent print examination, one would might likely invoke some recursive functioning or at least think at times in recursive terms. Not that we could truly emulate the process, for we can never say what factors are at work and how they work. That's what makes it expert evidence.