Tuesday, January 01, 2013

The Role of Uncertainty Within the Application of Inductive Processes in Forensic Science


        Friction ridge individualization is a formalized experienced based reasoning process.   This process is a form of recognition that has been formalized to allow for the organization and illumination of supporting proofs regarding a specific conclusion.   An inductive type reasoning process is rather different from the deductive process.  Essentially, deduction is the investigative process that proceeds from the general to the particular. [1]  When properly applied deduction is, in a sense, self-supporting cascade of related facts.  Known facts can be used to take the deduction process to the next cognitive and relevant level.  Induction type reasoning is in most aspects the opposite of deduction.  Induction type reasoning scenarios utilize inference that proceeds from the particular to the general.  More specifically “Induction is the inferential processes that expand knowledge in the face of uncertainty.” [2]  A general hypothesis or theory is organized from the examination of particular relevant details and how these details are correlated.  [3] The supporting fact structure is in reverse.  Not incorrect, just built from the particular axioms and logic.  Another difference between deductive and inductive reasoning type processes is in the formal proofing or testing of a hypothesis.  We are all familiar with the summation of deductive problem solving.  Proofs are logical, orderly, and are generally systematic.   Induction processes, on the other hand, is the inverse and has a conceptual similarity to quantum theory.   The parallel is a reliance on the probability theory in that all information relationships and their values cannot be sharply defined and must, in an absolute sense, be probability modeled, whether in strict sense or intuitively.   In the words of the physicist Richard Feynman “nobody understands quantum theory.” [4]   Interestingly, there is no established theory specifically for the induction inference process, and nobody truly understands it either.  Complexity, probability, uniqueness, stochastic variables and the related principle of non-specificity, highlighting the uniqueness of the process, outline its use. [5]  Axioms must be applied.
            According to theorists, a theory for the induction process would include two main goals.  First is the possibility of prediction.  Secondly, such a process would allow for direct comparison where observations can be made.   Prediction and direct comparison are fundamental components of the friction ridge examination process.  [6]  The problem with identifying a theory then must be within the application and proof of an inference based hypothesis.   The missing link is the consideration of uncertainty.  This uncertainty is the variability in the quality of information, which is a component of complexity.  Understanding this complexity, the principle of non-specificity, measurement theory, minimum sufficiency, information theory, and the notion that absolute proof, which always proves elusive, is the key to fine-tuning this science. 
Regarding friction ridge examination, inference hypothesis proofs are often based on specific characteristics of nature found within inter-related concepts such as randomness and uniqueness.  When the reasoning process is applied, prediction and direct comparisons must be employed in order to understand the relevance of the information analyzed.  With comparative forensic sciences, the principle of individualization is the name associated with the concept or axiom of “practical proof.”  This is essentially, the truth of the matter.  The principle of individualization is that the difference between tangible objects provides a practical proof about the subject’s uniqueness and the fact that it can be individualized within such limitations [7] outlined by such things as measurement theory, probability, and communicated via information theory.  With that said, why would uncertainty play such an important role? 
Quantum theory states that our ability to finely measure things such as electrons and photons is not only beyond our measuring abilities, but rather impossible.  A threshold is crossed from classical knowledge, such as measuring the ballistic dynamics of a bullet, into a realm that requires statistical modeling to offer probabilities on specific measurements.  Yet, if we think about it, all of our measurements are approximations, which mean they contain some degree of error.  Furthermore, we must consider that in most cases we utilize an intuitive form of probability comprehension that regrettably, enhances the system noise or stochastic variables we associate as uncertainty.   With information theory uncertainty is a measure of entropy.  
When accuracy is paramount, we push to measure things to tighter and tighter tolerances yet we still never reach perfection. Only when we deal with pure mathematics do we see absolutes.  The concept of 1/2 is absolute, but to measure exactly 1/2 of something is a different problem.  This variability is similar to quantum theory in that accuracy of information regarding inference cannot be precisely quantified in absolutes.  This is partly due to complexities or stochastic variables of the process, yet it is mainly due to the fact that there can be an infinite number of ways to approach and reason a solution with inductive based problems.  This makes each application of these cognitive complex functions unique in themselves, this is the "principle of non-specificity" and is measured as aspects of the stochastic noise and uncertainty in the process.  A measure of entropy in other words.  All information including, both the application of experienced based reasoning and the analysis of the subject itself, is on a variable informational scale of availability, discovery, quantity and quality which is called noise in information theory.  Accordingly, absolutes actually become sufficiencies or “practicalities.”  Practicalities are hypotheses or particular points at issue that fall within expected and acceptable statistical parameters of accuracy relating to the issue.  In other words, the greater the accuracy of the information used in the process, the more valuable a resulting conclusion or hypothesis will be.   Error correction within this process can realized with high levels of quality control to include such aspects as evaluation and testing. This is not a perfect solution.  It is a real world solution in light of elusive perfectness.  If distortion of information can be accurately understood within a practical sense, the information has a higher degree of value.  This is not unlike most scientific theories.  They are fine-tuned, yet known to contain some degree of error.  However, their main hypotheses are not necessarily in error.  This type of logic can help us understand how comparative forensic comparison science works, is supported, and how it can be validated in the face of uncertainty.  Information theory and error correction protocols fine-tune for forensic comparison help mitigate uncertainty.  
Confirmation of an inductive type hypothesis is isotropic in that all information may be valid and relevant.  Researchers have noted that it is impossible to put prior restraints on what might turn out to be useful in solving a problem or in making a scientific discovery. [8]  Again, this is related to the introduction of stochastic variables and the principle of non-specificity where this new information and process will be unique.  Friction ridge examiners are cognizant of the quantity vs. quality variability of information used in the development of a hypothesis.  The point here is that the very application of reasoning is variable in quality as is the information itself.   Recognition of uncertainty within the process itself is the missing detail needed for a proper theory the inductive reasoning process, specifically forensic comparison science.
How can variability in information quality be quantified?  If it cannot be measured directly, then it must be assigned values according to statistical models.   Thus, a “statistical understanding of information quality variability” must be folded into a theory of induction if the theory is to have practical value.  Thus far, the role of uncertainty has been regulated to a very informal intuitive position within the forensic sciences.  Accordingly, four items or considerations would be needed for a practical induction-inference theory are as follows:

Four Considerations Of An Induction- Inference Based Theory

1.     Predictability.
2.     Direct Comparison with Observation.
3.     Stochastic Variable Mitigation / Error Correction
4.     Degree of Uncertainty

The fourth consideration is essentially the need to understand how the variability of information quantity and quality, its degree of uncertainty known as noise, affects induction-based hypotheses within a failure criterion frame of reference.  This includes both specific issues and the hypothesis as a whole.   Thus, degree of uncertainty is a probability based accountability for the variation of the information’s quality, quantity and error within the process.  With friction skin examination, this may consist of such familiar details as; specific error rates, training, experience, and the holistic application of reasoning skills in all phases of the methodology.  These points would need to be quantified in some practical manner in order to make a hypothesis’s value fully understood.  Testing the proofs of an examiner’s conclusion of individualization or exclusion would have little meaning if the degree of uncertainty were not considered and appropriately understood in context.  The value of information components utilized must taken into account.  A false analysis inevitably leads to a false conclusion. 
With friction ridge individualization and exclusion, the introduction of low quality information has a high probability to lead to false positives.   This is where the concept of “trained to competency” enters the equation.  Training and experience are paramount in minimizing the inclusion of poor quality or inaccurate information into the comparison process.  It also reduces inferior reasoning. 
            Friction ridge examiners are well versed in the first two topics of such an induction theory, yet the last two points of non-specificity and uncertainty are often the point of contention regarding a hypothesis and proof of specific issues, including the illustration of errors.  Whether the issue is the validity of a single characteristic or the proof of individualization, all the information being used in the process must be practically evaluated to ensure that it falls within expected and acceptable parameters.   Our perception regarding factors of uncertainty should be addressed in a scientific manner.  When the goal is accuracy, accuracy of information, and accuracy of reasoning are the means.
            While we have outlined what a practical theory of inference must contain, we have not dealt with the issue that if such a theory is actually possible or how it must specifically fit within Information Theory.  It is my perspective that the parallels of rational consciousness and inference are unmistakably intertwined and incomprehensibly complex in their holistic nature, yet can easily be defined as unique with unique being a definition of complexity.   This leads us to the notion that if consciousness in non-computable, the same must be said of the isotropic inference processes.   The infinities of the systems and processes cannot be simply cancelled out as they are the embodiment of the process or system.  However, methodologies utilizing this cognitive process can be of value with their power of discovery and information organization.  
            Our hypotheses utilizing the inference process can be verified with repeatability of the results rather than false expectation of "exact repeatability" of the cognitive process as prohibited by the non-specificity principle.  The final comparisons of forensic comparative hypotheses will then allow for a practical evaluation regarding the role of uncertainty.  We can be certain that such an axiom based complex system cannot be perfect.  However, it is formally scientific, practical, and when the methodology is properly applied, it is found to be sufficiently accurate.  To toss out inductive processes as unscientific is to toss out a large amount of reality.  "Sufficiency" is all we can hope for in any formal cognitive process.

Craig A. Coppock  CLPE
May 11, 2005 / Updated May 14, 2017

  1. O’Hara, Charles E and Gregory L. 2003: Fundamentals Of Criminal Investigation,  Charles C. Thomas Publisher, Springfield: p. 885
  2. J. A. Scott Kelso: 1999 Dynamic Patterns: The Self-Organization of Brain and Behavior.
Brandford Books, Cambridge p. 38
  1. O’Hara, Charles E and Gregory L. 2003: Fundamentals Of Criminal Investigation,  Charles C. Thomas Publisher, Springfield: p. 887
  2. Gribbin, John ; 1995 : Schrodinger’s Kittens and the Search for Reality : Back Bay ; New York ; p. vii
  3. Coppock, Craig 2004 :  A Detailed Look At Inductive Processes In Forensic Science:  The Detail 4-2004,  and 5-2005  clpex.com, updated 8-24-2008 (Complexity of Recognition)
  4. Holland, J; Holyaok, K; Nisbett, R; Thagard, P. : 1986 Induction: Processes of Inference, Learning, and Discovery. MIT, Cambridge  p. 347
  5. Coppock, Craig 2004 :  A Detailed Look At Inductive Processes In Forensic Science:  The Detail 4-2004,  and 5-2005  clpex.com, updated 8-24-2008  (Complexity of Recognition)
  6. Holland, J; Holyaok, K; Nisbett, R; Thagard, P. : 1986 Induction: Processes of Inference, Learning, and Discovery. MIT, Cambridge  p. 349

No comments: