How to account for error in our evaluation has been the subject of several publications. Studies have also shown that it is difficult for laypersons to combine the information regarding the rarity of the genetic feature, for example, the rarity of a DNA profile, and information regarding the probability of error. Because this combination of probabilities is difficult, this should be done or at least be guided by specialists, that is, forensic scientists. In cases involving reported DNA statistics, forensic scientists sometimes provide very tiny numbers, as in the Jama case, along with claims that DNA typing is invaluable. In 2016, the PCAST report has acknowledged that errors happen. In this report, one can read the following: "Examiners have sometimes testified, for example, that their conclusions are "100 percent certain;" or have "zero," "essentially zero," or "negligible," error rate. As many reviews, including the highly regarded 2009 National Research Council study, have noted, however, such statements are not scientifically defensible: all laboratory tests and feature-comparison analyses have non-zero error rates." It is, therefore, important to avoid affirmative statements that ignore the possibility of error. It is important to consider both the potential of error and information on the rarity of the profile. This is true for all traces, but the effect is particularly important when the occurrence of the features of the trace is very small, less than one in 10,000, for example, as is the case for DNA or finger marks. Research has been conducted with jury eligible persons in order to study how well they would combine the probability of error and information on the rarity of corresponding genetic features. Not surprisingly, combining these two probabilities turns out to be not straightforward, and laypersons generally do not perform this task very well. Jurors seem to overweight extremely small probabilities for two reasons. First, as shown by Koehler and co-authors in 1995, in their paper entitled "The Random Match Probability and DNA Evidence: Irrelevant and Prejudicial, it seems that very small statistics, such as one in millions or billions, maybe quite telling and noticeable. This vividness would therefore exert a large impact on jurors' opinions. The second reason why people might overweigh small probabilities is that often, they combine probabilistic items of evidence in a rather intuitive way, like by averaging them. But this, from a logical point of view, is not correct. If one intuitively combines the probabilities, then the effect of error might be thought to be very small. Indeed, the experiment by Koehler and co-authors have shown that when the probability of the profile was given, it had a strong effect on mock jurors' verdicts, whether or not laboratory error probability was provided. Thus, information on the potential of laboratory error appeared to have little impact. However, if we want to account for both the potential of error and the profile probability, one has to combine these two factors coherently. The point is that if, for example, the probability of observing this profile is one in a billion, but the probability of error is one in 10,000, so, several orders of magnitude greater, then, it can be shown, but we won't go into the mathematical details here, that the overall probative value is, by and large, dominated by the probability of error. In particular, the conclusion ought to be that the probative value is way more moderate than what it would be without considering a probability of laboratory error or pollution. So, the results of the research by Koehler and co-authors suggest that the current practice of providing jurors with very small profile probabilities only is potentially prejudicial. But where should we go from here? The point is this. On the one hand, with some reasonable assumptions, we can readily calculate and support profile probabilities of one in several millions, so, very small numbers. On the other hand, from current procedures of assigning error probabilities, it can immediately be seen that it is hardly possible to argue that error probabilities are as tiny as profile probabilities. As we have seen in real cases, for example, Jama, we should also consider the possibility that a trace is polluted by biological material from the suspect. Such a pollution can occur anywhere in the process. Besides, one could also imagine that there has been an exchange between the profile of the suspect and that of a victim in a case in which a majority of traces come from the victim. In these cases, the consequence will be a correspondence between the DNA profile of a trace and that of a suspect. This kind of error is most discomforting and clearly affects the value of the results. We know that laboratories can commit errors. This has been shown both during proficiency testing and real case work. So, if errors occur, how do we account for them? How do we assign the probability of error in the individual case? Notice the particular formulation of the sentence I have just said. The focus is on the probability of error in the case at hand. It is not about an abstract proportion of errors in a sequence of distinct but somehow similar cases. This may be of interest in the study of the phenomenon of error in general, but it is an entirely different question. What are probability for the occurrence of an error in a given case ought to be? Very clearly, it is dependent on and should be based on the particular circumstances of the case at hand. Why? Because it is of no relevance how good a laboratory may work in trials under controlled conditions if the examined trace material in the first place has been exposed to problematic circumstances such as improper handling procedures, incomplete chain of custody, and so on. Stated otherwise, laboratory performance data may not adequately reflect the conditions of the case at hand. So, as I mentioned, there are data available from proficiency testing, but it is very difficult to draw any pertinent conclusions from the various errors observed in these tests for a given case at hand. On the one hand, if a laboratory commits an error once, then this knowledge is not directly usable for providing a value for the probability that an error will occur in a future case. It also does not allow one to draw immediate conclusions about the probability of errors committed by other laboratories. In addition, the simple notion of error is a term that is too loosely defined. As such, it does not allow one to engage in a reflective discussion and lead to concrete insights. If one tries to define it, one realizes that it can cover a variety of phenomena with very different consequences, as seen throughout this video. Besides, one should also remember that laboratory errors may not exclusively lead to false inclusions, but also to false exclusions. In summary, we can thus say that errors that lead to false matches are possible. Their occurrence may be below one in 1000. However, it is not zero. But generally, it is little fruitful to discuss the question of error probability beyond context. The probability of error has to be assigned given the particular context of the case at hand; and that context may be such that the probability of error ought to be assigned as very high, whatever the general performance measures of the laboratory may say. Above all and by all means, there is little doubt that this figure is anything near the tiny profile probabilities that can currently calculated for DNA profiling results obtained with more than multiplex kits. In conclusion, to be balanced, fair, and comprehensive, we need to account for the possibility of error. This ought to be assisted by forensic scientists and the impact of errors needs to be conveyed to the recipient of expert information. What we shall remember is that the probability of the results given the defense proposition cannot be less than the case-tailored error probability of the laboratory examining the DNA in question. Thank you, Alex. To conclude this part, let us summarize the two fundamental points. First, it's inappropriate to present the probative values of a correspondence between the profile of a suspect and that of a crime stain by only considering profile probabilities that are astronomically low, without a case-tailored probability for laboratory error or pollution at the scene or at the hospital or in the lab. As seen in the Jama case, in the report on the inquiry into the circumstances that led to the conviction of Mr. Jama, Judge Vincent commented as follows: "DNA evidence appear to have been viewed as possessing an almost mystical infallibility," and that, "The unreserved acceptance of the reliability of the DNA evidence allowed all involved to leap over a veritable mountain of improbabilities." Second, error and pollution should be taken into account. When we do so, the result have a value that is more modest, due to the unavoidable reality of various sorts of error that can lead to a false positive conclusion. This level is currently unknown, but is an integral part of real case work. Thank you for watching.