Making the Court Understand and Use the LR From Gut Feeling to Rational Evaluation of Odds Copenhagen, April 2015 Jakob Larsen, ph.d [email protected] Making the Court understand and use the LR Slide 1 ISFG Recommendation Even if the legal system does not implicitly appear to support the use of the likelihood ratio, it is recommended that the scientist is trained in the methodology and routinely uses it in case notes, advising the court in the preferred method before reporting the evidence in line with the court requirements. The scientific community has a responsibility to support improvement of standards of scientific reasoning in the court-room. Slide 2 Overview To speak ”legal” and ”forensic” – two different languages Education of users of forensic DNA investigations What the users of the LR should know Courses Cooperation with lawyers/prosecutors/judges Multidisciplinary publications e-learning/FAQs Communication of DNA evidence in Denmark Mathematical basis for understanding DNA evidence LR or verbalised translation? Illustration of random match probability Illustration of LR The court's use of Bayes theorem The major fallacies Does it work? Slide 3 The court and the scientist do not speak the same language Slide 4 For a person with a scientific background NOTHING is completely sure Are you sure that the sun will rise tomorrow…? Slide 5 No … not completely. But it is a very good theory! Lawyers are comfortable with words, scientists are happy with numbers Is it proved Maybe… almost. beyond reasonable The probability is doubt…? That was not 0,99! much…? It’s usually 1.000.000, isn’t it…? Slide 6 Education of users of forensic DNA investigations in Denmark National Danish Police 3 days Crime scene investigators The Public Procecutor 1 day Procecutors, judges, defense attorneys The Public Procecutor ½ day New Procecutors Court Administration 1 day The Judge Acadamy ½ day Judges Judges National Police School 1 hour Experienced investigators The attorneys Society/ Varies Defense attorneys Danish attorneys Slide 7 The court should… Be able to communicate evidential weight correctly Be able to use Bayes theorem Recognize and avoid prosecutors fallacy Recognize and avoid defense attorneys fallacy Be aware of risks Contaminations Mix-ups Close relatives Other errors Be able to distinguish between cases where the defendant is found due to a DNA database hit, and where the defendant was suspected before the DNA evidence was known Slide 8 Typical reactions from the law people • ”Is this something you have made up at the christmas party”? • ”We are going to decide whether the defendant should behind bars for many years. We need proof – not probabilities” • ”Isn’t the DNA evidence 100% proof? Then it is only circumstantial evidence – in contrast to the fingerprint” • ”Actually, I have always evaluated evidence this way – I have just used Bayes theorem with my gut feeling” • This is just what I have been missing! It can – in principle - be used to assess all kind of evidence in a rational way. Why do we not learn this in college? Slide 9 Before the course: Exercises How can DNA evidence typically be used in connection with a crime case? What aspects of DNA evidence do you find most difficult to use and communicate in court - and why? What specific benefits do you expect from the course on DNA evidence? Slide 10 The purposes of the pre-course exercise are multiple Forces the student to think about forensic genetics in advance – increases the yield Forces the student to realize the expected benefits of the course Tells the teacher(s) what aspects of forensic genetics, that are the most difficult to understand Slide 11 Cooperation with a specially trained lawyer Who? Independent lawyer who has special knowledge about DNA evidence, and can translate from “forensic" to “juridical". University Judge How? Contributions to readily-available presentations Contributions to teaching Monitoring of legal trends Judge Kari Sørensen, Court of Frederiksberg Slide 12 Multidisciplinary publications: Bayes theorem, prosecutors and defense attorneys fallacy Slide 13 Multidisciplinary publications: DNA profiles from one or more persons, frequency databases, transfer of DNA, uncertainty, etc. Slide 14 E-learning https://intranet.ku.dk/Sider/default.aspx Slide 15 If you choose to report the likelihood-ratio Numbers are unambiguous Easy to use in Bayes theorem Slide 16 Difficult for the court to understand Readily leads to the fallacies "We should not use probabilities, we need proof” If you choose to report verbalized translations of the LR The court is accustomed to verbal declarations of weights Easy to communicate in court Slide 17 Different people perceive words differently Necessary to use intervals Difficulty in applying Bayes theorem Mathematical basis for understanding DNA evidence Probability Conditional probability P (6 even) Independent events Odds P( E ) Odds( E ) 1 P( E ) Slide 18 What does the LR mean to the court? How the court may apply the LR “mathematically correct” Likelihood of DNA profile if stain originates from defendant Likelihood of DNA profile if stain originates from random alternative LR = PROBLEM IN COURT ODDS = Probability that stain originates from defendant given info and DNA Probability that stain does not originate from defendant given info and DNA ”If LR is 1.000, odds will increase by a factor of 1.000, when the DNA-evidence is taken into account Bayes theorem: Slide 19 POSTERIOR ODDS = LR x PRIOR ODDS Odds for guilt, ”odds for innocence” and doubt Odds for guilt = Probability that stain originates from defendant given info and DNA Probability that stain does not originate from defendant given info and DNA Odds for innocence = Probability that stain does not originate from defendant given info and DNA Probability that stain originates from defendant given info and DNA Odds for innocence = doubt ”If LR is 1.000, doubt is reduced 1.000 times, when the DNA evidence is taken into account” Bayes theorem: Slide 20 POSTERIOR ODDS = LR x PRIOR ODDS Sit down, and stay seated if you are not... Match probability Wearing a belt Having blue eyes Went skiing last year The owner of af black car The parent of three or more children Having coins in your pocket Born in January Had oat meal this morning The owner of three or more fishing rods Wearing glasses 0,80 0,30 0,08 0,20 0,10 0,20 0,08 0,20 0,12 0,30 Combined match probability 1:12.000.000 Slide 21 Relate the size of the match probability to probabilities the court is familiar with P(Unlock a 3-digit pad lock) P(guess a PIN-code) P(make ”yatzy” in one roll) Slide 22 Relate the LR to evidential weights the court is used to evaluate The weight of a testimony (assume the witness is wrong one out of 25 cases) The weight of the last two digits of a license plate from a perpetrator car matching that of the defendant The weight of a fingerprint match The weight of the defendants statement: “Not guilty" Slide 23 It’s 1000 times less than 1.000.000 Slide 24 It’s 1000 A likelihood ratio that is times What 1000 do you – is this think?? a lot? more than 1 Procecutors fallacy The DNA profile of the stain, is one million times more likely if the stain actually comes from the defendant than if the stain comes from a random other person in the relevant population. LR = Likelihood of DNA profile if stain originates from defendant Likelihood of DNA profile if stain originates from random alternative It is one million times more likely that the stain comes from the defendant than the stain comes from a random other person in the relevant population. ODDS = Probability that stain originates from defendant given info and DNA Probability that stain does not originate from defendant given info and DNA Slide 25 The horse analogy P ( 4legs horse) P ( horse 4legs ) Slide 26 Aha, if LR is 1,000,000, it must mean that the probability that a random person matches is 1:1,000,000. So there must be some five persons in Denmark, the stain is just as likely to originate from ...? Slide 27 Defense attorney’s fallacy! Defense attorney fallacy If the likelihood-ratio is 1,000,000, then the match probability is 1 in 1,000,000. In a population as the Danish, (approximately 5,000,000 inhabitants) there will most likely be five individuals matching the DNA profile. However, this does not mean that these five individuals are equally likely originators of the stain, because this would imply that the prior odds should be the same for all citizens in Denmark. The vast majority of the Danish population will for practical purposes be excluded as the originator of the stain, ie the prior odds are very low or zero. DAF is committed, if you put the weight of the DNA evidence in relation to the size of a population and thus assume that all members of the population have the same prior-odds. Slide 28 Exercise / repetition Identify “prosecutors fallacy" and "defense attorneys fallacy" in the following (not hard to find examples!) Published sentences Crime sections of newspapers Legal literature Try to consider the weight of other evidence with a likelihood-ratio, use Bayes theorem and see if it makes sense. Slide 29 Does it work? ”Prior odds”, ”Likelihood ratios” and combined probabilities are increasingly discussed in Danish courts. The testimony frequency is decreasing How much more in addition to the ”more-than-a-millionLR” do we need if we want to be ”beyond reasonable doubt”? Cases, where the prosecutor has estimated the LR of non-DNA evidence Slide 30 Thank you for your attention! Questions? [email protected] Slide 31
© Copyright 2024