The Black Swan: Second Edition: The Impact of the Highly Improbable: With a New Section: On Robustness and Fragility: 2 (Incerto)

Product Information

There have been plenty of other heroes along the way – head of recruitment Steve Walsh has crafted an outstanding scouting network so essential to the building of this history-making squad; fans’ favourite midfielder Andy King, meanwhile, has become the first player to win the top three divisions with the same club, and now prepares for a final-day trip to Stamford Bridge – the ground on which he used to be a ball boy in Ranieri’s Chelsea days. In the Venray murder case, the Dutch Supreme Court determined on what grounds courts may reject the alternative explanations offered by defendants and when they should justify their decision to do so. In this case comment I offered an interpretation employing Bayesian probability theory. In the following sections I will argue that prior probability only plays a key role in one of the three criteria that the Supreme Court mentioned, namely whether the explanation needs to ‘become plausible’. I will look at this criterion next. For the other two criteria we need different concepts, which I discuss in Sections 4 and 5 respectively. In Section 4 I argue that whether an explanation is ‘incredible’ depends on the credibility of the defendant. Finally, in Section 5 I argue that whether an explanation is ‘highly improbable’ depends on how obvious it is that the explanation is improbable. 4. Implausible explanations fail to become probable They know that much of their success this campaign has been down to other sides failing miserably, and that next season offers them all a shot at redemption; Manchester City and Chelsea with incoming managers, Manchester United too if Jose Mourinho gets his way, Arsenal in what might be Arsene Wenger’s final season at the club, and even Liverpool where Jurgen Klopp now has a full summer of fine-tuning ahead. Mauricio Pochettino’s Tottenham, who ran them so close, are built to last. There are two reasons why an explanation can be improbable: due to the evidence in the case (likelihood) or due to its prior probability. If an explanation is improbable due to the likelihood, it conflicts with (reliable) evidence that was already brought forward in court. In such cases the court can point to that evidence when justifying its decision to reject the explanation. Yet the Supreme Court’s ruling is about cases in which courts cannot point to such evidence. So, the court presumably describes situations in which an explanation is improbable because of its low prior probability. The prior probability of an explanation is its probability before any evidence is observed. The lower the prior probability of an explanation is, the stronger the evidence has to be to make that explanation probable. If the evidence is not strong enough (in terms of the likelihood) then the prior probability will not be raised sufficiently to create a reasonable doubt. In that case, the explanation has not become ‘plausible’.

When reasoning about which story to accept, rejecting one story and accepting the other often means finding ‘discriminating evidence’, i.e. evidence that fits better with one story than another ( Van Koppen, 2011, pp. 52–55). In Bayesian terms this means evidence where the likelihood ratio strongly favors one story over the other. Such evidence discriminates between the two explanations because we would expect the evidence much more if one explanation were true than if the other was. If the likelihood ratio is sufficiently much greater than 1, the probability of one explanation will be high and the probability that the alternative explanation is true will be low. In such a cases the court can point to the discriminating evidence as a reason why it rejects the alternative explanation. Before the promotion-winning season of 2013/14, there had been so little to enjoy over the previous decade. League One was fun as it turned out, if only to rediscover a winning feeling lost for six years, but the two Championship play-off semi-final heartaches that followed weren’t so much. In fact, Leicester have won an incredible 14 matches by a single strike this season – seven more than Arsenal, and eight better than Tottenham and Manchester City. What does an explanation with a low prior probability look like? First, it may have parts that do not fit well together. For instance, the explanation might imply that the defendant was in two places at the same time. Alternatively, the defendant may tell a story in which motive and action do not fit well together, such as a story about a robbery where nothing was stolen. Finally, the explanation may consist of a number of independent and individually unlikely events ( Lettinga, 2015, p. 53; Josephson, 2000). Second, an explanation can also have a low prior probability because it does not fit well with our generalizations about how the world typically works. For instance, we may believe that innocent bystanders do not run away from the police, that a suspect cannot cross the city in 10 min or that the police rarely forges evidence. The more an explanation violates such generalizations, the lower its prior probability is. 5. Incredible explanations are told by an unreliable storytellerJust don’t call it a fairy tale. If the phrase has proved a worn cliché over the last six months anyway, now it’s simply not true. The Collins Dictionary defines ‘fairy tale’ as being “a highly improbable account”– but this has happened. This is very real.

All this is supposed to be satirical but it has little or no resemblance to life as we know it, without which satire is impossible. Just as I was trying hard to become interested in the advertised state-of-the-nation-satirical-political novel, it suddenly turned thriller, and pretty daft thriller at that: the prime minister’s senior private secretary is murdered because he threatens to blow the whistle; two other people from the political classes are hit on the head, one with a wrench and the other, if I remember clearly, with the butt of a gun. The young woman who was hit on the head with a wrench, Jen Lewis, is the daughter of Myfanwy Davies-Jones, who is, I think, supposed to evoke Molly Parkin. Various grandees such as historian Lord Briskett and the mysterious gay fixer and strategist Alois Haydn pop up. Haydn, who is bankrupt, decides that, although he is the chief adviser of the yes camp and the architect of the conspiracy, he will leak what has happened to the opposition and at the same time short Britain, making himself many millions when the markets crash. So this becomes a financial fraud novel as well. Haydn’s Indian partner leaves him in disgust, perhaps to demonstrate that there is some decency and sense in this world. Vardy broke a Premier League goalscoring record, Mahrez won the PFA Player of the Year award and next season they have the sweet sensation of Champions League football heading to the East Midlands – as England’s Pot 1 representatives, no less. But they know things won’t be like this again. At the heart of the Supreme Court’s ruling is the idea that courts can reject a defendant’s explanation even in cases where the evidence does not refute this explanation. While rejecting the story by referring to a ‘smoking gun’ (i.e. refuting evidence) may be the ideal, other responses are possible too. The Supreme Court distinguishes three categories. First, some explanations can be rejected because they ‘did not become plausible’. I argued that whether an explanation needs to ‘become plausible’ during the criminal proceedings depends on its inherent plausibility at the time it is offered—its prior probability. If an explanation with a low prior probability does not become probable by means of the evidence, then the explanation fails to create a reasonable doubt. Second, some explanations are ‘incredible’. Whether an explanation offered by a defendant is probable partially depends on evidence about the credibility of the defendant. Finally, some explanations are so ‘highly improbable’ that the court does not have a duty to respond to them. I argued that what distinguishes these explanations from explanations that the court should respond to is that their improbability is obvious. When an explanation is obviously improbable, the court would not serve the goals of making its decision understandable by offering a response. A duty to respond would then only reduce the efficiency of the decision process.So, the Supreme Court’s ruling is about how courts should deal with cases in which the evidence does not ‘refute’ a defendant’s story, but they still wish to reject this alternative explanation. Before moving on to my interpretation of this ruling, I want to discuss both situations in which the court can point to evidence that refutes the defendant’s story and situations in which the court can reject this explanation though no refuting evidence exists. I analyse both situations using a Bayesian framework. 3. Rejecting stories with and without evidence Bit credit to Ranieri as well: subs all made a positive impact, and Dyer in particular just gave us another dimension.

As I noted, courts should only reject a defendant’s alternative explanation if its probability is low. Recall that on Bayes’ rule, a low posterior probability of a hypothesis can depend either on a low likelihood, P(E|H) or on a low prior probability, P(H) of the hypothesis. So, we might assume that, if the evidence does not discriminate between explanations in terms of the likelihood, an alternative explanation’s low posterior probability can only be because that explanation has a low prior probability. Is this how we should read the Supreme Court’s ruling? To put it differently, when courts reject a defendant’s explanation for being implausible, incredible or highly unlikely, is this always a judgment about that explanation’s prior probability? And what should we then make of the distinction between these three terms?So, there is little gain to justifying why we reject obviously improbable explanations. Yet spelling out such arguments does take time and effort and impedes the efficiency of decision-making. The costs of explicit justification will then outweigh the benefits. In the Venray case, a man was accused of stabbing his wife to death. At first, the man called upon his right to remain silent. He only offered an alternative explanation after one and a half years had passed. At that point, he knew the results of the forensic investigations. He then claimed that he had found his wife dead and hypothesized that criminals might have killed her because of an argument they had with him. As the court of appeal noted, this explanation fitted with the limited available evidence (blood stains and shoe prints) at least as well as the story that he killed his wife. In fact, the court remarked, the explanation may even have fitted better with the evidence, for instance because no blood was found on the defendant’s clothing, which is typically difficult to remove. However, the court did note that the defendant’s story was somewhat hard to believe, especially because the defendant waited so long to come forward with it. Nonetheless, it did acquit him. The court reasoned that it could only convict if there was evidence that refuted the alternative explanation or if it was so implausible that it needed no explicit refutation. According to the court of appeal, neither was the case.