Is it a Crime to Belong to a Reference Class?

tags

Reference Class Problem

Notes

there was no ``specific evidence'' that Shonubi smuggled the quantity of drugs on which the sentencing was based.

NOTER_PAGE: (2 0.3027972027972028 . 0.1880252100840336)

data consisted of the quantities of heroin seized from 117 Nigerian balloon-swallowing drug smugglers arrested at JFK airport between the dates of Shonubi's first- and last-known drug-smuggling trips: September 1, 1990 and December 10, 1991.

NOTER_PAGE: (2 0.38181818181818183 . 0.551470588235294)

statistician concluded that there was a 0.99 chance that on the seven previous trips Shonubi smuggled at least 2,090.2 grams

NOTER_PAGE: (2 0.4965034965034965 . 0.4506302521008403)

general consensus amongst commentators on this case seems to be that the Second Circuit Court's decision at Shonubi IV to vacate the sentence of Shonubi III was a poor one.

NOTER_PAGE: (3 0.24405594405594405 . 0.16491596638655462)

real importance of the Shonubi case is in the lessons it holds for statistical evidence generally.

NOTER_PAGE: (3 0.32517482517482516 . 0.2647058823529412)

concern is that the rejection of the statistical evidence in this case amounts to the wholesale rejection of statistical evidence

NOTER_PAGE: (3 0.7713286713286713 . 0.2741596638655462)

we do not believe that the statistical evidence in this case was as good as statistical evidence gets,

NOTER_PAGE: (4 0.0986013986013986 . 0.6911764705882353)

the real issue is whether Shonubi should have been sentenced based on evidence gathered from other people. That is, should he have been sentenced based on his membership in a particular reference class?

NOTER_PAGE: (4 0.5944055944055944 . 0.5105042016806722)

let us suppose that 99 per cent of people from a certain reference class cheat on their taxes. Does this mean that we are justified in charging and sentencing someone in this class with tax evasion, without further evidence?

NOTER_PAGE: (5 0.26083916083916087 . 0.0819327731092437)

we require further evidence not because we wish to raise the probability from 0.99 to something higher (after all a probability of 0.99 seems a good candidate for beyond reasonable doubt). Rather, we require further evidence because the reference-class evidence is not specific to the individual in question.

NOTER_PAGE: (5 0.34265734265734266 . 0.4695378151260504)

the reference class of Nigerian balloon-swallowing heroin smugglers is seen, by some, to be intuitively relevant, whereas the class of toll collectors, for instance, is not. We need to find an argument in support of this intuition,

NOTER_PAGE: (6 0.10069930069930069 . 0.13970588235294118)

It was Shonubi's membership in the class of apprehended drug smugglers that landed him in court. The mode of transport, country of origin, and, indeed, the kind of illegal substance were not relevantÐhe would have ended up in court if he were caught smuggling cocaine from Australia in stuffed toys, say. Indeed, he would have ended up in court if he had been caught engaging in any illegal activity.

NOTER_PAGE: (6 0.24475524475524477 . 0.15966386554621848)

suggestion that the class of balloon-swallowing heroin smugglers is more robust than others. That is, taking the intersection of this class with others (such as the class of toll collectors) will not significantly change the estimate of the quantity of drugs smuggled. If correct, this amounts to the claim that the reference class in question is homogeneous with respect to the quantity of drugs smuggled. The problem is that such homogeneity has not been established by any of the statistical evidence presented, and to simply assume homogeneity is to beg the question. Whether this reference class is homogeneous, and, in particular, whether it delivers the correct answer for Shonubi, is precisely what is at issue.

NOTER_PAGE: (6 0.3993006993006993 . 0.16176470588235292)

counterfactual dependence between carrying a certain quantity of heroin and being a member of the reference class in question: had Shonubi not carried the quantity of drugs on which he was sentenced at Shonubi III, he would not have been a member of the reference class in question. The problem with these responses is that in the former variant, causal connections that have not been established are being assumed, and in the second variant, homogeneity of the reference class is again being illicitly invoked.

NOTER_PAGE: (6 0.762937062937063 . 0.23529411764705882)

we are interested in a single-case probability, not a frequency relative to some reference class.

NOTER_PAGE: (7 0.34195804195804197 . 0.18067226890756302)

According to frequentists such as von Mises (1957), we cannot even make sense of single-case probabilities,

NOTER_PAGE: (7 0.3832167832167832 . 0.18592436974789917)

frequentist view seems to be lurking in the background of the criticisms of Shonubi IV.

NOTER_PAGE: (7 0.479020979020979 . 0.4422268907563025)

hard to see how any view that did not allow single-case probabilities could be of any use in decision theory

NOTER_PAGE: (7 0.5454545454545454 . 0.5703781512605042)

We now sketch a way of making sense of the Second Circuit's request for specific evidence that does not involve the rejection of all statistical evidence.

NOTER_PAGE: (8 0.5993006993006993 . 0.29936974789915966)

no matter how you cash out the phrase ``specific evidence,'' there is an obvious candidate for such evidence in the Shonubi case: Shonubi's previous behavior.

NOTER_PAGE: (8 0.6587412587412588 . 0.2815126050420168)

III. A DECISION-THEORETIC PERSPECTIVE

NOTER_PAGE: (9 0.66993006993007 . 0.25735294117647056)

IV. CONCLUSION

NOTER_PAGE: (12 0.7153846153846154 . 0.3897058823529412)