Saturday, August 11, 2007

Embedded Generalizations and Deeply-Embedded Generalizations

Some "generalizations" -- some general principles about the workings of the world (along with some sorts of rules for applying them) -- are embedded in the human brain -- or, more broadly speaking, some rules of inference (including, perhaps, "generalizations") are embedded in the human organism. When such principles are at work, we sometimes speak of "reflexive" thought and action. Consider the following example (provided by the authors of an extremely interesting book in order to make a somewhat different point):
You step off the curb. A horn blares to your left. You step back onto the curb. A good thing, too, you think, as a pick-up truck whizzes by.
John Woods, Ralph Johnson, Dov Gabbay & Hans Ohlbach, "Logic and the Practical Turn," in HANDBOOK OF THE LOGIC OF ARGUMENT AND INFERENCE 1, 10 (2002) (vol. 1 in series STUDIES IN LOGIC AND PRACTICAL REASONING, eds., D. Gabbay, J. Siekmann, J. van Benthem & J. Woods).

The "generalization" or generalizations that are at play in such a situation (the one involving the pick-up truck) probably are not exclusively a result of genes. A human being from, say, a remote settlement in a region of the world without automobiles, trucks, and horns might not react exactly the same way if placed in a city and presented with the blare of a horn from a pick-up truck.

But some "reflexive" rules of inference (and resulting actions) may be less dependent on inferential skills developed as a result (in part) of interactions with a specific kind of environment. Consider the "instinctive" ability of the human organism (eye, brain, etc.) to construct a three-dimensional image of, say, a ball when certain certain light signals pass into the human eye and meet the retina. See, e.g., Steven Pinker, "The Mind's Eye," HOW THE MIND WORKS 211-298 (1997); Daniel Chandler, Visual Perception; James Todd, The visual perception of 3D shape (March 2004); P. Tillers, Perceptual Errors. Many such inference rules are so deeply embedded, so to speak, that it is quite hard to know what those rules of inference are. (Sometimes, it is thought, such inference rules do not even exist. But this way of thinking about perception is error.) Consider: human beings often have the ability to distinguish human screams that are generated by fear and terror from human screams that are generated by joy and pleasure. Although it may be possible to figure out how human beings make this distinction -- researchers are hard at work on similar kinds of problems -- it will take much study and research to figure out the rules that human beings use to infer which kind of scream -- fear-inspired or terror-inspired, or some other kind -- they are hearing.

Are explicit generalizations in explicit inferential reasoning a superstructure plunked on top of a submerged iceberg of tacit inference? Or are explicit inference and the generalizations involved in them more autonomous? Fully autonomous? Relatively autonomous? If not always, sometimes? What is the usual relationship between tacit and explicit inference in the sorts of factual problems encountered in legal settings such as trials?

Note: today -- and was it not always so? -- almost all people are familiar with the phenomenon of visual illusions a/k/a optical illusions and other perceptual illusions, perceptual tricks and delusions.

Penrose Triangle
(public domain)

Question: what does the existence of such illusions prove? Answer: not much. The important question, and the hard question, is when visual errors and illusions occur and what if anything we can or should do about them.


Generalizations, it is said, are crucial to inferences about the workings of the world. But there is a deep mystery about how and why generalizations work (if they do work). Consider John Woods, Ralph Johnson, Dov Gabbay & Hans Ohlbach, "Logic and the Practical Turn," in HANDBOOK OF THE LOGIC OF ARGUMENT AND INFERENCE 1, 16 (2002) (vol. 1 in series STUDIES IN LOGIC AND PRACTICAL REASONING, eds., D. Gabbay, J. Siekmann, J. van Benthem & J. Woods):
How is it possible that there be a range of cases by which projections from samples are so nearly right while at the same time qualifying as travesties of what the logic of induction requires? The empirical record amply attests to a human being's capacity for pre-inductive generalization and projection.

Thursday, August 09, 2007

Dramatic Use of Statistics in a $280 Billion - $800 Billion Cigarette Case

In Schwab v. Philip Morris USA, Inc., 449 F.Supp.2d 992, 1239-1241 (E.D.N.Y. 2006), Judge Jack Weinstein wrote:
A. Aggregate Proof

Plaintiff's use of aggregate proof does not violate defendants' constitutional rights. The appropriateness of such proof has been analyzed in numerous memoranda. See, e.g., Blue Cross, Appendix B at Part VII and D, infra. Experience with several years of discovery in this and related tobacco cases, and two full trials, has strengthened the conclusion that statistical proof combined with other evidence is a necessary and pragmatic evidentiary approach that reflects full due process in this and many other massive tort cases. It is consistent with defendants' constitutional rights and legally available to support plaintiffs' RICO claims. This case, involving alleged mass market fraud on smokers whose economic injuries, if any, are calculable on an objective basis not involving individualized questions of medical harm, is even better suited than previous cases to such aggregate proof. Cf. Blue Cross, Appendix B, infra (damages sought for all increased payments made to individual insureds by health plan); Simon II, Appendix D, infra (punitive damages class sought based on estimate of compensatory damages due to personal injuries incurred from smoking).

The idea that due process and jury trial rights require a particularized traditional form of evidence for each element would make this case and cases like it impossible to try. There is little harm in retaining a requirement for "particularistic" evidence of causation and damages in sporadic individual accidents where there are but a few medical histories and witnesses; such evidence is almost always available and convenient in such litigation. See, e.g., In re "Agent Orange" Prod. Liab. Litig., 597 F. Supp. 740, 832-34 (E.D.N.Y. 1984). Even in such cases use of almost any experts, whether doctors or DNA experts, depends upon the implied or express probabilistic underpinning of their professional judgments.

In mass fraud cases with hundreds of thousands or millions of injured the cost of one-on-one procedures is insuperable and unsuitable for either a jury or a bench trial. The consequence of requiring individual proof from each smoker would be to allow a defendant which has injured millions of people and caused billions of dollars in damages to escape almost all liability. As Professor Rosenberg noted almost a score of years ago, such restrictions in the form of admissible evidence is impractical and unnecessary. "The concept of 'particularistic' evidence suggests that there exists a form of proof that can provide direct and actual knowledge of [the parties' conduct]. 'Particularistic' evidence, however, is in fact no less probabilistic than is the statistical evidence that courts purport to shun." David Rosenberg, The Causal Connection in Mass Exposure Cases: A "Public Law" Vision of the Tort System, 97 Harv. L. Rev. 851, 870 (1984) (footnotes omitted). Many commentators agree. See, e.g., Peter Tillers, Symposium: Artificial Intelligence and Judicial Proof, 22 Cardozo L. Rev. 1365 (2001) (describing tendency of evidence scholars to rely on mathematical and quantitative methods, such as probability theory, statistics, and decision theory); Louis Kaplow & Steven Shavell, Fairness Versus Welfare, 114 Harv. L. Rev. 961, 1203 n.580 (2001); Laurens Walker & John Monahan, Sampling Liability, 85 Va. L. Rev. 329 (1999) (using statistical evidence is a reliable and practical method for mass trial); Robert G. Bone, Statistical Adjudication: Rights, Justice, and Utility in a World of Process Scarcity, 46 Vand. L. Rev. 561 (1993); Jonathan J. Koehler & Daniel Shaviro, Veridical Verdicts: Increasing Verdict Accuracy Through the Use Of Overtly Probabilistic Evidence and Methods, 75 Cornell L. Rev. 247, 248 (1990) (although courts should carefully determine the validity of probabilistic evidence, "overtly probabilistic evidence is no less probative of legally material facts than other types of evidence"); Michael J. Saks & Robert F. Kidd, Human Information Processing and Adjudication: Trial By Heuristics, 15 L. & Soc'y Rev. 123, 151 (1989-1990) ("Much of the testimony that is commonly thought of as particularistic only seems so. It is far more probabilistic than we normally allow jurors (or judges) to realize."); cf. The Evolving Role of Statistical Assessments as Evidence in the Courts 78-79 (Report of the American Academy of Science Panel on Statistical Assessments as Evidence in the Courts) (Stephen E. Feinberg ed. 1989) (noting the contradiction between some courts' insistence on evidence that seems certain, and such "probabilistic" institutions as plea bargaining, in which decisions are made on the basis of "probable" outcome). But see the objections (generally rejected by academics and courts) in Laurence H. Tribe, Trial by Mathematics: Precision and Ritual in the Legal Process, 84 Harv. L. Rev. 1329 (1971). See generally Federal Judicial Center, Manual for Complex Litigation (Fourth) Section 11.493 (2004) (guidelines for use of sample surveys); David H. Kaye & David A. Freedman, Reference Guide on Statistics, in Reference Manual on Scientific Evidence (Second) 83-178 (Federal Judicial Center, 2000).

The Federal Rules of Civil Procedure and the Federal Rules of Evidence grant district judges broad authority to shape the nature and scope of admissible evidence for trial. Scientific evidence--such as sampling and statistical extrapolations--is well suited to mass tort actions. It is particularly appropriate in massive consumer fraud cases, so long as it passes the gatekeeping criteria described in the Federal Rules of Evidence and Daubert v. Merrell Dow Pharms., 509 U.S. 579, 113 S. Ct. 2786, 125 L. Ed. 2d 469 (1993) and related cases. Many states have provided special mechanisms for handling consumer fraud claims in the aggregate, recognizing that many such claims cannot be economically tried individually. When, as in the case at bar, the plaintiffs are a widely spread group complaining of injury from a common course of conduct by defendants, statistical analysis may provide a more accurate and comprehensible form of evidence than would the testimony of millions of individual smokers. See Blue Cross & Blue Shield of N.J., Inc., 133 F. Supp. 2d at 172 (explaining propriety of statistical extrapolation for entity suffering damages in aggregate); Blue Cross & Blue Shield of N.J., Inc., 36 F. Supp. 2d at 575 (E.D.N.Y. 1999) ("The aggregation of millions of alleged injuries in the instant suit can be expected to yield more accurate results with respect to the causation issue since projections based upon a large statistical base will be available, thus reducing the size of the possible error.").

Resolving mass tort disputes on a case-by-case basis may create a systematic bias against plaintiffs because, "[w]hile defendants spread the risk of adverse judgments across all test trials, each trial decides the fate of each plaintiff party on a single roll of the dice." David Rosenberg, Mass Tort Class Actions: What Defendants Have and Plaintiffs Don't, 37 Harv. J. on Legis. 393, 430 (2000); Marc Galanter, Why the "Haves" Come out Ahead: Speculations on the Limits of Legal Change, 9 L. & Soc'y Rev. 95 (1974) (importance of defendants' roles as repeat players). The defendant who successfully resolves a mass tort dispute with aggregate tools enjoys the economic benefit of a final resolution to all proceedings, not just a single case. Cf. Simon v. Philip Morris Inc., 200 F.R.D. 21, 43-46 (E.D.N.Y. 2001) (discussing public policy supporting aggregation).

Professors Laurens Walker and John Monahan of the University of Virginia Law School discuss and defend Judge Weinstein's rulingin their article Sampling Evidence at the Crossroads. The abstract of their article reads this way:
The use of sampled evidence in mass tort cases is at the crossroads. In McLaughlin v. Phillip Morris, Judge Jack B. Weinstein of the Eastern District of New York certified a nationwide class action on behalf of an estimated 50 million "light" cigarette smokers. In the face of the overwhelming cost of gathering evidence from tens of millions of plaintiff class members, he announced a dramatic trial plan to completely abandon individualized adjudication in favor of aggregate factual determinations based on evidence from statistical samples. The central issue in the interlocutory appeal now before the Second Circuit is the legality of statistical sampling. This Article strenuously defends Judge Weinstein's strategy of relying on evidence produced through proven methods of sampling. Such an aggregate strategy is not only much more cost-effective than case-by-case evidence gathering, it is also much more likely to arrive at an accurate determination of damages. By laying out a comprehensive proposal for the use of sampled evidence in litigation, the authors hope to hasten the day when legal decision makers finally accept a practice than has been a hallmark of scientific decision making for the past sixty years.

Wigmore Lifetime Achievement Awards

Professor Michael Risinger, Chair of the Evidence Section of the Association of American Law Schools, announced yesterday that Wigmore Lifetime Achievement Awards would be given to
William L. Twining
Judge Jack B. Weinstein
Very appropriate!

Tuesday, August 07, 2007

Stuart Taylor, Jr., Is One Smart Cookie ...

... but I did just see him use the following phrase in The Atlantic online:
The kind of DNA evidence that can conclusively prove innocence or guilt ...
Oh dear, oh dear, oh dear!

For an explanation, please see here and here (comments provoked by the Central Park Jogger case).

And, dear Reader, just in case you are too lazy to click one of those two links, allow me to quote myself:
Even if we assume that DNA technology and laboratory procedures have improved so much that when a DNA test shows a "match" or "exclusion" we are effectively compelled to conclude, respectively, (i) that two samples containing DNA have a common source or (ii) that two such samples do not have a common source -- even if we assume that DNA evidence and laboratory procedures have gotten this good, the probative force of DNA evidence on a question such as criminal guilt or innocence always depends on a swarm of surrounding assumptions, evidence, and facts. Bottom line: DNA by itself never establishes a proposition such as "guilty" or "innocent."
I think that the Innocence Projects on the whole do God's work. (No one except a saint always does God's work.) My sense is that almost everyone thinks that the Innocence Projects are wonderful, divine, even saintly. Is it the desire to convey this sentiment -- to pass out congratulations and encouragement and to be counted among the good guys and gals -- that leads normally-thoughtful observers to exaggerate the epistemic (inferential, evidentiary) authority of DNA?
Think for a moment. Just for a minute. Please! (Try it. You might like it.) For example, consider questions such as these:
1. Does DNA [or a fingerprint] indicate -- conclusively indicate, that is -- the mental state of the person who is the source of the DNA [or fingerprint]?

2. Does the existence of one person's DNA [or fingerprint] in a particular place rule out the possibility of some other material in that same place that has another person's DNA [or fingerprint]? If not, will both samples -- with DNA or fingerprints -- always be found? If not, what follows?

3. Can a sample with a particular person's DNA [or fingerprint] be "planted"?

4. Does a DNA sample [or a fingerprint] conclusively establish when or how the DNA [or the fingerprint] came to be where it was found?
Suppose a sample with Person A's semen is found on a window sill near the victim's body. What if anything does this discovery conclusively prove? Does it conclusively prove how the DNA or fingerprint sample got there? When it got there?
5. Is the process of DNA identification [or fingerprint] fully automated? (Answer: No.) If it were, would the possibility of error be ruled out? Do machines ever malfunction? Do they ever wear out? Are they ever manipulated? (Is it hard or impossible to make a process so "automated" that it cannot be manipulated by a human being?)

6. Can a sample bearing a person's DNA be (literally) blown by the wind? Can a sample bearing a person's DNA be carried by a dog? (Can human hairs be carried by the wind? Do dogs move about and can they carry human hairs?) Can person A's DNA by transferred to person B by a handshake between A and B? (I don't know the answer to this question. But my guess is that the answer is "yes.") Can person A's DNA find its way onto person B if person B wears unwashed clothing that A has recently worn?

7. Does the presence of man A's semen in a woman X's vagina "conclusively" prove that man B did not have intercourse or attempt to have intercourse with woman X?

8. Do I need to keep asking such questions about such appalling situations to prove my point? (Alas, apparently I do.)

What is it that leads intelligent people to exaggerate the power of DNA evidence beyond all reason and common sense? Is it an anxiousness to affirm the great service that has been done by the Innocence Projects? Perhaps. But is such noble-minded exaggeration dangerous? I think it's possible -- I think it's probable -- that the answer is "yes."More is at stake here than a persnickety concern about the precise use of language.

If one believes that something is certain (whether it's guilt or innocence), one naturally has a tendency to think that such a certainty is not worth investigating. Why bother?

And what happens if a certainty is overthrown -- if, for example, a person said to be certainly innocent is shown to be probably guilty, or vice-versa? What kinds of feelings might one then have if "one" is, for example, a trial judge, a juror, a crime victim, a legislator, or just a member of the public who has a recollection of the past?

Incidentally, are prosecutors and expert witnesses for the prosecution always the only sinners? Should we trust defense counsel and expert witnesses for the defense for the same reason that we are asked to trust Google -- because we know or believe that they're the good guys and gals and wouldn't do anything wrong or even think of doing anything wrong? (And we're sure, aren't we, that the good guys and gals who run Google now will always run Google -- or that if the good guys and gals at Google prove to be mortal, their replacements will be just as good as the originals? Hasn't every President of the United States been as good as old George [Washington, I mean]? Q.E.D.)


Well, now I had better read the rest of Taylor's article. He is, after all, one of the best journalists around. He's ordinarily first-rate. But all of us get snookered now and then. Including me. It's entirely possible I've made a mistake in this very post. (That's a bit of a paradox, I suppose. But there you have it. And I still think -- for the moment -- that what I've said here -- or most of it -- is correct. But I could be wrong in thinking that.)