Noam Chomsky on AI, Bayesianism, and Big Data

Since I don't share Noam Comsky's political views, I wish he would stick to topics such as language and logic. There is a wonderful article in the Atlantic that purportedly deals with Chomsky's views of (the limitations of) "artificial intelligence" a/k/a computational intelligence. The article includes a perceptive introduction by the author (Yarden Katz) and a transcript of an interview with Chomsky. (A link to a video of the Chomsky interview is also provided.)

In the course of the interview Chomsky makes a variety of points about a great variety of matters. One of the matters he discusses is associationist psychology, Bayesian techniques, and statistical approaches -- "big data" -- to an understanding of natural and social phenomena. He says, for example:

"Suppose you want to predict tomorrow's weather. One way to do it is okay I'll get my statistical priors, if you like, there's a high probability that tomorrow's weather here will be the same as it was yesterday in Cleveland, so I'll stick that in, and where the sun is will have some effect, so I'll stick that in, and you get a bunch of assumptions like that, you run the experiment, you look at it over and over again, you correct it by Bayesian methods, you get better priors. You get a pretty good approximation of what tomorrow's weather is going to be. That's not what meteorologists do -- they want to understand how it's working. And these are just two different concepts of what success means, of what achievement is. In my own field, language fields, it's all over the place. Like computational cognitive science applied to language, the concept of success that's used is virtually always this. So if you get more and more data, and better and better statistics, you can get a better and better approximation to some immense corpus of text, like everything in The Wall Street Journal archives -- but you learn nothing about the language.

"A very different approach, which I think is the right approach, is to try to see if you can understand what the fundamental principles are that deal with the core properties, and recognize that in the actual usage, there's going to be a thousand other variables intervening -- kind of like what's happening outside the window, and you'll sort of tack those on later on if you want better approximations, that's a different approach. These are just two different concepts of science. The second one is what science has been since Galileo, that's modern science. The approximating unanalyzed data kind is sort of a new approach, not totally, there's things like it in the past. It's basically a new approach that has been accelerated by the existence of massive memories, very rapid processing, which enables you to do things like this that you couldn't have done by hand. But I think, myself, that it is leading subjects like computational cognitive science into a direction of maybe some practical applicability..."

Comment by Tillers: I have always liked what I take to be Chomsky's neo-Kantian and neo-Platonic approach to an understanding of "human behavior" such as language. In a broad sense, he takes the position that some sort of a logic dwells within the human animal (for example?) that generates, or causes, or explains, what the human animal does (with, for example, language). 

• If you share Chomsky's general theoretical orientation, I think it follows that Bayesian accounts of factual inference about human behavior must be supplemented by "nomological structures" (usually called "generalizations"), which, in the case of the human animal, must describe -- or attempt to describe -- the internal (cognitive) "operating system," or mental world (both tacit and explicit) that the human animal in question uses. Cf. Peter Tillers, P. Tillers, "Are There Universal Principles or Forms of Evidential Inference? Of Inference Networks and Onto-Epistemology," in William Twining, Philip Dawid & Dimitra Vasilaki, eds., Evidence, Inference and Enquiry (Oxford & British Academy, 2011) (SSRN prepublication version of the paper is available here; you can also access the paper here).

&&& The dynamic evidence page

Evidence marshaling software MarshalPlan

Child Abuse, the BBC, the New York Times, and Mark Thompson

The Mark Thompson case -- Mark Thompson is, or was, the "incoming" CEO of the New York Times and before that was head of the BBC -- the Mark Thompson case will test the New York Times' professed commitment to principle -- its professed abhorrence of the sexual abuse of children. Mark Thompson's defense that he had "no idea" what Jimmy Savile (and presumably others) had done to children while at the BBC is almost laughable: if Mark Thoompson did not know, it's because he did not want to know. See, e.g., Andrew O'Hagan, Light Entertainment: Our Paedophile Culture  London Review of Books (Nov. 8 [sic], 2012) (recounting widespread sexual abuse of minors by prominent personalities within the BBC but laying the blame at the feet of "culture").

See also Michael Moynihan, Will Mark Thompson Survive as New York Times CEO amid Jimmy Savile Scandal? Daily Beast (Nov. 2, 2012):

"A rough précis of the increasingly complex scandal: BBC presenter Jimmy Savile, who died last year at age 84, is alleged to have sexually assaulted hundreds of woman and underage girls, which some have claimed was an open secret at Broadcasting House. Soon after his death, the BBC’s program Newsnight was set to air an investigation detailing horrifying allegations of rape and pedophilia against Savile. But  for reasons that are still unclear, higher-ups intervened and the program was dropped. Who intervened—and why—is a matter of furious debate.

"Thompson, then the top man at the BBC, claims to have known nothing of the Newsnight investigation and, therefore, was in not involved in spiking the Savile segment. But Thompson’s line on what he knew and when has been modified, shifted, and qualified since the story broke. (See The Daily Beast’s previous coverage of Thompson’s seemingly contradictory account of the Savile timeline here.) But new allegations in London’s Sunday Times suggest Thompson’s office was twice contacted about the Savile investigation by a freelance journalist last April. Thompson claims this information languished on his secretary’s desk, never reaching his ears, a defense that has provoked renewed scrutiny of his role at the Times. As New York magazine writer Joe Hagan put it, if conclusive evidence surfaces that 'Thompson had prior knowledge of the BBC report on Savile, the Times’ reputation would be damaged, not least because of its own tough reportage on molestation cover-ups inside the Catholic church and at Penn State.'”

....

....

  

The New York Times was not nearly so solicitous of what it took to be the defense of wilful ignorance in the case of Roman Catholic bishops and cardinals such as Bernard Law. (This is different because, after all, Thompson is not a Roman Catholic cleric?)

Is it time for someone to investigate whether the New York Times tolerated criminal sexual misbehavior by people within its own ranks? Or does the Times think that a journalist's privilege bars such an investigation?

  &&& The dynamic evidence page

Evidence marshaling software MarshalPlan

The "Child Sex Abuse Scandal" Now Extends to the BBC and the New York Times

I once thought that after the "[Catholic] clergy sex abuse scandal" had run its course (and tort lawyers had largely emptied the pockets of the U.S. Roman Catholic Church that they could reach into) the education establishment would be next; I intimated that a "teachers sex abuse scandal" would be next on the agenda. When the Penn State-Sandusky sex abuse scandal erupted, my prediction initially appeared to have been vindicated. But now we have what might be called the "BBC-NYTimes sex abuse scandal." See Jack Shafer  The New York Times, the BBC and the Savile Sex Scandal (October 25, 2012). So perhaps we we will now witness a "mass media sex abuse scandal" -- involving organizations such as the New York Times or the Boston Globe. Or perhaps there will be a series of "sex abuse scandals." In any event, the Catholic clergy look less and less like unique sexual "predators." This is not surprising. Child sexual abuse permeates society. As I have repeatedly suggested, for the media and tort lawyers to single out Catholic clergy as targets reeks of religious prejudice and financial opportunism.


&&&

The dynamic evidence page

Evidence marshaling software MarshalPlan

A Hypothetical Question of pro se Defendant = Not Evidence

From State v. Selecky, 2012 WL 75088 (Jan. 11, 2012) (unpublished opinion):
Lorraine M. Selecky "drove with her thirteen-year-old daughter to a 7–Eleven store in Roselle Park to rent a video from an automated Redbox video vending machine located outside the store. As they approached the Redbox, defendant and the daughter were engaged in an intense argument. Once there, they encountered off-duty police officer James Cantrell, who was renting a video with his children. At this point, a heated argument occurred between defendant and Cantrell, either because defendant thought Cantrell's children were taking too long in making their choice or because Cantrell interjected himself in defendant's mother-daughter dispute. In any case, the exchange escalated to the point that defendant threatened to call the police, but she did not do so, nor did Cantrell disclose that he was a police officer. Two days later, defendant received in the mail a summons, issued by Cantrell, for parking in the handicapped parking spot located next to the Redbox vending machine.

"[Lorraine M. Selecky], appearing pro se, contested the ticket in municipal court. At the trial held in the matter, Cantrell testified that he had observed defendant parking illegally and that he notified dispatch of the violation. However, he did not personally raise the subject of defendant's allegedly illegal conduct with her at the time."

"On cross-examination, [Ms. Selecky] asked Cantrell why, if she had parked in a handicapped spot, Cantrell did not ask her to move her vehicle. Her question was as follows:

Officer, since you took an oath that it is your honor to[,] you know [,] protect the public and dispatch anything that you see [as] wrongdoing, why would you not have approached me at that time if you in fact say this is what I did, parked in the handicapped spot right next to this red box, having seen I made this violation, why would you not interject and say to me, Ma‘am, why are you parking there, you don't have the handicapped sticker?

"When directed to rephrase her question, defendant asked:

Let's say I'm parked in the handicap spot such as Officer Cantrell is saying here that I'm parked there, and you see me coming out of the car and I approach the red box, why would you not stop me then and say, Ma‘am, you shouldn't park there, move your car. Why would you not do that if that were the case as you're suggesting?

"Cantrell responded that defendant was deemed to know the law, so there was no need on his part to inform her of her violation. In the circumstances presented, he could nonetheless do so, or he could issue a summons. He chose the latter course."

In a bench trial in a municipal court, Ms. Lorraine M. Selecky was convicted. "In reaching his decision on the matter, the municipal judge observed that, in this case, he was required to determine that “somebody's facts are right and somebody's facts are wrong.” He then held: 'Based upon the inferences that are to be given to the State and to the police, I find Officer Cantrell's testimony to be credible.' The judge continued by stating:

The one thing that I did find interesting to your [defendant's] detriment is when you asked the officer, I guess it was a hypothetical, but you asked him, well even if a person had parked in that spot, would you not interject with the person and ask him to simply pull—basically pull out of that spot rather than issue them a summons.

My—it made me think a little bit what would trigger that question and it made me think possibly, possibly that you were in that spot and that you were hurt or offended by the fact that a simple courtesy would have been to say hey, why don't you back out of the handicapped spot.

I don't know if that happened or not. But it made me have a thought process of why you would ask that particular question.

"Following some additional musings on the import of defendant's question, the judge determined to accept Cantrell's testimony regarding the episode, and he ruled that the State had met its burden of proving a statutory violation.

"Defendant retained counsel and appealed to the Law Division. On appeal, counsel argued that no deference should be given to the municipal judge's credibility finding, which was based upon his erroneous conclusion that deference had to be accorded to the State and its police witness.
"...
"Following issuance of an order stating that defendant's appeal was “denied,” defense counsel moved for an order granting defendant a new trial or permitting defendant to supplement the record with the testimony of her daughter. In a written opinion denying counsel's motion, the judge addressed an argument by counsel, set forth in a certification accompanying the motion, that neither judge had made a credibility finding with respect to defendant. The judge stated:

On the merits of defendant's claim, I, along with [the municipal court judge] did, in fact, take into account her testimony. Both courts took note of the surprising hypothetical that defendant posed when she asked Officer Cantrell why, in lieu of issuing a ticket, he would not have just asked an individual parked in a handicapped spot if they would simply move from the spot. [The municipal court judge] found the hypothetical suspect, as if she 'were hurt or offended by the fact that a simple courtesy would have been to say hey, why don't you back out of the handicapped spot.' I adopted [the judge's] concerns.

... I properly took the testimony of both the defendant and Officer Cantrell into account in coming to the determination that the State properly met its burden in proving that defendant illegally parked the vehicle in a handicapped spot."

Defendant appealed to the Appellate Division of the New Jersey Superior Court. The Appellate Division reversed Ms. Selecky's conviction for parking in a hnadicapped parking spot. The Appellate Division explained:

"[W]e are particularly concerned by statements by the judge in his opinion denying defendant's motion for a new trial or supplementation of the record, indicating that in determining credibility he, like the municipal judge, considered defendant's hypothetical questions as an admission that she, in fact, had parked illegally. We know of no precedent that would permit a hypothetical question posed by a pro se party during cross-examination of a witness at trial to be considered as substantive evidence in the proceeding. Indeed, for it to be so construed would present an unlawful trap to a defendant, untutored in the law, representing his or herself in a legal proceeding. Moreover, evidence has been defined as:

Any knowable fact or group of facts, not a legal or a logical principle, considered with a view to its being offered before a legal tribunal for the purpose of producing a persuasion, positive or negative, on the part of the tribunal, as to the truth of a proposition, not of law or of logic, on which the determination of the tribunal is to be asked.

[1 Wigmore on Evidence § 1 at 8 (Tillers Revision 1983).]

"Defendant's hypothetical question cannot be considered a 'fact' as that concept has been defined. Moreover, we do not regard defendant's hypothetical question either 'suspect' or 'surprising .' Rather, it constituted an entirely logical inquiry designed to test the officer's credibility.

"We thus conclude that, because a basis for the judge's credibility determination was plainly improper, the determination thus reached was sufficiently lacking in foundation as to warrant judicial intervention and correction. State v. Johnson, 43 N.J. 146, 162 (1964). Retrial is therefore required.

"Given the well-recognized importance of a municipal judge's evaluation of witness credibility, and out of concern that the evidentiary mistakes occurring in that regard in the initial municipal trial in this case would infect future proceedings if held before the same municipal judge, we direct that a different judge be assigned to the retrial of this matter."

....

Of course, probably the real reason for Ms. Selecky's victory was that the Appellate Division felt that Officer Cantrell abused his authority by punishing Ms. Selecky for getting into an argument with him in the Redbox store. This is, in any event, a completely reasonable conjecture. And Officer Cantrell will think twice before doing again what he did to Ms. Selecky. And Ms. Selecky had reason to be gratified with the outcome in the Appellate Division. We should all be grateful to have citizens who  have as much grit as Ms. Selecky did.

&&&
The dynamic evidence page

Evidence marshaling software MarshalPlan

Evidence and "The Shift Toward Law School Specialization"

Victor Fleischer, The Shift Toward Law School Specialization (blog post), Dealbook (Oct. 26, 2012):

In the traditional model of legal education, schools offer a general professional degree in law. No majors or concentrations. Schools provide a strong foundation of legal analysis and grounding in the common law, on the assumption that law firms will teach new associates the specifics of what they need to practice law, whether that means drafting deal documents or taking a deposition.

In the emerging model, law students must add on a degree, certificate or other indication of readiness to engage in a particular practice area or industry. N.Y.U.’s strategy committee described this goal as providing “professional pathways that prepare students to operate in a world that demands increasing specialization.” (Full disclosure: I was a visiting professor of law at N.Y.U. in 2010.)

Comment by Tillers: But if the new N.Y.U. program does not attempt to teach evidential analysis (statistical analysis is a subset of evidential analysis), N.Y.U.'s new program may turn out to be a retrogade step. Evidence mediates between legal principles and real-world conditions.

&&& The dynamic evidence page

Evidence marshaling software MarshalPlan