Semantic Compositions

How much is a typo worth?

A friend directed SC to this article in Business 2.0 about a man named Kevin Ham, who built a fortune on trading in Internet domain names. Mr. Ham engages in many of the usual practices employed by people who buy domain names by the ton -- ads go up on any page he's got parked until he receives an offer to buy the domain name -- but this is a gentleman who is far smarter than the typical search engine spammer. Indeed, the article opens with a retelling of an auction he attended, where the reporter put himself inside Ham's head to ask the question, "If it's a typo, is it a mistake a lot of people would make?". The fact that Kevin Ham (and his partner Robert Seeman, whose name is on the patent application to be discussed) came up with an answer to that question that nobody else did makes him a very smart man indeed.

The traditional way of making money on other people's typos has been to register domain names which are a letter or two off of sites that people want to visit. The idea is that you'll grab all the "direct navigators" who just idly type in a name and hope something sensible pops up (sort of like putting your car in drive and hoping there's a road ahead of you, and that it goes in the same direction you had in mind -- which is not to say that there's anything wrong with typing an address if you actually know it). For example, newyorktime.com takes you to a site which has nothing to do with the New York Times except that the owner has thoughtfully placed a subscription link for the newspaper near the top of the page (and the rest of it is pretty much a conventional ad farm). "Typo-squatting", as the practice is known, has attracted the hostility of legitimate business for both diluting their trademarks and damaging their reputations by hosting malware (interesting how Google is on both sides of this one). The latter article mentions some interesting research done by Microsoft to systematically trace typo-squatting, and you can read the technical report from Microsoft Research here.

What Kevin Ham and his partners have done is not to focus on typos in the names themselves, but to focus on the domain codes that end every Internet address. The trick isn't to buy up domain names, but to get the domain administrator to redirect traffic for unregistered names to a site you own. And there might not be a richer prize for this typo than the country code for Cameroon -- .cm. The story actually broke almost a year ago, but until the Business 2.0 article, nobody knew who was behind it. As the story notes, this is a nifty legal gimmick that gets around the question of violating trademarks, because you never purchase any misspellings which are arguably the property of someone else (although it should be pointed out that the property rights aren't in the misspelling itself, but in the clearing of a certain edit distance around strings like your own). It's working so well, Ham is working on adding .co (Colombia) and .et (Ethiopia).

It's hard for SC to imagine being the sort of person who mistypes an address, gets a link-farm instead of what he was expecting, and says, "Oh, gee, this looks just as good. Let's click through random ads instead of trying again." Clearly, though, enough people must do this to make it worth Mr. Ham's while, which is not to say that this is the sort of business your host would want to build if he was running a company, especially if he had training in something like medicine. We can get at least one useful piece of information out of the article, though; we're told that he pulls in about $70 million a year , that he gets about 30 million unique visitors a month, and that he has about 300,000 domain names (we'll count .cm as a typo-domain for our purposes, since it doesn't matter what came before that). While the $70M figure includes unspecified "other ventures", we'll assume they're all typo-related, which at least gives us an upper bound on the value of each typo-derived hit. That $70M/year works out to $5.83M/month, and so with 30M visitors each month, that means he's looking at about 19 cents per unique visitor. If you spend a little time playing with the Google AdWord price estimator, you'll find that will buy you first-page Google search placement for an awful lot of words -- like, say, "wedding shoe" (actually, only 18 cents to appear in the 4th-6th positions for that phrase -- such a deal!). According to the article, Kevin Ham is paying about $7/year in overhead to maintain these sites.That suggests that the average Kevin Ham "typo tax" pays off after just 35-40 hits. It's not obvious to SC quite how you would fix this problem from the advertiser's perspective, since the initial error and subsequent judgments are made by the customer, but if you're paying 19 cents for every visitor who would have happily come to your site at no cost to you if only they could type straight, that's got to be worth money to someone.

Reposted Pinkberry article

Your host did not realize that the recent Pinkberry post had been cut off during the posting process. Thank you so very much, TypePad. Alas, the original text was lost, and had to be reconstructed from memory and browser cache records of the missing links. That's been done, and you can now read the completed post. Sorry about that.

What was your first clue, lady?

Once upon a time, your host applied to a graduate school and requested that he be considered for a position teaching Spanish as part of the aid package. He then discovered that they had a particularly effective strategy for figuring out just how good his Spanish really was: one morning, he received a call which woke him up, from a lady speaking only in Spanish. Being groggy and annoyed at being woken, poor SC proceeded to flub the interview, which largely consisted of trying to elicit a conversation from him in Spanish. Oops. In retrospect, SC has mixed feelings about this as an interview strategy -- it weeds out the non-fluent quite well, but it doesn't really account for the possibility that someone might not want to be cooperative for reasons having nothing to do with their fluency, and everything to do with getting what sure sound like wrong-number calls from strangers at inconvenient times.

However, SC has lately tried to be more accommodating of people who call speaking a language other than English. In the last 24 hours, your host has received a veritable flood of phone calls looking for someone who obviously: 1) speaks primarily Spanish, and 2) is in something less than the good graces of the people inquiring about his availability. Victor Whatever-your-last-name-is, you are so dead when the people calling for you catch up -- good thing they keep getting someone who has never made your acquaintance instead.

What these events have in common is the bizarre conversation that just occurred, and which SC has thoughtfully transcribed for your amusement:

SC: Hello? [no answer] Hello?
Caller: Hola?
SC: Yes?
Caller: Hola?
SC [now impatiently]: Yes, hello, what do you want?
Caller: No habla español? (Don't you speak Spanish?)
SC [getting it now]: Sí, yo hablo español, pero no es mi lengua primera. (Yes, I speak Spanish, but it's not my first language.)
Caller [angry]: Perdona? (Excuse me?)
SC [otra vez, con enfasis]: Sí, yo hablo español, pero no es mi lengua primera. (Excuse you!)
Caller: [click]

Based on the increasingly enormous call volume in Spanish in the last 24 hours, your host is reasonably sure that the lady on the other end would have liked to speak to Victor. SC would also like to speak to Victor, and would probably do to him for free whatever these folks are being paid for. But the fascinating thing about this call isn't that it was an obvious wrong-number issue (well, a fraudulently given number, not so much a wrong one). It's that rather than trying to ask for Victor, the lady on the other end tried to figure out whether or not SC might be hiding him by determining whether or not his language use marked him as the sort of person Victor would be associating with. Caller ID tells SC that this wasn't the first time this particular person has asked SC about Victor -- at some point, shouldn't she figure out that if the person on the other end of the call doesn't sound like Victor, and doesn't behave like Victor, it probably isn't Victor?

Clicheighty

In one of the bigger hints your host will drop this month, he spent much of Wednesday at an event titled "Managing Innovation", sponsored by the San Diego Software Industry Council. It was a highly worthwhile use of his time, and he learned a lot about things that you ought to do to make your colleagues happy, productive, and most importantly, creative. Much, much more to say on these topics later.

However, the linguistic tic of the day was a cliche that perhaps doesn't quite rise to the level of snowclonehood, although it's working on it: the flagrant abuse of the number 80. If you haven't heard of the mythical "80-20 rule", it's a business cliche derived rather spuriously from an observation by the great Italian economist Vilfredo Pareto. Pareto's observation was that 80% of Italian wealth was concentrated in 20% of the population. The problem is not with Pareto's observation, which was derived from actual data, but with the cliche-driven thinking of business types who like round numbers, and who then advance spurious (and as far as SC knows, experimentally unsupported) claims like "80% of your business comes from 20% of your customers" or the computer science version, "80% of the work can be done with 20% of the effort". As a number of writers point out, the fact that the numbers add up to 100 is responsible for a lot of sloppy thinking. After all, there is no reason whatsoever that 80% of effects couldn't come from 5 or 10% of causes just as easily.

One example of this sloppiness came today in a question-and-answer session. During the panel discussion leading up to the session, a CEO of a small business mentioned that his investors wanted him to apply an "80/80 rule" to his development budget (i.e., target the top 80% of product features wanted by 80% of his potential customers). This led to a questioner asking him why he was willing to only go after 64% of his market -- clearly the result of multiplying the 80s together thoughtlessly. That's nonsense. The proper interpretation of the 80/80 rule is that you pursue 80% of your potential market, but deliver a product that customers will want to see upgrades to. The CEO got this; the questioner was in the grip of the idea that these rule-of-thumb numbers were both meaningful and additive, and could not be convinced of the error of his thinking.

That last link demonstrates just how broadly the 80/20 cliche has impacted people's thinking; a list of other applications found there includes the ideas that you should lose weight by eating until you're 80% full and eat what you should 80% of the time (and what you want 20%), that 80% of horse races are won by 20% of jockeys (this one is at least backed by some data, but not enough to convince SC of its broad applicability), 80% of the effort you put into a job search is wasted, a good job (as in employment, not as in congratulations) should be acceptable to you 80% of the time, and so on. This is largely nonsense, not backed by anything remotely like the economic data that triggered these claims to begin with.

Contrary to what you might think, though, SC isn't most bothered by the superficial similarity of these "rules" to the species of jokes that run "50% of all statistics are made up". It's more the intellectual laziness of attributing actual meaningfulness to these idly thrown-about numbers. It wouldn't be hard to write a similar post, with a little more research, about the miracle number 3, and how all successful business strategies are apparently built on 3 principles (not the same ones -- it's just that many people seem to like to put triangles on slides, label each edge with a word or two, and declare this to be insight). There's no deep meaning to the repetition of 80% or 3 in business-speak; they're just numbers that the community of managers have agreed are safe examples, and if you use them, nobody will question your judgment -- even though using them mindlessly might be the best sign that "your" judgment is something of a misnomer.

By any other name just as sour

It's time once again to write about Pinkberry, even though SC has not eaten a single thing of theirs since the last time (and has no plans to ever, ever change that). As incredible as it may be, SC wishes to write in their defense over a language issue of such sublime ridiculousness ([and it IS ridiculous -- lime isn't even a flavor there -- Ed.]) that no reasonable English speaker could fail to agree. As will shortly be obvious, the candidates for unreasonable English speakers far outnumber the reasonable ones, and hold positions of power and influence within the government and legal system.

The story this time involves a lawsuit that first came to SC's attention through an article on the website of the local ABC affiliate, headlined "Lawsuit Claims Pinkberry Isn't Really Yogurt". No, really? As far as your host is concerned, it's not even food, but that's not actually the issue in question here. Here's the crux of the matter:

Steve Lyle, CA Dept. of Food and Agriculture: "The fundamental issue we have with Pinkberry is they manufacture the product on premises."

You see, Pinkberry's super secret ingredients are prepared at each of the companies [sic] 18 stores, not at a state licensed dairy facility. And that is a "fro-yo" no-no.

This is, to put it mildly, crazy talk. SC is not saying that it is ridiculous to grant exclusive licenses to use names based on place of manufacture, but that the reasonable grounds for doing so are wholly absent in this case. One example of an exclusive name use that SC has no problem supporting is Champagne (along with wine appellations more generally); although California wine producers used to fight vigorously to be able to call their products champagne instead of sparkling wine, with a somewhat mixed history of getting results by going to the World Trade Organization, they've largely thrown in with the makers of Champagne, Port, Sherry, and Chianti to agree that they'll all get ahead by protecting their names. (This may have been motivated by a dispute a few years ago where California winemakers found themselves on the defensive against Italian producers over the name Zinfandel.) More generally, wine appellations are part of a larger class of trademark law known as "geographical indications", protected by the World Trade Organization TRIPS agreement, which covers "indications which identify a good as originating in the territory of a Member, or a region or locality in that territory, where a given quality, reputation, or other characteristic of the good is essentially attributable to its geographic origin". Along these lines, SC doesn't have much problem with people wanting to protect feta cheese, Parma ham, and basmati rice, all names referring at least in part to geographic facts about production, and not only to the products being produced.

"State licensed dairies", though, enjoy no such consideration from SC. A state licensed dairy exists wherever the state and a dairy producer decide that they do. The idea that frozen yogurt is yogurt because it was produced in a factory instead of on the premises where it is to be consumed is simply insane. However, as often happens with stories that appear too idiotic to be true, there's another angle. While the Department of Food and Agriculture spokesman quoted above makes the issue out to be one of whether or not the product is manufactured on the premises, the lawsuit in question also turns on whether or not Pinkberry's products meet some more relevant criteria of yogurthood; as Clara Peller might have asked, "where's the bacteria?"

The California Food and Agricultural Code provides a fairly detailed series of requirements for what frozen yogurt must contain to be labeled and sold as such, specified in Section 36991-36994. Specifically, frozen yogurt must: be made from milk, be fermented with one of three families of bacteria, not be pasteurized post-fermentation, have a titratable acidity of 0.5% (lower if it contains fruit), contain at least 3.5% milkfat, and weigh at least 4.0 pounds per gallon of. All of these are lab-testable requirements, and would appear to rule out Pinkberry's yogurt on the simple grounds that it doesn't cover nonfat versions of frozen yogurt, which Pinkberry most certainly claims their product to be (you can view the nutrition facts at their site, but it's Flash-based and so  they can't be linked directly). However, plenty of foods are labeled as nonfat frozen yogurt for sale in California, and the 3.5% figure corresponds to an assumption of preparation with regular market milk (see Section 35784 -- that 3.5% is simply the standardized amount of fat in whole milk) so that would be a very weak argument to hang Pinkberry on. In fact, the more sophisticated argument against Pinkberry-as-yogurt turns on the question of bacteria. According to the L.A. Times, the privately-filed lawsuit spurring this issue claims that Pinkberry's yogurt does not pass the bacterial analysis needed to qualify as yogurt.

In order for an authoritative determination of Pinkberry's bacterial contents to be made, they would have to disclose their ingredient list and publish an accompanying lab analysis. Technically, Pinkberry is in violation of the Food and Agricultural Code here, too -- Section 33704.5 requires that if yogurt is manufactured and directly served on the premises, an ingredient list must be posted in a conspicuous location. Since their ingredient list is presently secret (not a tenable position, with only limited exceptions), from a consumer perspective there's no way to make this determination. This has led to all manner of innuendoes about exactly what goes into the product, some humorous, some more sinister.

So, what ingredients can be kept secret? Not much -- for a summary of Federal labeling requirements, look at page 2 of this consumer guide (prepared by the California Department of Public Health), which notes that everything has to be listed in descending order of contribution to the product by weight. Only those items identified by the Food and Drug Administration as "harmless markers", which basically means flavors or scents which have no discernable health impact (for the exact details, see paragraph (g) of 21CFR501.100, a section of the Code of Federal Regulations dealing with labeling exceptions). If Pinkberry is hiding a proprietary flavoring compound, they can apply for an exception to not disclose it publicly, but they can't hide a lack of bacterial content by claiming a trade secret.

At this point, the fiendish genius of the "no-state-licensing" attack becomes clear. Either Pinkberry admits that they're illegally running unlicensed dairies, or they admit that they're not really producing frozen yogurt. Both of these outcomes are damaging to their business, in a way that not all their competitors suffer from, and not necessarily for particularly sensible reasons. Contrary to the assertion in this L.A. Business Journal article, it is not the case that legally valid frozen yogurt "can’t be reconstituted from powder". It's just that the reconstituted yogurt has to have been prepared in a licensed dairy, and pass the lab analyses. It took SC only five minutes of looking in order to come up with a couple of "dry mix" yogurts that specifically pass the bacterial requirement (generally the problem in these cases -- see the ingredient list in the dry mix in the last smoothie on this Jamba Juice page for specific details). The technology for getting the bacteria requirement met in dry powders has been around long enough that the relevant patent has gone public domain, as well (albeit only recently). Of course, if Pinkberry is doing it this way, they need to be paying for dairy licenses, probably paying restitution for prior violations, and putting up their ingredient list as per the legal requirements. Their competitors, who have most definitely not all been posting their own ingredient lists, should be doing the same, even if they're buying their yogurt mixes from licensed dairies operating outside their stores.

On the other hand, if it isn't yogurt by the technical definition, they shouldn't be going down alone. While we dismissed the nonfat argument as weak previously, it's by no means reasonable that Pinkberry should be the only ones to take the fall on account of a failure to meet an arbitrary technical definition. Anybody who advertises "nonfat frozen yogurt" is contradicting themselves under the terms of the law as written, and it would be entirely reasonable for Pinkberry to take the position that if their "nogurt" is going down, they're taking the industry with them. The mere fact that the law is out of date shouldn't be the entirety of the issue -- if the state is going to get involved only over the question of where the powder was mixed, considering that there is absolutely no geographic indication in the term "frozen yogurt", it is simply absurd that their competitors should be able to advertise a noncompliant product and get away with it.

In conclusion, as much as it pains SC to say so, he sides with Pinkberry. That leaves an original-flavored sour taste in his mouth.

Holy expressions of surprise!

We take a break from doing posts which hint at what's coming next to present the first-ever SC blog post written specifically for the purpose of entering a contest. Specifically, the contest can be found here, or could when it was open (having closed one minute before this entry went up), and is in regard to the Sam and Max series of video games -- quite simply the best thing to happen to PCs in the last year. The only rule of the contest is to produce something creative relating to the first season of games (season is an appropriate term here; rather than releasing one big game that took 20-30 hours to finish, Telltale Games developed 6 short ones that each took 4-5 hours). While they expect most of their fans to produce video clips, drawings, or other typical fan art, that's not the SC style, and so we now present: "Holy Dog and Rabbity-Thing!: A Syntactic, Semantic, and Pragmatic Analysis of Expressions of Surprise in Sam and Max, Season One".

In the 1960s, long before it could be terrorized by a giant robotic Abe Lincoln, the nation was introduced to expressions of surprise of the form "Holy [Noun Phrase]" via the Batman television series. Batman's sidekick Robin would utter these expressions, followed by an announcement of something terribly obvious to the viewer (technically speaking, "Holy Toledo!" preceded the series, as demonstrated by the Greater Toledo Vistors' Bureau, and was as humorous as anything Robin said, as the notion of Toledo being holy could only be a joke). In this essay, we analyze expressions of surprise across Season One of Sam and Max, in order to demonstrate how Sam's humor transcends that of the Batman series which inspired these expressions. Both in order to control for context, and because the author did not have the foresight to prepare transcripts of the entire series before this contest was announced, we will examine expressions of surprise that occur in the opening dialogue of each episode, during the sequences where Sam and Max receive their assignments through phone calls with the faceless commissioner.

Syntax: Expressions of surprise in the original Batman series always took the form "Holy [Noun Phrase]!" One website providing quotes from the series lists 356 examples (and warns that it may not be an exhaustive listing), including "Holy jelly molds," "Holy priceless collection of Etruscan snoods," and "Holy contributing to the delinquency of minors!" These expressions would often be followed by some additional sentence to state the obvious; for example, "Holy Alps!  I'd better brush up on my geography!" after realizing he doesn't know where Mount Gotham is.

In contrast, Sam's expressions of surprise do not generally include an explicit "holy" (although he does utter "holy cow" in Episode 3), and generally take the form of a complex noun phrase consisting of a name followed by a locative prepositional phrase. These phrases became progressively more complex over the duration of the series. Whereas Episode 1 featured the simple "Jiminy Christmas in a padlocked sweatbox" and "Great gouts of steaming magma on a beeline for the orphanage!" (both as part of the same phone call sequence), Episode 4 stepped up to the very complex (and highly alliterative) "Sweet suffering Saint Sebastian on the Sousaphone in a short story by Susan Sontag!". While both Sam and Robin clearly make use of a grammatical template that provides a repeatable underlying structure for their utterances, Sam's syntactic creativity clearly exceeds that of Robin.

Semantics: While Robin shamelessly attempted to curry favor with future linguists studying his work, uttering the phrase "holy semantics" in one episode, he ruined his credibility by using it to describe a conversation that had just taken place regarding an issue of phonetics. Otherwise, Robin's semantics are largely uninteresting, as his statements usually simply reference something in the immediate visual context of the show. By contrast, Sam makes repeated references to objects without clear meanings, which force the gamer to stretch his imagination. For example, in Episode 2, he states "Sweet mother of double jeopardy backstroking in butterscotch!" which is suggestive of phrases used to refer to the Virgin Mary, as well as the legal concept of double jeopardy (legal concepts being something that a freelance policeman would be at least marginally acquainted with). It is not actually clear what the mother of double jeopardy would look like, though, nor how she gave birth to double jeopardy, nor what she would be doing in a pool of butterscotch. Sam's semantics impose demands on the gamer which again make for a more rewarding experience than Robin's statements.

Pragmatics: Philosopher Paul Grice defined four maxims for communicative effectiveness, which we may summarize as "be truthful", "make your statements as informative as necessary, no more or less", "be relevant", and "be clear". Robin's statements are usually clear, at least marginally relevant, not obviously false, but never informative. Sam is willing to go to greater lengths for his humor. Episode 3's statement of "Holy cap-wearing catfish flopping a crime beat!" is not obviously true (there are not likely to be any cap-wearing catfish), informative (not related to any facts of the case), relevant (ditto) or clear (huh?). In fact, Episode 6's statement, "By the Greek goddess Selene in a chariot with dual overhead cams and "Silver Foxx" mudflaps!" references a moon goddess, which is the only time any of his statements obey even one of the Gricean maxims. While faithfulness to these rules makes for productive conversations, it makes for lousy humor, and thus we see that Sam once again outworks Robin in the pursuit of humor.

In conclusion, in every dimension of linguistic analysis we might bring to expressions of surprise, Sam is just flat-out funnier than his predecessor, Robin. Also, Max threatened to shoot me if I didn't reach that conclusion.

On the apparent common etymology of "Verizon" and "Amtrak"

Keeping up the theme of progressively dropping greater hints about the major upcoming change around here, SC would like today to direct readers' attention to three things: 1) the lexical sub-geniushood of the folks responsible for coining "Verizon", 2) the similarly sub-genius nature of the folks who brought you "Amtrak", and 3) how these two organizations conspire to afflict SC mightily.

Verizon is alleged to be a portmanteau derived from "veritas" and "horizon". Per the link (to an official corporate page), the people behind it cannot distinguish between common nouns and proper names; hence the statement in the link that "[T]he root names for Verizon are "veritas" and "horizon." This is nonsense, of course, or at least a very strained back-formation; there's no obvious connection between the company and either of those things, just as "Altria" isn't a derivative of the Latin word "altus" (except perhaps in that strange world where pseudorandom combinations of terms from astronomy and physics are inherently desirable names). The same people who tried this stunt with the Verizon coinage have more recently tried to claim that "Idearc" is: 1) a plausible word pronounced as though it stands for "Idea Arc" (and not as though a hypercorrecting New Yorker had gotten hung up on the word "idea"), and 2) the best idearc anyone could come up with for the name of a new company. Based on a 6-hour ordeal yesterday where your host's $60/month "broadband" connection entirely failed to operate, SC has finally come to appreciate that they might be telling the truth about the "horizon" part, as in "He's ready to punt his Verizon-labeled Blackberry over the horizon".

The ridiculousness of the Verizon portmanteau pales by comparison to the choice of "Amtrak", though, which is supposed to stand for "National Railroad Passenger Corporation". OK, that's not entirely fair -- it actually is a portmanteau of "American" and "track", as in "we are run by the American government" and "we like to have trains sit on a track for 4 hours at a time without moving an inch."

At this point, it should be abundantly clear that SC spent 6 hours on an Amtrak train yesterday, 4 hours of which involved total immobility due to the train breaking down, and that he was unable to do anything productive due to the wholesale failure of his Verizon-provided Internet connection. Your host apologizes for the rant, and will wrap it up by tying it in with what it is hinting at; like the genuinely brilliant mind behind Despair, Inc., says, "If we don't take care of the customer, maybe they'll stop bugging us". Amtrak and Verizon have spent a long time trying to get SC to stop bugging them.

Linguistics for fun and profit -- mostly profit

From a rather unlikely source, National Review's Jonah Goldberg, comes an excellent Bloomberg News article on the state of the art in natural language processing and artificial intelligence for financial applications. NLP and finance being longtime interests of your host, this was an article he found well worth his time.

There's a bit of the usual hand-wringing about how artificial intelligence didn't live up to its promises in the 1960s, and the resulting skepticism which followed. And SC really wishes the HAL 9000 references whenever AI is discussed would just go away, because it frustrates your host no end that people end up thinking that a murderously malfunctioning computer is an aspirational goal (although if the AI community wants to perpetuate this meme, SC can only shrug). On that note, check out Goldberg's post for a reasonably funny HAL-in-finance parody. Having said all that, there are a couple of projects of interest that you can read about in the article:

• Professor Michael Kearns' Penn-Lehman Automated Trading Platform, involving machine learning with a stock market simulator to test real-time trading strategies
• Collective Intellect, a Colorado-based startup, offers a system for filtering financial blogs to find information that might give you a trading edge
• Professor Kathy McKeown's Newsblaster (you'll need to request a username/password to play with it, or see this example) which hasn't been used for financial applications yet, but which is mentioned for ongoing research in conducting "what-if" analyses of how news events might impact markets

A common feature of these lines of research is that they're largely oriented towards the short term; projects done by Prof. Kearns' students relied on an evaluation procedure that didn't allow overnight holding of positions, never mind anything that might qualify for long-term capital gains (that's 12 months). This strikes SC as not being a wholly desirable approach to things, because while everyone and their brother wishes they could pick off events like the giant spike in Dow Jones stock from two days ago (and some options traders may have done so illegally), there's a lot more to financial analysis than simply trying to guess where the upticks will occur. Prof. McKeown's research sounds like a stab in the direction of real financial fundamental modeling, but at least as presented by the Bloomberg article, right now the emphasis seems to be on minimizing risk and maximizing returns through short-term exposure to high-probability trades. That's not a bad strategy -- the importance of proper risk management cannot possibly be overstated, and it would take a real jerk to complain about a reliably profitable approach -- but it leaves a whole lot of other investing approaches on the table.

Maybe it's not always nice to say "thank you"

The telescreen received and transmitted simultaneously. Any sound that Winston made, above the level of a very low whisper, would be picked up by it, moreover, so long as he remained within the field of vision which the metal plaque commanded, he could be seen as well as heard. There was of course no way of knowing whether you were being watched at any given moment. How often, or on what system, the Thought Police plugged in on any individual wire was guesswork. It was even conceivable that they watched everybody all the time. But at any rate they could plug in your wire whenever they wanted to. You had to live -- did live, from habit that became instinct -- in the assumption that every sound you made was overheard, and, except in darkness, every movement scrutinized. -- George Orwell, 1984, Chapter 1

A coworker sent SC a link to this article from the Daily Mail in Britain, concerning the use of "talking" cameras for public surveillance, whereby operators in a remote control room can speak to the people they are watching. As the article puts it, the cameras are for "publicly berating bad behaviour and shaming offenders into acting more responsibly".  The civil liberties issues with such a system should be reasonably obvious -- although SC hesitates to comment, not being familiar with British jurisprudence on what would certainly raise questions about "unreasonable search" in the U.S. -- but the way in which our assumptions about verbal behavior are upset are remarkable. Take this example of a voice-from-the-heavens order to someone to get off his bike:

The Mail on Sunday watched as a cyclist riding through a pedestrian area was ordered to stop.

'Would the young man on the bike please get off and walk as he is riding in a pedestrian area,' came the command.

The surprised youth stopped, and looked about. A look of horror spread across his face as he realised the voice was referring to him.

SC would certainly be jumping off his own bicycle at that moment. Shortly thereafter, he might start questioning whether or not he had gone insane, and whether anyone else could hear the voices. The absence of both a visible interlocutor and an established context for being part of a conversation without being able to see someone (i.e. phones or walkie-talkies) must be terrifying. It's not clear from the article if signs are posted anywhere to let people know that they're being monitored, or if any other sort of publicity is being employed to defuse a potentially terrifying situation; such signs are routinely the case in much less interactive situations in the U.S., such as the near-ubiquitous cameras at stoplights around Los Angeles.

There appears to be a style guide to try to make the language used very formal, perhaps in the belief that this will make people feel like the process is somehow automated and impersonal, which would be a lot less unnerving than the thought that someone is watching you -- nay, judging you -- when you neglect to cover your mouth during a sneeze:

The control room operators have been given strict guidelines about what commands they can give. Yelling 'Oi you, stop that', is not permitted.

Instead, their instructions make the following suggestions: 'Warning - you are being monitored by CCTV - Warning - you are in an alcohol-free zone, please refrain from drinking'; and Warning - your behaviour is being monitored by CCTV. It is being recorded and the police are attending.'

Mr Bonner [the program manager overseeing the cameras -- SC] said: 'We always make the requests polite, and if the offender obeys, the operator adds 'thank you'. We think that's a nice finishing touch.

That "thank you" sounds just like something out of Robocop, but SC guesses that Mr. Bonner wouldn't appreciate the humor of that thought.

It's not at all the case that we don't have monitoring systems like this in place already in the United States. Prior to moving to Los Angeles, he used to change trains at a station which consisted of little more than a concrete slab on the side of the tracks, an associated parking lot, and a couple of automated ticket machines. In spite of the absence of any obvious personnel, there would nevertheless be periodic announcements which were clearly made by live people with a view of the platform. Some of these announcements were of the type, "Sir, please step away from the tracks". One day, stopped at another station, your host found the control room where the platform was monitored by security cameras, and all became clear.

The train station camera isn't actually a good comparison, though, because there was also a security guard always present at the station, and just about all of the behavioral issues would be handled by him (which presumably nobody would disagree with). This gets us to the crux of the matter: is public misbehavior something we want stopped with perfect reliability? There isn't a day that goes by in Los Angeles when your host doesn't say, "If only a cop had seen that", but these events almost entirely occur while driving. Advocates of cameras for traffic enforcement claim that because driving is a regulated activity, this is a wish that both can and should be fulfilled reliably. Even granting that this is so, walking around in public is not yet something that requires a license, and despite the pragmatic intent of wishing on a cop (venting frustration), SC highly doubts that most people would actually want that wish fulfilled.

Perhaps with enough experience, we could cease to regard being publicly criticized by disembodied voices as strange, in much the same way that many people (not SC) have come to stop viewing people apparently talking to nobody (but actually using cell phones with wireless headsets) as mentally ill. It seems far less likely, though, that we would ever come to view those scoldings as anything but intrusive.

Let the countdown begin: 31

Today begins the countdown towards a new and exciting chapter in your host's life. Coming June 1, Semantic Compositions gets:

• A new design
• A new direction
• A new link in the sidebars

To celebrate this event, your host is planning 30 days of shameless hint-dropping and a series of posts on various topics which will eventually even make sense together. Starting tomorrow, watch this space.