The open-access journal “Semantics and Pragmatics” is up and running, and is accepting submissions for publication. The journal can be accessed here:
http://semprag.org/
I am very excited about this development, which has been long overdue.
Archive for General Science
But We Don’t Talk That Way
In the study of linguistics, some of the most profound discoveries about the nature of our linguistic capacities have been made by formulating sentences that we would, in the normal course of events, never say. Some of these are ungrammatical: “Who did John see Mary and?” Some are grammatical, but kind of weird: “What did you wear without ironing?” Others are grammatical, but it’s not clear why anyone would say such a thing: “Either the bathroom is upstairs or there is no bathroom.” It is sentences like these that have given rise to some of the best work in linguistics, and this fact has led many to conclude that there’s something seriously wrong with linguistics in general, and with linguists in particular. “But we don’t talk that way” is a common complaint. Why not try to develop theories whose aim is to handle naturally occurring sentences, such as those found in large corpora?
The complaint, though it has some force to it, is clearly off the mark as a criticism of linguistic methodology. If physicists, for instance, had set themselves the task of accounting for the world as we normally observe it, it would be a rather uninteresting endeavour through and through. Indeed, some of the best work in physics, as in other sciences, requires that we, in the words of Francis Bacon, ”twist the lion’s tail.” In other words, we need to manipulate nature to learn her secrets.
For instance, consider recent work on the “ultracold,” the physics of the very, very cold, the physics of near absolute-zero. There has been a burst of research in this domain. Wolfgang Ketterle, Director of the Center for Ultracold Atoms here at MIT, shared the Nobel Prize for having discovered a new state of matter, the Bose-Einstein Condensate (BEC), which required cooling into the nanokelvin range. The existence of this state was predicted by Bose and Einstein about eighty years ago, but no one was able to create the right conditions for it to reveal itself. Indeed, Steven Chu, professor of physics at Stanford, is reported to have said: “I am betting on Nature to hide Bose condensation from us.”
What is important about this discovery in the ultracold is that it would never have been made had physicsts satisfied themselves by merely observing nature as is. Instead, much money, time, effort, creativity, and ingenuity was spent in “twisting the lion’s tail.” The goal of scientific understanding is to discover what nature is about, whether or not she reveals herself to us in the normal course of events. This usually requires not merely passive observation, but direct intervention. One needs to play with the lion, to manipulate the lion, to coerce the lion, that holder of Truth, to reveal to us whatever window of truth we happen to be interested in. Sitting on the sidelines watching from afar generally leads to very little by way of understanding.
On Optimality and Efficiency
I recently had a conversation with a friend of mine who works in the banking industry. We talked about wage distributions, (assuming, for the sake of discussion, that the existence of wages does not amount to a form of slavery), outsourcing, “globalization” (a highly misleading term), sweat shops, and many other issues. We disagreed on just about everything. Interestingly, his arguments often rested on the notion of economic “efficiency,” a rather superficial and interest-relative term, not only in a rhetorical sense, but in a strict, mathematical sense. Let’s begin by briefly discussing the concept of “efficiency,” or “optimality,” in the better understood sciences, such as physics or biology, to get a general sense of what the term does for us in our understanding of complex systems.
We often say that some system of the natural world is “optimal” in the sense that it is given some problem to solve, and the solution it has found is the best one for solving that problem, given some measure of goodness. Take, for instance, the problem of protein folding. Given some string of amino acids, under certain conditions (temperature etc), the linear string will fold into a determinate three-dimensional configuration, what is called a “protein.” The linear string determines the 3-D shape, which then determines the protein’s function. Much depends on understanding this 3-D configuration — if we could figure out how we get from the linear amino acid sequence to the protein shape, not only would we have come to understand something deep about biology, the medical applications would be endless. This puzzle, the puzzle of how a linear amino acid sequence determines a protein’s shape, is called “the protein folding problem.”
Under current understanding, one proposal suggests that what happens is that the structure attempts to minimize energy. The problem is: find some configuration that minimizes energy, and the solution is the one that leads to the proteins we find in nature. So, we have the following pair: <Problem, Solution>. That’s the basic picture of any optimization problem — one has an optimal solution only to the extent that there is some well-defined problem to which the solution is optimal. In the protein folding example, had the problem posed by nature been different, our proteins might well have had a different look than the one they have now.
For instance, under our abstract understanding of the mathematics of computation, the solution to the protein folding problem is “NP-Complete,” which means it’s computationally intractable in the worst case. But nature seems to not care — its problem is to minimize energy, not computational complexity. Had the problem been: find some shape using a procedure that is computationally efficient (in the strict mathematical sense of using an algorithm that runs in “polynomial time”), then the solution may have been something else entirely, and our biological makeup could have been drastically different from what it is now. Thus, the solution is what it is because of the problem it’s trying to solve. There is no “efficiency” or “optimality” in and of itself — we must always ask, efficient with respect to what purpose? What problem is being solved efficienctly?
Or take the theory of foraging. It is generally thought that foragers attempt to maximize caloric intake. Consider birds straying from their nest in search of food. If they take the first food item they come across, it may be so small as to not be worth the trip out and back. If they wait for large food items, they may be so rare and time consuming to find that the risk of starvation becomes too magnified to make it worth it. The optimal solution to this problem is to find larger than average food items, but not much larger than average.
It turns out that birds do not do this. They do search for larger than average food items, but not as large as that predicted by the optimal strategy produced above. It seems that the reason for this is that the problem we formulated above was the wrong problem — instead, what they’re optimizing is a balance between maximizing caloric intake and not being away from their nest for too long a time (so as to protect their nestlings). The birds’ foraging strategy is the optimal solution to something like the following problem: maximize caloric intake while minimizing time away from the nest. (See, for instance, Richard Lewontin’s 2000 “The Triple Helix: Gene, Organism, and Environment,” Harvard University Press).
In general, optimization problems abound in nature. The key is often in determing the right problem, and it’s a safe bet that nature has found the best solution. Take even simple problems such as getting from A to B, any two points you like. In physics, of course, shortest path principles are robust. But take human action. If you inspect your own actions, you will find that you usually find something like the shortest path between A and B (out of all the paths known to you). Of course, you may sometimes take “the scenic route,” but that’s because something forces you to (such as wanting to take the scenic route). In the absence of any other motivating factors, you will find something like the shortest path to get to where you want to go. When you’re at an intersection, and want to cross the street, you never walk up one block, cross the street, and walk back down. Instead, you just cross the intersection in a more or less straight line. Or if you want to get from your bedroom to the kitchen, you never walk along the walls of each room you pass as you travel to your destination. Instead, you walk in what amounts to the shortest possible route from the point in your bedroom to the desired point in your kitchen. The problem is: get from A to B minimizing some cost function (energy, time, or whatever), and the path you take is more often than not the optimal (or near-optimal) one.
Now, in economic theory proper, we find all kinds of “optimality” solutions for problems of all sorts. Take the concept of “Pareto-Optimality,” or “Pareto-Efficiency,” for instance. An allocation of resources is said to be Pareto-Optimal (Pareto-Efficient) if there is no other allocation that makes anyone better off without making someone else worse off. The conclusion is that resource allocations should strive to be Pareto-Efficient. Now, the problem is, of course, that one can have a Pareto-Optimal resource allocation while having an outcome that offends our sensibilities. For instance, imagine a group of three people dividing $100 they find on the ground. Two of them get $45 each, while the third gets $10. Well, such an allocation is Pareto-Optimal, because the only way to improve anyone’s allocation is to take some money away from someone else. But that hardly counts as an “optimal” solution in some other sense, say, in the interests of fairness, or in terms of proper rewards, or other measures that need not be tied to value or justice.
Now, the crucial difference between the problem of allocating resources and protein folding is that we are not in a position to *do* anything about the latter. Nature has designed a set of optimization problems, and proteins, subject to the laws of physics, must obey them. And we have no say in the matter. In matters of social design, on the other hand, if “democracy” is to have any meaning, then we do have a say in the matter. The problem we formulate need not be one of maximizing shareholder profit only, while ignoring all other factors that motivate human action. This does not mean, by the way, that government intervention is needed to determine resource allocations. One can have market forces in a more or less decentralized fashion operating to determine resource allocations, but the problems that we’re trying to solve can be different (eg. minimize disparity in the distribution of resources, or maximize the capabilities of all agents to participate in the market, or maximize the freedom of individuals to live healthy, pollution-free lives, or maximize people’s potential to create and produce, or minimize children’s exposure to advertising, or minimize waste and destruction of natural resources, or minimize the existence of “prisoners’ dilemma” type games, and many others). It is up to us to decide the problems that are worth seeking solutions to.
Thus, appeal to the concept of “efficiency,” in economics as in the hard sciences, is always relative to some optimization problem. As such, any argument in the social sciences that rests on “efficiency” must justify the problem that the efficiency is an answer to. “Efficiency” in and of itself is vacuous, and since nature is not supplying all our social optimization problems, any proposed problem must be justified as a reasonable one for us to seek solutions to. And one finds that a system whose sole purpose is to maximize the capitalist’s capacity to exploit, both nature and other people, hardly qualifies as a foundation on which to rest the current socio-economic order.
Clay Public Lecture by Mike Sipser
Tonight, Mike Sipser, head of the mathematics department at M.I.T., will deliver this year’s Clay Public Lecture at Harvard University. The talk is titled “Beyond Computation,” and will focus mostly on the P vs NP problem in theoretical computer science.
I am very excited about this talk. This semester, several of us from the department are following Sipser’s graduate class “Theory of Computation,” cross-listed in Course 18 (Mathematics) and Course 6 (CSAIL). His lectures are a work of art — the experience of watching the master explain his craft is truly a gift to be cherished. Tonight’s performance should be no different.
The lecture takes place in the Science Center, Hall B, at 7pm.
Skeleton not a Hobbit, but a Human
A study published in yesterday’s Proceedings of the National Academy of Sciences debunks an earlier claim that a skeleton found in 2004 on the Indonesian island of Flores near the Ling Bua Cave belonged to an undiscovered species of hobbits, or Home floresiensis. Instead, researches claim that LB1, the name of the skeleton, is actually a human with abnormal development.
LB1 is about one metre tall, and has a smaller brain size than most humans. The study attributes its short height to a condition called microcephaly, which leads to smaller brain and head sizes. Since no traits unique to the skeleton were found, LB1 could not be claimed to belong to a new species. For now, it seems that hobbits remain a product of Tolkien’s brilliant imagination, and not natural history.
Truthful Speech
April 8, 2007 at 3:08 pm · Filed under General Science, Linguistics, Socio-Political Commentary, Uncategorized
Why should people speak the truth? Why should they lie? Do people speak the truth? Is there such a thing as truth? The last question is rather pointless, and we’ll leave it to the post-modernists to grapple with. People do speak the truth, and they also lie. Do we have any understanding of what the conditions are under which they do one as opposed to the other?
Much of modern pragmatic theory, the theory of language use, begins with the assumption that people speak the truth. This is stipulated as a primitive of our communicative machinery, and it is on the basis of this assumption that algorithms are constructed for computing inferences based on speech acts (such as the inference that if I say “Mary ate some of the pie,” you conclude that Mary didn’t eat all of the pie).
On the other hand, work in game theory, in particular on various kinds of “signalling games,” asks a more fundamental question: Given agents with various goals who find themselves in the strange predicament of having to communicate, under what conditions will honest signalling arise? The answer on offer seems to be that honest signalling will arise to the extent that the agents’ interests are aligned. Thus, when applied to certain kinds of games of information exchange that humans play, such as when we’re really trying to exchange information with our interlocutor and get them to come to believe the meaning encoded in our sentence, truthful speech is what we should expect. It is no wonder, then, that linguists and philosophers have been inclined to posit truthful speech as a primitive, for the kinds of language games that have tended to interest them most have been coordination games of information exchange (such as when you ask me how much of the pie Mary ate, and I tell you that she ate some of the pie, intending to get you to believe that Mary ate some but not all of the pie).
However, it seems that the theory doesn’t exactly apply to humans. Experimental results suggest that humans tell the truth more than they “should,” where by “should” I mean “should according to game theory.” In other words, even when it’s not in the agents’ best interest to speak truthfully, they will, as if driven by some deeply rooted impulse to tell the truth. But people do of course lie. However, as it stands, the conditions under which truthful speech will or will not arise are not well-understood. What is clear is that human communication is not guided solely by any narrow construal of “self-interested” utility maximization. But that is to be expected, since human motivations are diverse and plentiful.
However, what is true of humans need not be true of inanimate agents, such as firms or other organizations. Consider, for instance, the corporate media, such as the New York Times. Can we understand when they will or will not speak the truth? Since the motivations of such firms are simple, guided only by the need to maximize profit, the game theoretic apparatus can be more readily applied to them. Since most of the New York Times’ profits are driven by advertising, the game-theoretic results suggest that when the needs of the advertisers clash with the need to speak the truth, truth will find itself going the way of the dodo. And this is indeed what we find, to a large extent. Thus, we can expect a self-generated, self-enforcing, decentralized propaganda system to emerge solely from the fact that large media firms are driven by the profit motive, with profit generated from advertising. As such, the game theoretic results support Ed Herman and Noam Chomsky’s “propaganda model,” which suggests that propaganda in the Western media emerges out of market forces, and not through some centralized decision maker imposing informational restrictions on what the public can and cannot see.
George Orwell had this idea well over sixty years ago. He once wrote: “Is the English press honest or dishonest? At normal times it is deeply dishonest. All the papers that matter live off their advertisements, and the advertisers exercise an indirect censorship over news.”
There is of course a lesson to be learned here, namely, if we, the public, want truthful, reliable information sources to inform us about the goings-on in this nasty world of ours, these sources should be stripped as much as possible from motives that interfere with the emergence of truth. Another way of saying the same thing is that there should be an incentive for an information source to be truthful, in that the extent of its utility is derived from the extent of its truthfulness. A media system designed to sell human brains to advertisers is destined to become a liar from the beginning. Its fate is sealed from the get-go, a rather unfortunate thing, for no one wants to be a liar. The liar draws the venom of all for having abused us in such a vile way. Surely, media firms deserve a better fate than that, as does the public, whose thirst for truthful, reliable information remains unquenched.
Permalink Leave a Comment