Archive for June, 2012

h1

Motivational Misalignment

June 30, 2012

Interesting about causal inference and the techniques for trying to avoid biases in observational studies; I look forward to your further exposition.  Is Pearl’s book on your reading list as well?  On a related note, do you know much about surveys and how data from surveys can be used in observational studies?  There are two reasons for my query. 

The first is that I was recently reading a paper on methodologies for surveying organizations so as to avoid biases and another on how to elicit probability estimates from humans using a novel graphical interface.  It seems to me if one can avoid certain biases and acquire “soft” information, that is perhaps half the battle, but I don’t really know if there is mathematics to support this feeling.  For the second, if you follow my meandering path, you’ll see how surveys and observational studies show up: in a previous post I had said I would talk about walking around Cambridge in April and so let me do that. 

I was there for the first Collective Intelligence Conference and it was rather inspiring and much more interdisciplinary than I have been to before.  The speakers/attendees included social scientists, computer scientists, people from business schools and law schools, and others.  My own talk used methods from cognitive history to try to understand potential differences in the nature of deduction between individual intelligences and collective ones by examining particular artifacts.  IBM is also getting big into collective intelligence.  But that is not what I wanted to talk about in detail. 

Instead, I wanted to bring to your attention the talk by the legal scholar Yochai Benkler and ask you what you think.  This especially in light of the statement you had made:

Business is business and life is life; the technologies being developed for both are similar, but their objectives are fundamentally different.

Broadly speaking, he was talking about designing human systems that involve the application of power by some participants or organizers on other participants, such that power will not lead to motivation misalignment.  A legal system is one design methodology for imposing power that may avoid motivation misalignment due to forces of legitimation.

Many empirical studies show that trying to apply power by acting on material interests can undermine intrinsic motivations.  The reason motivational misalignment arises is that most people’s motivations cannot be reduced to a simple scalar-valued “economic utility” calculus.  Benkler lists four major classes of motivations:

  • Material interests
  • Moral commitments
  • Emotional Needs/Affective Responses
  • Social motivations/connections

(I know for me, the three intrinsic motivations are typically stronger drivers than material interests, and perhaps this is broadly true for many.)  Benkler then describes some types of motivational misalignment that can occur.  One is normative framing, where the introduction of explicit incentives can frame a decision as one where self-interest is the appropriate moral framework, as opposed to one where other-regarding moral commitments are appropriate; I know this can definitely be troubling to me.  The next, as he says, is:

The second is interparty negative signaling, that is to say, if the behavior is one in which in social situations we think a reasonably well-adjusted or competent person would do because it is the right thing to do and then you offer a reward or you threaten a punishment, you’re signaling to the other person that you think they are incompetent, inappropriate and then they basically get turned off and walk away and do something else.  So that’s the negative signal, you add the reward, it’s not self-signaling, it’s not that it’s no longer a useful way for me to understand myself, it’s a way for me to read from you who I think have information about whether or not I am competent at this, to say no I won’t.

Later, he goes on to say:

Once people get used to being pro-social, once there are rewards that are social rewards and social recognition, you see a shift over time with increased levels of pro-social behavior or if you have a context in which everything is interpreted and rewarded as something that’s purely about material interests you have a decline and then shifts over time.

and then argues that social norms are a great way of preserving common community and legitimating actions of power.  I think this class of ideas can be useful not only for society but also for social systems for knowledge work.

So that finally brings me back to surveys and observational studies.  In a trip to Cambridge about six months ago (one previous to the one in April) I met up with a friend of the blog, Peter Jones, and we went candlepin bowling and had a bit of discussion about Robert Putnam’s book Bowling Alone.  I finally picked up a copy and have just finished reading Section II on Trends in Civic Engagement and Social Capital.  The several chapters in that section lay out an impressive array of evidence—much of it survey-based, all of it observational—on the decline of social capital and civic life in the United States since the 1950s and 1960s.  The next parts put forth potential explanations and possible fixes.  Thus I wonder whether either causal inference or the system design concepts that Benkler introduced will come up.  I hope so!

Although spending time with the collective intelligence community was very intellectually stimulating, I’m looking forward to spending time with my “home” academic community of information theorists at ISIT, my national community on the banks of the Charles, and my family at lunch on Tuesday.  So I guess I’ll see you in Cambridge, eh?

Just a warning, I probably won’t write while I am away in India, so feel free to write as many consecutive posts as you like, but know that it is not a case of what Putnam (p. 152) writes quoting Robert Wuthnow, “[in small groups], the social contract binding members together asserts only the weakest of obligations.  Come if you have time.  Talk if you feel like it.  Respect everyone’s opinion.  Never criticize. Leave quietly if you become dissatisfied.”

Advertisements
h1

Causal Inference

June 29, 2012

How are things going Señor Vishnu Vardhan?  Sorry, I don’t know anything about Lagrangian points in orbits being related to frames in linear algebra. 

It is interesting that for some people, a problem makes sense and lights up their mind if presented as a physical problem, but for other people, if presented as just a math problem — the gravities between celestial bodies being one example.  Even if different physical or non-physical problems take the same mathematical form, the former class of people will still think through the common physical problem.  One example close to your heart is of treating all information-related problems as ones of communication.  There are many other examples like thinking of various problems as ones of electrical circuit analysis.  One that has been in my mind, but not in the strict way as would be for the former class of people, is thinking of figuring out the productivity of newly hired salespeople as being the physical problem of deblurring the cameraman image

At the end of your lifting session, you brought up randomized trials and the issue of sample selection bias, on which Jun will be presenting some work of ours at the IEEE International Conference on Service Operations and Logistics, and Informatics in Suzhou pretty soon.  When trials are not randomized and the experiments are observational studies in natural environments rather than the type of experiments conceived by Ben Franklin, the type of statistical methods that are applied are known as causal inference.  The common physical problem for causal inference is medical treatment doses and patient responses.  Inference problems in various fields can be thought of as the estimation of dose-response functions.

I want to learn about causal inference and dose-response estimation, especially as thought about by Donald Rubin et al., and hope to put up a few more posts as I learn about it.  As a starting point, here is a paraphrasis of a Wikipedia article:

Rubin’s model is based on potential outcomes and assignment mechanisms.  Every patient has different potential outcomes depending on their assignment. A person may have one income at age 40 if he attended a private college and a different income at age 40 if he attended a public college.  To measure the causal effect of going to a public college as opposed to a private one, we should look at the outcome for the same individual in both alternative futures, which obviously can’t be done.  What can be done is a a randomized experiment, in which any difference between populations can be attributed to the assignment because that is the only difference between the groups.

In a randomized experiment, the assignment mechanism is completely random. In observational data, there is a non-random assignment mechanism: in the case of college attendance, people may choose to attend a private versus a public college based on their financial situation, parents’ education, relative ranks of the schools they were admitted to, etc. If all of these factors can be balanced between the two groups of public and private college students, then in Rubin’s model, the effect of college attendance can be attributed to the college choice.

h1

Lifting the Weight of History

June 23, 2012

As you know, I was back in Ithaca the last couple days to help run the 2012 North American School of Information Theory.  It was definitely a good time.  I got to spend hours in Phillips 101, see the sundial, and hang out with some of the organizers, lecturers, and their families at places like Aladdin’s and Purity’s.  I also met up had a nice chat with the historian Ron Kline, and also took a few minutes to go into Olin library to smell the stacks.

I was also doing a bit of cleaning here: the age old problem with the weight of history.  As I was doing so, I came across the November/December 2002 issue of the magazine American Heritage.  In there is a review of the book The Myth of the Paperless Office by Sellen and Harper.  It has an interesting point about the advantages of paper in terms of prioritization.  As has become my custom, let me quote from it.

We’ve heard plenty about the advantage of computerized information handling, but we never hear about the virtues of paper.  Sellen and Harper have identified them.  Take how they describe the disposition of papers on a typical desk, perfectly capturing my own messy office and perhaps yours as well: A primary pile of papers, adjacent to the open workspace at the center of the desk and probably next to the phone, contains what they call hot files, documents to be acted on immediately.  Warm files, still active but of less urgency, occupy the periphery, teetering on a corner of the desk or stuck in a desk drawer.  Cold files, the great preponderance of documents that don’t need immediate attention, fill file drawers father away.

and perhaps my seeing this article ten years after first reading it is an example of the thing itself. I do do much of my technical reading on a computer screen, but there are indeed stacks of paper around my desk(s).  Regarding the shape of the typical computer screen (at the turn of the century), a column in the May/June 2000 issue by Frederick E. Allen, says:

Likewise the shape of your computer screen.  It came out of Thomas Edison’s laboratory, though Edison would surely be as surprised as anyone to learn that.  It was conceived by  a young laboratory assistant named William Kennedy Laurie Dickson during the invention of motion pictures…. By December 1891 he was experimenting with the latest celluloid film and settling on strips 1 3/8 inches wide and fifty feet long.  He then decided on an image one inch wide and three-quarters of an inch high—the proportions of your computer screen.  Why that ratio? “He never said why in his letters,” says Paul Spehr, a motion-picture historian who is writing a biography of Dickson.  “But he was first of all a photographer.  He had been Edison’s photographer for seven or eight years.  He had a strong sense of composition and was accustomed to photos where the bottom was always a little shorter or longer than the vertical. It was an aesthetic decision.”… (Movies changed to wide-screen in the 1950s only to be different from TV.)  And when people started building computer monitors, they naturally used TV screens, and Dickson’s format lived on yet again.

Of course, widescreen is now pretty common among computer screens due to the movies.  In general, the column that Allen wrote, Behind the Cutting Edge, is fairly interesting.  For example, the July/August 2001 issue discusses and pretty much debunks the book IBM and the Holocaust.

You’ve talked about this before, and in a related manner, John Steele Gordon in a column from July/August 1999 entitled “Lies, Damned Lies, and the Dow: Statistics help us comprehend the world—sometimes” writes :

Statistics are both ancient and astonishingly modern.  When the shepherds of Plato’s Attica counted their flocks, they were, of course, making a statistical analysis.  But the word itself entered the English language only in 1787.  It is no coincidence that that was also the decade that James Watt patented the rotary steam engine.  As the factory system, which had evolved in the middle of the eighteenth century, and the new power spread, enterprises could operate on a far larger scale.  This meant that seat-of-the-pants management became less efficient.  The owner of a business that employs hundreds or thousands of people needs statistics.  As the mathematics of statistics, and thus their power to illuminate an increasingly complex world, rapidly advanced in the nineteenth century, governments began to utilize them more and more as well.

Another interesting gem with some relevance to this blog is from the October 2004 issue, from an article “Franklin’s Forgotten Triumph: Scientific Testing”:

What is known as the blind protocol influences our lives in a thousand ways.  Its basic elements are simple to understand.  Take a group of people.  Randomly assign them to one of two groups.  One receives the real medicine; the other gets a placebo, and the researchers and the patients alike are blind to which is which…. Doing science this way is important because what researchers want or expect can influence what they observe, or how they interpret what their data says.  If no one know what he’s observing until the data collection and analysis are completed, then the potential for bias is eliminated.  That is why the blind protocol has become the gold standard of the life sciences.  But where did the idea come from?  Even scientists are surprised to learn that it was created by Benjamin Franklin.

Although randomized trials e.g. for judging human behavior is certainly a golden methodology, I think the most interesting of statistical questions actually arise when this is not possible.  What say you; I think you have some recent interest in selection bias, right?

h1

Problème des Trois Corps

June 9, 2012

Thanks for the infographic on celestial mechanics: very interesting.  Definitely saved me the effort of looking up my ephemeris tables, my planetary size table, and my history of science/technology papers.  Though I must say, when I worked at Syracuse Research Corporation one summer, I enjoyed getting TLE from NORAD and generating animations with the Satellite Tool Kit.  Interestingly, both now have new three-letter names: SRC and STK

The alignment of three bodies like Venus, Earth, and Sun on Tuesday is somewhat related to a classical problem of mechanics: the three-body problem, where the goal is to predict what will happen under gravitational forces given initial conditions.  One of the nice results that comes out of the three-body problem is the existence of so-called Lagrangian points, which are stationary points of the dynamics.  Do you happen to know if there are any frame-theoretic interpretations of the Lagrangian points?

Mechanics and mechanical engineering are old subjects, but you can still come up with new things.  I was just reading an article in the special issue of IEEE Spectrum on money, and came across an intriguing technology for transferring money from one person’s cell phone to another person’s.  Let me quote from the article:

Silicon Valley start-up Bump Technologies recently announced Bump Pay, which cleverly uses accelerometers and geolocation to let users exchange money between phones.  To transfer money to a friend who also has the Bump Pay app, you just enter the amount, then bump the two phones together.  Bump Pay’s system uses geolocation to limit the search to nearby phones.  Next, it compares the timing of accelerometer events to match up two phones that recorded an impact at the same time.  Finally, it executes the transaction via PayPal.  Compared to NFC, it’s an extremely convoluted solution for proximity sensing.

Although I would probably enjoy fist bumping transactions, I hope that near field communication (NFC) also grows a little more, especially since I have been working with Pulkit Grover and Anant Sahai on securing inductively-coupled communication.

The problem studied in the work with Pulkit and Anant is the wire-tap channel, with three bodies: transmitter, legitimate receiver, and eavesdropper, but also makes use of the underlying physics of inductively-coupled channels.  Indeed, a lot of classic multiterminal information theory problems are “three-body” problems and many of them are not completely solved.

In comments to a previous post, Gireeja had brought up the new robust optimization approach to multiterminal information theory which sacrifices exact characterization for computability.  The deterministic approach to information theory has a similar philosophical basis and somewhat harkens back to the days of Nyquist and Hartley.  Likewise with network equivalence approaches.  Thus, information theorists seem to be following, in some sense, the approach taken by physicists to hard problems: approximation.  What would be nice, however, are information-theoretic analogues (e.g. about single-letterization) to the fact it is impossible to solve the n-body problem in general only using the method of first integrals.

Now that would get a fist bump from me!  I’d add a little Bump Pay to the Ashvins Fist Bump Prize, but perhaps that isn’t particularly useful.

h1

Transit of Venus Infographic

June 2, 2012

How are you Señor Lasse Kjus?  Interesting point about hard work making us happy, but only if it is the hard work that we choose for ourselves.  I wouldn’t necessarily call it hard, but I did some work last night that absorbed me and I think played to my strengths. As you know, I’ve been dabbling a bit in visualization (with the help of the Cambridge office).  Something that is becoming very popular these days (but in a way has been popular for a long time), is the use of infographics.  I decided to try my hand at making one and the result is below.  The subject of the infographic is the transit of Venus that will occur this Tuesday just before sunset in the United States and starting at sunrise on Wednesday in Europe and Asia.

Transit of Venus Infographic