Game Theory
November 19, 2000
7:30 a.m.
I'm reading a book titled "Prisoner's Dilemma" by William Poundstone. It's a biography of John von Neumann and an attempt to describe game theory--and specifically game theory's tie to military behavior. I didn't know much about game theory before reading it, and I'm only halfway through. However, game theory is essentially an attempt to look at strategies employed in what are loosely called "games" (essentially a game is an interaction between two or more entities wherein each side has to choose an approach--or strategy as defined by specific actions--in order to achieve a goal). In it, a person or entity is described as "rational" or "irrational" based on whether they are actively trying to do what is best for them based on their action's payoffs (and those payoff's worth--or utility--as ranked by the player or entity), and an assumed strategy of the other participants in the game. In other words, a person is playing rationally when they take an action that truly serves their best interest--which may or may not be a cooperative approach. When they do something that will truly hurt their own self interest (like voting for Perot in 92 and Nadar in 00), then they are acting in a fashion that the theory describes as irrational. (Yeah, I know, I know. Nadarites are willing to sacrifice all for the thirty year plan). Oddly, or not so it turns out, people don't behave in a "rational" fashion all the time. In fact, most dilemma's posed in the cases I've read has people acting "irrationally" upward of 30% of the time--for a myriad of reasons. At this point in the discussion, game theory gets more complicated than I want to get into here. If you want to understand its details, go read the book. But this is, I hope, a strong enough foundation to lay for the point of my thoughts today. A few days back, Linda Dunn used the sidebar to comment on the study I did on one of my latest stories. To paraphrase, her point is that "You can't trust surveys because I don't answer them correctly." It's an interesting situation--one that professional poll-takers and surveying groups deal with on a constant basis. She also comments that surveys are taken out of context by those that read them. This is the position that says "People will see whatever they want to see in the data." It is correct. There are many groups who run a poll or survey in order to justify a belief that they already hold. I'm always leery of a person at work who says that they ran a customer survey and it validated what they already knew. There is no doubt that the position the observer takes is important in the process. So it all boils down to what the respondee will do vs. what the observer do. Both are important to the quality of the process (where quality is described a "true" accuracy). The respondee can choose a truthful strategy because they are actively trying to help (or just because this is the "right" thing to do), or they can choose a deceitful strategy because they are purposefully trying to jury rig the system (or maybe because they don't trust the conditions given in the adminstration of the survey or because ... well ... just because). The observer can do their best to read the information without personal bias in order to understand what is truth, or the observer can take the information and use it for their own nefarious purposes (thereby making the respondees earn what game theorists call the sucker payoff). If the respondee trusts the observer to work for the common good, they cooperate and respond truthfully. If they, for whatever reason, don't trust the observer, they defect and respond in deceitful fashion. If the observer trusts the respondees, he takes their information as "true." Otherwise, he interprets. Either way, if he has pure intentions, the observer uses the information in the fashion that was advertised. If the respondees have defected, then he gets the sucker reward, and works with false data as if it is real. Only with dual strategies of cooperation does everyone get a "best deal possible" type of payoff. It's a classic game theory situation. I cannot tell you specifically what strategy people used to fill out their surveys. But I can tell you that I've assumed they have chosen a strategy that says they trust what I was trying to do, and responded with "true" answers. I think this is a fair assessment, in that it was a self selecting group. In general, I don't believe too many people would go out of their way to spend a couple hours to derail a struggling writer at the draft stage. I don't see anything in their best interests to purposefully deceive. If most of the respondees selected a "I'll defect and be deceitful" strategy, then I'll admit I'm hosed in looking at the result. But I trust the folks that responded. I don't think that's a problem. But there's another problem that affects the quality of surveys. Somewhere in there, she complains about a question I asked that requests a person map characters to existing people/characters, specifically stating that this is not something she does while reading. She would have picked a family member, but didn't respond because she didn't think that was relevant. In other words, she responded to the question by trying first to figure out what I wanted to do with the information...in other words, my presence as the observer kept her from answering at all. The way a question is asked drives the responses. Can't disagree with that. I'll note that, though Linda's discussion comments on the fact that people did associate individuals with movie characters in their responses to my question, I also received answers that equated people with personal friends and relatives. So, some people jumped over the barrier that kept Linda from responding--not saying Linda did anything wrong here at all. She was trying to answer as best as she could, and her response at the survey level was something like, I'm skipping this question because I don't know how to answer. It was a great answer, because I knew that her answer wasn't that the characters didn't grab her or weren't real people, but that she honestly didn't understand the question in the same fashion as I had stated it. As I said in my study, I was only really looking for whether someone would get an image at all. So for purposes of the analysis, I basically ignored her response in this area, because it was essentially an abstention. Easy. No biggie. I think the survey worked there because over the collection of people that responded, I received enough replys to give me an idea of which characters were "working" and which may have needed a little effort. So, anyway...as I said, the strategy I took was generally to assume helpful intent. My sample size was small enough that I could look at every response and judge whether I thought it was done within the "structure" of the questions I asked. And I did my best. If people purposely provided false information, then the raw data I received was "untrue" and the analysis I performed in judging the validity of their positions/interpretations and deciding how to go about any next steps was flawed. If people responded in a cooperative fashion, with intent to help (i.e., they gave their most honest answers both good and bad), then the data was "true" and the analysis I performed--which is of course another imperfect process--was at least working with valid data. It's really as simple as that.   Have a great day.  Daily Persistence is © Ron Collins |
|
Nearing the end of "Ghandi" -- draft 2 |
BACK TO 
|