Dealing With Information Cascades
Throughout Info 2040 I have been learning many concepts that explain large groups’ actions and how large groups deal with information. For the most part, importance or correctness can readily be evaluated by the breadth of people referencing or accepting a certain view. We have learned about Google’s page Rank system that ranks pages importance and relevance on outside links to that site. This is a way of google using the internets’ consensus to determine importance. In fact, one could make the philosophical case that correctness in many cases is merely what the majority of people think is correct.
One of the most interesting cases of using consensus to choose a most optimal strategy is that of information cascades. The first information cascade we examined was a game where many students pull colored marbles from a bag that contains either 2 red marbles and one blue (majority red) or two blue and one red (majority blue). The students then guess whether the bag is majority red or blue and each student only knows what color they pulled and what color the students guessed before them. What becomes interesting about this problem is that very quickly the students stop basing their guesses on any new information. Once two students in a row get the same color, probability tips towards that color being the majority; the more quickly this occurs the less information the rest of the students will base their guess on. If the first two students both get blue then the third student should guess blue regardless of what color they get. After this point every guess should be blue and will be based off of only these first two guesses.
The concept of information cascades raises a very interesting problem: something wrong can be deemed correct by many people if only the first few people are wrong and the rest are unable to communicate. As Dr. Welsh told John Tierney, the author of the blog post I have linked at the bottom, this issue arises because: “It all comes down to the fact that many choices are primarily binary.” In the context of the marble example this means that the third student must choose blue or red even if they aren’t sure about either. If the third student pulls a red marble they must make a decision with more certainty than they have knowledge.
The binary nature of choices is the heart of the issue if students can only communicate through their choices. However, if each student could say what they are choosing and why, the information cascade would be avoided altogether. Thus, it seems that information cascades like this occur because of a lack of communication, not information.
The last point that Dr. Welch brought up was the value of outside thinkers. He pointed out that people who go against common knowledge are often not serving their own best interest but are very valuable in society as they can uncover errors made with information cascades. Coming back to the example we used in class, if the third person decided to choose red because he pulled red, he would most likely be wrong, but every person after him could base their decision off of four data points and not just three, potentially changing the class’s consensus. In this way, conformity despite knowing better is selfish as it has the highest expected pay off for the individual, but does not benefit anyone else.