Bayesian Learning and the Media
The economist this week had an article this week about a new paper coming out in Psychological Science that discusses how people may use Bayesian analysis to make predictions about events in the world. The estimates that people were tested on ranged from the estimated length of a congressman’s term (given how long they had been in office) to how long to bake a cake (given how long it had been in the oven), all things that people estimate given prior experience. Can the media affect this?
The key to Bayesian estimation is a body of prior knowledge; an individual experiences something over and over again and builds a body of data regarding it. We are then capable of estimating things based on what we know using a kind of Bayesian analysis. This is the same kind of analysis is that is currently popular in designing “intelligent” software and especially popular in email applications attempting to root out spam. Given “knowledge” about what words appear in spam, proximity of those keywords, and with what frequency, software can estimate the likelihood of an email being spam.
But this kind of analysis can malfunction given the wrong set of data or when two things are mistakenly thought to correlate when they don’t. Superstitions may prove to be an example of this. To me, this begs that question, can the media warp the data so that our estimates of outcomes of events are askew? Several papers have drawn the conclusion that people (using Americans in their data sets) overestimate their chances of being a victim of a crime. From one of those papers written by psychologists from The Journal of Social Issues: “These data lend empirical support to the assumption that people typically predict more victimisation than they eventually suffer.”
What is needed then is a similar study in which the respondents are first screened for media saturation. Start with people who watch CNN all the time. Ask the same questions of people who generally don’t read the news or watch television and see if there is any difference. As of yet, I have found the study that attempts this, but I am willing to speculate the result. Although the mind clearly has an ability to distinguish between fact and fiction (at least in most of us) it is not at all clear that facts in the media do not end up becoming part of our perceived personal experiences, leading us to believe things like “All politicians are crooked” or “I will be the victim of a crime someday” or even “that guy on Banapana is always right”. All right, maybe not the last one.