How Inside Information Effects Information Cascades
https://hbr.org/2015/04/having-inside-information-leads-to-worse-decisions
When we discussed information cascades in class, we mostly focused on how people make decisions based on only the evidence that they observe. In the urn examples, each person observed the marble that they pulled out from the urn as well as the responses of previous people who had picked out of the urn. Based on the evidence they had, they would either announce their guess as to whether the urn was majority blue or majority red. One of the problems on one of our problem sets involved the current person making a decision being explicitly told the color of someone else’s marble. We assumed that the person currently making the decision weighted all information equally and only went with their own information based on ties.
The article above analyses some evidence that people may weight inside information higher when making decisions, often to their detriment. It turns out the fact that the information is private makes our brains biased towards weighing it more. In the example in the previous paragraph, this would translate into the person currently making the decision to treat the secret information they got as more important. This could mean that even if everyone has said that the urn is majority blue, if someone explicitly said that they saw a red marble, it could change the probabilities in favor of majority red for the person currently making the decision. The article also gives an example where if an employee gets forewarned about corporate restructuring could start looking for a new job, even if there were already good indicators that they wouldn’t get fired. In that scenario, the fact that the information was private made the person ignore the indicators that they wouldn’t get fired. The act of starting to look for a new job could make the person “check out”, which could actually cause them to be fired in the end, even though there was good evidence to the contrary.
The above example highlights how information cascades can be modified by human psychology. This can be important for trying to understand differences between real-world data and models. This information could also be used to update models to account for human psychology. For example, it is likely that in the real world, most people place more emphasis on their direct observations than on the indirect observations from others. Since we are dealing with probabilities in most of these information cascade problems, we could probably increase the probabilities assigned to some events and decrease the probabilities of others to better match what we actually do when making decisions.