Eliminating Negative Consequences of Cascades in Groups
https://hbr.org/2014/12/making-dumb-groups-smarter
In class recently, we have been studying information cascades and various ways of looking at them. No matter what setting, be it a network or sequential decision making process, cascades occur because the participants neglect their own private information about whatever was going on, and would imitate what their predecessors were doing, as this action yields the greatest payoff, based on the total information the individual has. While cascades can lead to beneficial consequences (think wisdom of crowds), they do run the risk of yielding unfavorable ones. In this post, we look at the causes of bad decision making in groups, and potential solutions aimed at reducing the chances of such incidents.
The article “Making Dumb Groups Smarter” by Cass Sunstein and Reid Hastie is a discussion about the causes of erroneous decision making in groups, and potential ways of averting this phenomenon. To explain why groups can have tendencies to make bad decisions, the authors bring up several reasons tied to human psychology, but place the most blame on cascading. Expanding on cascades, the authors bring up two different types; “informational” and “reputational” cascades. In the words of the authors, “In informational cascades, people silence themselves out of deference to the information conveyed by others. In reputational cascades, they silence themselves to avoid the opprobrium of others”. Following the introduction of these terms, the article brings up an example of each. The example of the informational cascade occurred in a series of mock trials staged by Hastie, where the jurors, upon seeing the verdicts of other jurors, would forsake their own verdicts for the one that the majority of preceding jurors had endorsed. The reputational cascade involved an example where the conviction of one member of a new company project was followed by people who were actually skeptical about it, because the dissenters did not want to look potentially foolish or adversarial. Finally, the authors suggested several solutions aimed at curbing bad decision making in groups. Among them, the ones of greatest relatability to our course were rewarding group success, and the “Delphi Method”. The latter consists of several cycles of decision making, where individuals submit their decisions and then discuss them. After that, they decide and discuss repeatedly until some sort of consensus is reached among group members. An alternative version of the method is “a system in which ultimate judgments or votes are given anonymously but only after deliberation”.
The informational cascades mentioned above are obviously cascades caused by informational effects. The reputational cascades are related to direct-benefit effects, where group members stand to benefit from each other by maintaining good social standings. The solutions that were proposed by the author are relevant to the class in the following ways. The idea that groups can be rewarded for good decision making is a testament to the fact that individuals in cascades base their choices on potential payoffs. If the payoff for imitating is higher, as it is in cascades, then the individual proceeds with that action. Each decision made by an individual is a type of game, where depending on the choices that other group members made, the individual’s choice will yield different payoffs. So a participant would be likelier to engage in a different behavior than their predecessors, with the knowledge that they have “nothing to gain from a correct individual decision and everything to gain from a correct group decision”. The Delphi Method can prevent cascades by actually eliminating the need, or reducing the potential benefit, of imitating others’ behavior. Direct-benefit effects are removed, since the decision making may be made anonymously, and informational effects may be minimized since everyone may now disclose their own personal information. So the payoffs that each individual has change with this set-up.