Falsification in Sociological Threshold Models
In class, we talked extensively about threshold models, especially in their relation to collective behavior. Take, for example, the classic example of IPhone users versus Android users: if you, a long-time Android user, are deciding upon a new phone, you may take into account that 40% and rising of your friends now have IPhones, and you too may choose to switch for the benefits that accompany having a similar phone as your friends. This specific example is propagated in almost every aspect of life — take for example, the choice between a new PlayStation or a new Xbox (which system are your friends on?), using Facebook or Twitter (which one is used by the people you want to interact with?), or between attending one event or another (which one will your friends be at?). We split up the choice into two primary categories, direct benefits and informational benefits, and discussed at lengths the reasoning and the consequences of such models.
While reading through Mark Granovetter’s paper on Threshold Models of Collective Behavior, I chanced upon an example that was also brought up in our textbook: when a certain threshold of people, let us say 30%, put up their umbrellas at the same time, others soon bring their umbrellas out too. At first, this seems to mimic our traditional model — we have a certain threshold at which it becomes “better” (loosely) for observers to choose one action over another. In this particular case, when the number of people raising umbrellas is 30%, those who have not yet chosen to raise their umbrellas are hit with a informational benefit train of thought — most other people are raising their umbrellas, they must have some information I don’t, so let me parallel their action and also pull out my umbrella. From here, traditionally the choice begins to cascade as more people pull out umbrellas, which causes the threshold of more people to be reached and for them to also pull out umbrellas, and so on.
However, Granovetter warns of the falsification of such a model. We cannot be too quick to assign this situation to be a threshold model, as there are other effects at play. In this case, we must note that there may be an outside factor that is causing these people to bring out their umbrellas, aka, the fact that it is beginning to rain. We can see now that instead of people reacting mutually to one another raising their umbrellas, they perhaps are all just reacting to their need for protection from the rain falling on their heads. Granovetter goes on to give other examples where the threshold model falls short of explaining human behavior, even at times incorrectly predicting what is to happen due to a simple lack of complexity — see his example on riot behavior for more in depth analysis.
All in all, although threshold models can be very useful in predicting and explaining the behaviors we see in our every day lives, we must take care as to not simplify the world too much — there are plenty of other factors out there that can serve as counterexamples to prove our model false. Instead, we must use this idea developed in class of a threshold model as one of many tools, with a right time to use it and a wrong time to use it.
https://sociology.stanford.edu/sites/default/files/publications/threshold_models_ajs_1978.pdf