Have you ever picked up the phone to answer a call from a pollster? I can probably guess your answer, and that’s not uncommon. A tiny fraction of people actually answers those calls, which creates a predictably skewed result if political campaigns are truly relying on those calls to generate any useful insight. The modern solution to this that some researchers are advocating is to just have pollsters “call” artificial intelligence simulations of voters instead.
In early testing, these simulations seem to be largely accurate in depicting human responses to open ended questions. Given the notorious unreliability of information from polls already, many are arguing that doing polling this way can hardly be worse. And perhaps it could be even better because pollsters might finally get some insight about how “regular” people think rather than those at the fringes who hold the most radical opinions, or those who happen to have the most free time to answer polling calls in the first place.
