Why does push polling work?

As with all writing file in the /opinions section of this blog, this is strictly my thoughts on a subject and lack any formal empirical backing. These are just conclusions that I have arrived at through roughly the logic process outlined herein. I allow myself a lose leash on conjectors. As always, I gladly welcome ideas to the contrary.

Push polling is the act of presenting one's self as a poll (typically a political poll) and asking questions in a way intended to bias the response or better yet bias the opinion of the respondee.

I can think of three classes of push poll style questions.

The outright lie: "Are you aware that President Elect Trump has promised to release all child molesters from prisons?"

The irrelevant association: "How do you feel about the fact that your representative lives next door to a convicted murderer?"

The veiled implication: "Have you ever suspected Eric Garcetti of being corrupt?"

Push polling can be done over the phone, online, or via any channel one can conduct a poll of any kind. The ones I've encountered in my life seemed - to me - to be very flimsy and obvious. Although I have to confess, perhaps I'm only able to detect the most obvious of them and perhaps there are master craftsmen able to produce push polls slick enough to fool me. If I find it indistinguishable from a regular poll. My current beliefs that all push polls are obvious are very consistent with the world in which I am only able to detect the obvious ones and therefore assume all push polls are obvious.

I do think awareness of push polls are likely to make people more able to detect them. Yet, I'm not sure everyone is aware that push polls exist. Perhaps it would be a public service for a skeptic to create silly obvious push polls and put them out into the wild to make people cognisant that push polls exist. I'm on the fence about that, but as a sort of art project, I think it's a clever sort of guerilla education tactic.

The question is: do push polls work? My minimal web searches did not turn up any robust empirical studies on this topic. I'm sure there's something deep in the psychological literature, but I wasn't able to find it.

PBS did a program called Dirty Politics 2008 that covers this. Of particular interest is McCain's angry reaction when George Bush claimed to have no involvement in the push polls accusing McCain of all sorts of falsehoods. This alone is not evidence one way or the other, but it is one person of some importance who considered it a credible threat.

Call me cynical or call me a realist, but I believe there's a large part of our population that could easily be taken in by a push poll. I've heard too many times in my life that people say "I don't like politian so-and-so for reason X" where reason X is a demonstrable falsehood. I (sometimes begrudgingly) support everyone's right to pick whatever candidate they prefer, but I really wish people did so for factual reasons, even if they're inane or irrelevant reasons.

This got me thinking about the mechanism by which push polls effect people. We know that our memories can be flawed, and we have some biological understandings of why. Is susceptibility to push polls a design flaw in our species or is it just stupidity and lack of critical thinking? A design flaw can only be fixed through engineering. The absence of critical thinking can be fixed through education.

How would a Bayesian look at push polls?

Let's start with the prior. I believe there are two situations - new uninformed cases and new evidence cases. "Do you think Barack Obama's tax policies on the importing of barley have helped American farmers?" I have no friggin clue! I am deeply unqualified from having an opinion on this very specific matter distant from my area of expertise.

Yet after reading that, new information has inserted itself into my brain. I guess Obama must, in fact, have a tax policy about the importing of barley, when it is actually more plausible that his tax policies have no specific coverage or reference to barley whatsoever. If I consider my mind in a mechanical fashion, we can say that I have an internal state space representation of the world. This question has now surreptitiously injected a new state which capture's my belief over whether or not Obama has a good barley policy. No problem so far, this seems like good inferencing.

I do find a problem in what my mind does next, however. My beliefs should be a distribution over the probability that the policies help American farmers. My prior should be uninformed since this is all new to me. I believe the smart thing to do would be to have a uniform prior over whether or not I think it's a good policy. However, the very fact that we are discussing it makes me assume some controversy exists. Why would there be a discussion about irrelevant things? I don't yet have an opinion here, but my internal prior distribution seems to skew to the edges. Surely I will find my beliefs to be strongly in support of, or against the president on this matter if its a topic of some controversy, right?

Put more bluntly, one could ask "Was Alexander Hamilton a child molester?" Never considered it before. My prior should be the same as my belief over the entire population. I believe the number of people that are child molesters is fleetingly small. I hope far less than 1%. If anything, a public figure is less likely than my general population prior because they're watched more and people benefit from exposing their crimes.

Yet the question being asked implies controversy which seems to effect the forming of our priors in cases where we had no prior over the specific claim.

What about the Bayesian update? I think here we have the types of failings people are more familiar with. "Did you know that Adolf Hitler was known for drowning stray cats in his neighborhood as a child?" Hitler was a horrible enough human being without adding in fiction that distracts us from the attrocities he actually committed. Further, no one really wants to be in the position to have to defend Hitler, so incorrect information which happens to be consistent with our existing beliefs is generally not commented on at best, but often accepted.

I often find myself either disgusted with the "heathens" that don't know how to properly vet and absorb new information or other times compassionate to people who must not have had the same education as I did and failed to be exposed to logic. It's easy for me to get a bit elitist about updating beliefs. Even if I'm not perfect at it, I know the best mechanism. But I find the mechanism of my mind to be dangerously susceptible to this means of "hacking my prior".

Speculation is a critical component to intelligence. When evaluating a new idea, one must ponder what the world would be like if the idea were true, and what it would be like if the idea were false. Can we identify how those two worlds would be different and then check the real world at the points of disagreement to see which hypothetical world we live in?

That's good scientific thinking, but it also means that people whose minds are tuned to invoke that process are going to spend time, maybe even automatically, exploring some ideas whose truth cannot easily be established. "Did you know your representative is a secret member of Isis?" If true, they're going to do everything they can to avoid revealing it, so the ability to vet the information is going to be quite difficult.

True, at some point, the grandness of the conspiracy required for an unprovable idea to be true is so great, we must envoke complexity theory or just plain ockham's razor. But I find time spent daydreaming an idea can have a slight nagging bias, even for unlikely ideas.

Perhaps your readers are more in control of your minds than I am, and are able to deftly avoid this issue of an apparent controversy implying that some evidence must exist changing your prior in one way or the other. But for me, I do find an implication makes me think my posterior surely must be anything other than equal to my prior.

With that observation in hand, I must conclude that it's possible for clever people to author push polls which could be, to me, indistinguishable from regular polls. Hopefully I'm only vunerable to the top echelon world-class manipulators. But whatever my ability to spot and reject, I find this tactic to be a perfectly legal but credible threat to our societal intelligence. I don't think we can outlaw push polls, although some well thought out restrictions might be useful. I think the message has to be vigilance.

Every poll we participate in should be questioned about its agenda. That doesn't mean you have to spend an hour researching the firm and it's employees. It simply means we should ask these questions: Is it plausible the polling company was attempting to bias my beliefs? If they were successful, is that in any way negative to me? If so, what "new information" might I have implicitly or explicity picked up from the poll alone?