Constructed Preferences For Creationism
My starting point is a debate between P.Z. Myers and Wesley Elsberry about the phrasing of the questions. Elsberry points out that the depressingly low support for evolution (specifically, for the teaching of evolution as opposed to creationism or intelligent design in schools) shown in the poll contrasts with other polls, and he attributes the discrepancy to the polls' wording -- the question in the recent poll was framed in such a way as to make evolution less appealing as an answer. Myers agrees, but disputes that this shows that "framing works." If by "framing works" they mean "re-framing our message would be an effective political strategy," I don't think this poll discrepancy shows a whole lot. But if they mean "framing is a valid scientific theory," then the poll discrepancy is a good illustration of a phenomenon that, on the basis of a great deal of other research, is indisputably true.
Framing effects fall under the larger umbrella of the idea of constructed preferences. There is a temptation, in looking at such poll discrepancies, to try to figure out what question wording would be unbiased and thus reveal the true views of the people of Florida. This assumes that there is such a thing as "the true views of the people of Florida" prior to asking the question.
Rationalist theories, of the type that underlie traditional utilitarianism, neoclassical economics, traditional political science, and common sense, hold that people have preexisting preferences that are called up and expressed in response to a relevant question or choice situation. Such settled preferences would not be vulnerable to any framing effects short of outright trickery. Myers, for example, notes that he would have no problem picking the pro-evolution answer to even the slanted-against-evolution poll. This should come as no surprise, since if anyone should have a settled preference for teaching evolution in schools, it's an evolutionary biology professor who runs a popular blog largely dedicated to defending the theory of evolution.
But psychologists have found repeatedly that question framing makes a difference. People choose a low-risk, low-payoff gamble over a high-risk, high-payoff one, but rate the latter as more attractive. People are more willing to go a medical procedure that 9 in 10 people survive than a procedure that kills 1 in 10 people. People are willing to pay the same amount to clean up one polluted lake in Ontario as they are to clean up all of the polluted lakes in Ontario. People's preferred price for a bottle of wine can be doubled or halved by asking them whether they'd pay more or less than the last two digits of their Social Security Number, prior to asking them to name their price.
What has become clear is that this is not a case of "rhetorical tricks" fooling people into misstating their true preferences. Rather, on non-core questions -- questions to which your answer isn't central to how you see yourself and live your life -- people don't have preexisting preferences. Their preferences are constructed on the fly. Once presented with a question, they try to figure out what preference to have. In such a situation, they take cues from how the question is presented. Since different questions give different cues, they come up with different, yet still genuine, preferences in different situations.
It's quite likely that a large number of Floridians simply don't have settled preferences about which theory of human origins should be taught in schools. Instead, they wait until it becomes necessary to take a stand (e.g. when a pollster calls, or a referendum appears on their ballot), then figure out which position seems best in the context of the moment.
Myers is on the right track when he says that he prefers the slanted-against-evolution poll because it pushes people's "religion button" in the same way that the actual public debate over Florida's science education policy does. Insofar as the poll's framing matches the framing to be found in a particular real-world situation, the poll will lead people to construct their preferences for curriculum content in the same way and thus provide an accurate prediction. Pointing to the different results obtained by a differently-framed poll is not, however, a "blindfold" or "framing the problem away." The two polls in conjunction reveal important information about the prevalence of constructed versus settled preferences among the anti-evolution respondents to Myers' preferred poll. And it gives us a prediction of what the outcome would be if the influences on preference construction in the real world were somehow -- neither poll tells us how -- altered to match it.