There are many reasons to be discouraged about the state of personal “information responsibility,” including confirmation bias, the internet echo chamber, the malign effect that Texas has on the contents of American textbooks, et cetera. So it is nice to read some responsible, measured discourse indicating that things are sometimes not quite as horrible as typically reported. Not a cause for rejoicing, mind you, but you take what you can get.
And I recently heard a talk by Arthur Lupia ("Challenges and Opportunities in Open-Ended Coding", presented at a workshop on The Future of Survey Research) that made me even less willing to accept at face value claims of the form "Fewer than X% of Americans Know Y". Arthur reported on some forensic analysis, so to speak, of the internal records of the American National Election Study. He learned that the standard methodology, used in this and other surveys for asking, recording, and scoring open-ended questions (and especially open-ended recall questions), systematically underestimates respondents' knowledge.