looking at the human condition, I see plenty of data to the opposite.
I see it too - but you're only talking about the present. Put it into historical context and it indicates the opposite of what you think it indicates. The history of bullshit is that, while there is still too much of it, it started at vastly worse premodern levels that were really weird, and has been losing arguments ever since.
One of my favorite examples is the reports of the first Protestant ministers who went through the 16th century villages, talked to peasants about what they actually believed and, because they weren't evaluating their own work, actually recorded that. Turns out even after hundreds of years of Catholicism, these peasants had all sorts of ideas, and those ideas varied wildly from village to village or person to person. They'd have three gods, or eight, or even believe in forms of reincarnation. The Protestants went to homogenize that, of course, and in places like Brazil they still do. But even in Europe and the US, the number of distinct belief systems continues to decline.
Far from creating one shared space, today's interconnectivity seems to have led to a bunch of echo-chamber bubbles
Medieval villages and similarly unconnected societies are echo-chamber bubbles too. So the number of bubbles has been going down, sharply, and competition between ideas has clearly become tougher. If an exception like Homeopathy is growing, that means it has been unusually successful in the harder environment (a big part, in this case, was greatly reduced claims of effectiveness). But that shouldn't distract from the fact that lots of pseudotherapies that were comparable to it fifty years ago, such as Anthroposophic medicine and Orgone therapy, have gone down. And of course the quacks and witch doctors that we used to have before those were even more heterogenous and numerous.
And that's exactly what you'd expect to see in a world where whether someone accepts an idea very much depends on what else that someone already believes. People aren't usually choosing rationally what to believe, but they're definitely choosing.
There seems to be a hard-coded exception for young kids, who will believe any bullshit their parents tell them, and that, rather than active conversion, is how some religions continue to grow. Surely it also helps bullshit that isn't religion.
I'm obviously not doing this experiment you're talking about, because it is wildly unethical and incurs severe social cost. And even if it turned out I can convince people of bullshit just as well as I convince them of rationality, that wouldn't be relevant to my original assertion that convincing the unconvinced is not at all an "open and hard problem".
It seems that the CFAR workshops so far have been dedicated to people who have preconceptions pretty close in ideaspace to the sorts of ideas proposed on LW and by the institutions related to it. This is not a criticism; it's easier to start out this way: as has been said, in a different context and perhaps not in so many words, we should focus on precision before tractability. We're not going to learn a thing about the effectiveness of rationality training from people who won't even listen to what we have to say. Nevertheless, there will come a day when these efforts must be expanded to people who don't already view us as high in social status, so we still have to solve the problem of people being more concerned with both our and their social status than with listening to what we have to say. I propose that the solution is to divorce the consideration of social status from the argument.
There is a lot of talk of cognitive biases on LW, and for good reason, but ultimately what we are trying to teach people is that they are prone to misinterpreting reality, and cognitive biases are only one component of this. One of the problems with trying to teach people about biases is that people feel personally responsible for being biased; many people have a conception of thinking as an 'active' process, so they feel as though it reflects upon their character. On the other hand, many people conceive of perception as a 'passive' process; no one feels personally responsible for what they perceive. So, I propose that we circumvent this fear of character assassination by demonstrating how people can misinterpret reality through perception. Enter: the rubber hand illusion.
In case you're unfamiliar with this illusion, to demonstrate the rubber hand illusion, a subject sits at a table, a rubber hand is placed in front of them, oriented relative to their body as a natural hand would be, and a partition is placed between the rubber hand and their 'real' hand such that they are unable to see the 'real' hand. Then, the experimenter simultaneously 'stimulates' both hands at random intervals (usually by stroking each hand with a paintbrush). Then, the experimenter overextends the tips of a finger on each hand, the rubber hand about 90 degrees, and the 'real' hand about 20 degrees (it's not really overextension, and it wouldn't cause pain outside of the experiment's conditions). Measurements of skin conductance response indicate that subjects anticipate pain when this is done, and a very small selection of subjects even report actually experiencing pain. Also, (just for kicks) when subjects are questioned about the degree to which they believe their 'real' finger was bent, they overestimate, by an average of about 20 degrees.
As Dr. Vilayanur Ramachandran has demonstrated, the rubber hand illusion isn't the most general example of this sort of illusion: the human mind can even anticipate pain from injury to the surface of a table. In fact, there is evidence that the human mind's evaluation of what is and is not part of its body isn't even dependent upon distance: Dr. Ramachandran has also demonstrated this with rubber hands attached to unnaturally long rubber arms.
I think that there are also three beneficial side effects to this exercise. (1) We are trying to convince people that Bayesian inference is a useful way to form beliefs, and this illusion demonstrates that every human mind already unconsciously uses Bayesian inference all of the time (namely, to infer what is and isn't its body). To further demonstrate the part about Bayesian inference, I would suggest that subjects also subsequently be shown how the illusion does not occur when the rubber hand is perpendicular to the 'real' hand or when the 'stimulations' aren't simultaneous. (2) After the fact, the demonstration grants social status to the demonstrator in the eyes of the subject: "This person showed me something that I consider extremely significant and that I didn't know about, therefore, they must be important." (3) Inconsistencies in perception instill feelings of self-doubt and incredulity, which makes it easier to change one's mind.
Addendum: This post has been substantially edited, both for brevity and on the basis of mistakes mentioned in the comments, such that some of the comments now appear nonsensical. Here is a draft that I found on my desktop which as far as I can tell is identical to the original post: http://pastebin.com/BL81VQVp