I wrote the draft article in the linked PDF about four years ago. As is the way with all too many drafts it then worked its way to the bottom of my workload and stayed there, never making it to a final draft stage. I'm publishing it now because I think its main argument resonates with some current debates about public attitudes and public perceptions, particularly with regard to poverty.
The basic argument is that we should take widespread false beliefs about society and the mechanisms by which these beliefs propagate seriously as an object of study. Three main mechanisms are described: rational error, cognitive bias and spontaneous social theory (this does not claim to be exhaustive). I argue that all three of these mechanisms have an important social aspect and that the interaction of people in similar social situations can have feedback effects on all three, so widespread error can be a social phenomenon not just in a quantitative sense but in terms of the processes which generate it. This allows for an account of socially situated false beliefs which doesn't depend on a theory of ideology in any of its usual senses: we can get from the individual to the collective without positing collective beliefs, false consciousness or manipulation by elites.
Although the argument is illustrated using an outlandish example (claims that 'welfare mothers' in the States had smaller vocabularies than middle class toddlers) it's intended for more general application. For example, if we want to understand the impact on public opinion of media coverage of benefits (which is dominated by gross stereotyping of claimants as Ben Baumberg, Kate Bell and I showed http://www.turn2us.org.uk/PDF/Benefits%20Stigma%20in%20Britain.pdf) we need to ask about why audiences are receptive to this type of material. Otherwise we wind up open to accusations of the type: 'how typical of those on the left to think that people who hold views they don’t like must have been brainwashed.' http://mainlymacro.blogspot.co.uk/2013/10/rational-expectations-media-an... The rough account of 'social error' here provides a way of thinking about this which avoids the two extremes of pure elite manipulation ('brainwashing') and 'expressive beliefs', where false beliefs are held to be expressions of deep-seated concerns, which can have the unfortunate if unintended effect of partially legitimating them -see Bobby Duffy and Hetan Shah here http://www.newstatesman.com/politics/2013/07/muslims-benefits-and-teenag... .
The content makes no claims to originality: there are heavy debts to Jon Elster (on ideology), to half-remembered bits of Quine and Davidson (on the inescapability of theory) and to cognitive psychology (an area which I knew even less about at the time than I do now). These are just the debts I'm conscious of - I suspect there are also arguments I'd picked up elsewhere and forgotten the provenance of. The 'economic' argument about the costs of error versus the costs of information acquisition in section 2 is very similar to an argument made much more elegantly by Simon Wren-Lewis here http://mainlymacro.blogspot.co.uk/2013/10/rational-expectations-media-an.... (It was Simon's piece that inspired me to dig out the article.) A background influence is Condorcet's analysis of collective judgment and its associated 'sociology of error'. (The argument that impartiality can be a source of error is one I haven't seen before though.)
I wouldn't write this article in the same way today (but I don't have time to rewrite it from scratch). One of the many things that's wrong with it is that it doesn't take account of memorability as a factor in the dissemination of false beliefs: the sheer implausibility of the example used helps explain how it came to travel so widely, as once heard it is hard to forget. Dan Sperber's work on the 'epidemiology of representations' (http://www.dan.sperber.fr/wp-content/uploads/CulturallyTransmittedMisbel... ) which I didn't know at the time of drafting is crucial here (and informed the work on media coverage cited above). Another problem is that it uses an unanalysed naive concept of 'belief', not taking account of the fact that we hold beliefs to differing extents and arguably in different ways: did people /really/ believe that 'welfare moms' had the vocabulary of toddlers - in the same way that they believed they knew their own addresses? Probably not, but I'm not in a position to provide an account of these sorts of differences in belief-holding, and I haven't seen one that convinces me. Finally if I was writing it today I'd feel obliged to learn something about the new subject of 'Agnotology' (the study of ignorance) http://www.sup.org/book.cgi?id=11232 .