The Ellsberg paradox is often cited as evidence for unknowable “ambiguity” versus computable “risk”, and a refutation of expected utility maximization and “subjective” or “belief-type” probabilities. I have concluded the paradox is not convincing. The results can be explained by differences in distributions that show up in repeated “games.” The distributions behind Ellsberg’s thought experiments are different and economic agents should be expected to respond to these differences.
The 16-page paper (.pdf) can be found on SSRN, updated May 2011 or here on the Close Mountain site