I am happy to say that the author answered many important concerns in this new version of the paper, which now clearly states that it is not arguing against scientific personality psychology, unlike other situationist authors.
I still fear that some readers might read the paper and conclude that personality psychology is unscientific and rests on folk intuitions alone. The author encourages this impression by stating repeatedly that the scientific study of personality is an interesting endeavour to be undertaken in the future (p. 3-4, p. 22). This statement seem to ignore the fact that a science of personality has been existing for more than half a century, in psychology, genetics, neuroscience, etc. (even though it is not, of course, above criticism). It would be nice to read one sentence recognizing this fact.
Two points of concern:
On p.3, the author argues that correlations between traits and individual actions are undetectable because they are not very strong, as shown by Mischel in 1968, and because we are bad at spotting correlations of any kind, as shown by Jennings et al. Mischel's work is now much too old to be convincing. A decade later, Nisbett found that correlations were actually 30% higher. I think a reference from the 2000s would be indispensable here. (And, if the author wishes to imply that personality traits are much less important than situational factors, it should be noted that correlation sizes between situations and behaviors are quite often not stronger than that - or else, prove the contrary).
On the topic of the false consensus effect, p. 8-9, I think the author may have misunderstood Dawes and Mulford's critique. In the false consensus experiments, the subjects do not have the choice to make a prediction or not to make one. They have to answer the experimenter's question, even though they are given very little data to rely on, a fact which subjects can do nothing to change. Dawes and Mulford ask what the most rational option is, given those bad conditions: should people take into account their own decision in predicting others', or should they abstain from using that information? The second option means preferring to rely on no data at all. The first option means relying on just one data point. Dawes and Mulford conclude that the first option (use one data point) is more rational than the second (do not use the available data). It would, of course, be more rational to use better data, as the author points out. But the data is simply not there to be used, and the subjects have to make a decision (this is part of the experimental set-up). They make the best of a bad situation. Therefore, the authors should not so readily accuse human stupidity.