I strongly doubt that there is any distinction between ‘stereotype’ and ‘heuristic’, ‘inductive reasoning’ or ‘statistical inference’ to be found strictly in their information content.
Perhaps, as a group, stereotypes are more resistant to modification based on updated data (Just like a stereotype would be, to lazy, pig-headed, and generally ignorant to listen to the facts…); but people do lousy heuristic work all the time that isn’t ‘stereotyping’, just bad stats, and statistical truth is not generally considered to be a defense against charges of ‘stereotyping’, however strong the data.
Again, as with our discussion some time back about the ‘Boston Celtics’ vs. the ‘Washington Redskins’, it’s mostly a matter of the social power and hierarchy-reinforcing deployment of various verbal constructs, rather than the constructs themselves (which don’t even have to make any sense: much schoolyard trash-talk makes almost no sense at all, as a sentence that ostensibly conveys meaning, except that it is understood by all to be an attack on the target.)
By way of example: there isn’t any informational difference between the statistical/epidemiological reasoning that would cause a public health outfit to target anti-violence interventions at poor urban minorities, especially males, and the stereotype that young black men are violent and dangerous. The same would be true of a genetic counselor paying particular attention to a British couple of Pakistani immigrant background, vs. and EDL skinhead with some booze in him shouting about imbred pakis.
We usually call them ‘stereotypes’ when statements or beliefs in the form of an inductive inference (whether true or not) are used primarily for their value as social/hierarchical weapons. Informationallly indistinguishable (if, probably better informed on the whole, since stereotype users are not primarily truth-motivated, though they may use true statements when it suits them, while researchers do tend to be theoretically truth motivated, if sometimes misled, incompetent, or otherwise fallible) assertions not used for their value as social blunt instruments; but for more benign purposes, are usually classified as non-stereotypes.
(One other salient factor, though it can, again, occur as cognitive blinding on the part of somebody who we wouldn’t use the term ‘stereotyping’ for, might be the willingness of the person holding the inductive inference to treat individual cases who deviate from the statistical norm as individual instances. Statistical induction is a fine way, often your only one, of assessing things that you otherwise would have no data about; but it’s a rare population indeed where all members have the same traits as the ‘average’ member. If you are fully truth-motivated, and successful in being so(the human brain is not evolved to be a truth machine, don’t let it trick you.), you will replace statistical likelihoods with individual facts as the facts become available. If you are primarily interested in using things that look like statistical inferences as tools of power, it will behoove you to treat individuals, even if they exhibit signs of being statistically atypical, as though they, personally, are simply a local instance of the ‘average’ member of whatever group you’ve assigned them to.)
