The author misses the strongest case argument for stereotypes.<p>Cost of trait differentiation at scale.<p>In scenarios where the cost reductions of aggregation are larger than the gains received fron individual processing, stereotyping is the most efficient choice.<p>Example:
There are several bowls of M&Ms, they all have a low but varied percentage of poisoned M&Ms in them and these are difficult to detect. You are competing to eat as many as possible as fast as possible for as long as possible.
Does it make sense to scrutinize every individual candy? Does it make sense to only eat from certain bowls when possible?
If you mix all the bowls together and equalize the risk profile, is this of a higher benefit than dumping the highest risk bowls in the trash?
How do these answers change based on the severity of the poison and the distribution in the various bowls?<p>Arguments against this type of analysis are either moral, structural (conflict costs), or case-based.<p>The author doesn't delve into any of this.