Many of us experience such a high degree of empathy that we are constantly putting ourselves in other people's shoes and trying to see all sides in any controversy or conflict. Many of our problems with sensory and emotional overload derive from an excess of this ability, not a deficit.
1. Rachel challenges whether people with autism have 'theory of mind' difficulties and instead argues that people with autism have high degrees of empathy.I am not Rachel, but it seems to me like she wasn't denying the existence of those results at all; just saying that those results don't tell us much of anything about what the autistic people in the various studies were actually thinking that led them to do the test "wrong." We know that autistic people don't interpret social situations the same way non-autistic people do; what we don't know is how autistic people do interpret them.
This is however hard to reconcile with the scientific evidence. Literally dozens of studies from around the world have documented the theory of mind difficulties in autism. And the empathy difficulties are also well documented and widely replicated, both on performance tests (e.g., emotion-recognition tests from the face and voice) and on self-report measures (such as the Empathy Quotient or EQ).
Consider the latter, where 81% of people with autism score less than 30/80 on the EQ, by their own self-report, whilst only 12% of people without autism score at this low level. These results are mirrored when parents complete the EQ about their children, in many independent samples. So, whilst some people believe that theory of mind and empathy difficulties in autism are mythical, the results of many independent scientific studies suggest otherwise.
2. Rachel challenges whether people with autism have difficulty knowing when they have hurt others, and wishes I had not stated that children with Asperger syndrome (AS) are delayed in being able to figure out what might hurt another person. Indeed, she finds my statement hurtful.
As a working scientist, all I can do is summarize the empirical evidence. An example is the Faux Pas Test, where children are asked to identify if anyone said anything they shouldn't have said, whilst listening to short audio recorded stories. Children with AS as a group on average scored significantly lower than children without AS, despite being older than the comparison group. Indeed, the design of this experiment allowed us to estimate the size of the developmental delay in AS, since the 12 year old children with AS performed more like typical 9 year olds. So, although Rachel may not like hearing these results, this is what the science finds.
(I am basing the above paragraph on what Rachel has written in this post on the Sally-Anne test for Theory of Mind, as well as her three-part series critiquing the EQ. Her post on the Sally-Anne test, in particular, is interesting because it describes an alternate thought process for a hypothetical autistic child taking the test, and coming up with reasons why Sally might look in other places than where the researcher wants her to look. What's important is that it's not that the kid in this example can't imagine Sally's point of view, it's that ze is drawing on different thought processes and experiences to arrive at different predictions for what Sally will do. The problem with the test is that it treats all wrong answers as failures to imagine Sally's mental state.)
Because I understand Prof. Baron-Cohen's need for empirical validation of these possibilities, I've even come up with an experimental design he (or anyone) could use to evaluate the two-way-street hypothesis that I've been promulgating here.
You'd have a bunch of people, half of them autistic and half of them not, and you would group them into pairs, with each pair including one autistic and one non-autistic member, both the same sex, same age and roughly similar verbal abilities, and just have them interact together for a short while, like 5-10 minutes. You would record their interaction on video, and then you would ask each person, separately, some questions about what happened between them. What they thought felt at certain points (decided on by the researchers sometime between the actual exchange and the individual Q&A sessions), what they thought the other person was thinking or feeling at certain points. You would then compare the two participants' answers to see how well they overlapped. You would also do this with autistic/autistic and NT/NT pairings, and then compare the average degree of similarity of the paired accounts across all three permutations.
This would allow you to see whether autistic people seem to understand each other better than they do NTs, or whether NTs are equally baffled by autistics.
It would also end the stalemate between Baron-Cohen's "well, the evidence says autistic people just don't understand social situations" and autistic people's self-reported experiences of both understanding other autistic people, and of having non-autistic people spectacularly fail to understand, or empathize with, them.