I don't normally follow social and family dynamics psychology, but there's some research there that has interesting implications for understanding kids on the autistic spectrum better.
Researchers on aggressive children have identified a series of steps in social information processing. These steps include a) encoding social cues from the enviroment, b) interpreting this social information, c) clarifying goals, d) constructing possible responses, e) making a response decision, and f) enacting that social behavior.
The first step, encoding social cues, might include noticing a friend's quick frown and abrupt gesture. In the next step, the child tries to interpret this social information. Is my friend angry? At me, or at someone else? Preoccupied and not wanting to be distracted? Sad? Frustrated? Impatient? Wanting to be alone, or wanting to be comforted? The child draws on a huge body of contextual knowledge, general and specific: what has this friend done in the past? What do these gestures mean in general? For children his age? What was just happening in the immediate interaction that might be causing the friend's behavior? What's happening in the friend's life in general that might be causing it? A staggering amount of information gets processed here, very quickly. During the next step clarifying goals, the child decides what he or she wants from the interaction. Does he or she want to comfort the friend, play with the friend, not deal with the friend's problems? Does he like the friend that much at all? Maybe she'd become more popular if she snubbed the friend--how important is that to her, relative to the friendship?
Next, the child considers possible responses based on all this information. She might choose to try to comfort the friend, distract him by suggesting a game, snub him and go play with someone else, ignore the gesture and keep the conversation going as before, or ask the friend how he's feeling. Next, the child actually picks a response. In the process, he has to predict the consequences of each of these responses: if I try to comfort my friend, will that make her feel better or make her push me away? If I snub him and go play with someone else to be more popular, will I feel guilty?
Finally, the child has to enact the social behavior, choosing the right words, body language, and intonation or the wrong ones. The same social behavior, in the same social situation, could be enacted gracefully or clumsily, leading to very different consequences.
The reason researchers on aggressive kids break down social cognition so finely is that aggressive children seem to show abnormalities on at least one of these stages. For instance, their interpretations (stage 2) are more likely to be negative than other kids'; they're more likely to interpret ambiguous social behavior as hostile. They may also be biased towards aggressive behavior at the response decision step, because they feel more confident with their ability to carry out an aggressive action than an acceptable one.
But this typology has far greater significance than this. We've seen that problems with understanding others' behavior and feelings can mean many different things, leading to people with very different cognitive profiles being lumped together as ASD. A fine-grained model like this could help researchers study each link in the chain, and could help clinicians better understand where kids are having problems and target their interventions more precisely. If we can import this model into research on ASD, we could have a much clearer understanding of what it means to understand others--and why so many of us fail.