Of course, strictly speaking, ABA is simply providing a consequence to any action, either positive or negative, rewarding or punishing. Developmental type treatments actually use ABA techniques, but it is done using a more naturalistic approach. Based on what I have seen, good "natural environment" ABA is very similar to these techniques. The Early Start Denver Model, for example, purports to be a "developmental" treatment for autism when in fact it is really an ABA approach using natural environment teaching in dyadic exchanges, pairing sensory reinforcement with social interaction.
Social communication deficits are of course central to autism as a disability.
According autistics to whom I have spoken to as adults, and some I have read about who are considered "recovered" or "high functioning", they have not suddenly developed the innate ability that neurotypical people have to understand social situations. Rather, they have learned the appropriate response to give during social situations and have adapted to cope with their disability.
ABA has been criticized for making kids robots and if you buy the RDI view, does not give autistics the ability to think dynamically. If you also believe that some autistics are "cured" (I think there is a continuum of being able to cope - the better you cope the more like normal you appear), then theoretically there must exist some former autistics who now have developed the innate ability to read social cues. Do I think these people exist? Yes, probably, but not very many. Very, very few, in fact, far fewer than the number of people who "recover". Why? Because recovery is more often "coping", and not a cure.
Autism is defined by observable behaviour. If you don't have the behaviour, you don't have autism. But you can still have autism and simply control your behaviour to the point that no one notices. You may still not get jokes, understand facial expressions or body language or read emotions all the time, but if you do it well enough some of the time, you don't have autism.
Where does this all come back around to ABA? In my limited experience so far, ABA is most difficult to apply to social skills because it can become very rote and is difficult to generalize. However, the people who have "recovered" by learning how to operate in social circumstances have done so one situation at a time.
Having a computer science background, one of the most interesting concepts to me at university was artificial intelligence. One of the AI tests is called the Turing test, where a person converses with a computer or a real person but is not told to whom they are speaking. If the AI can trick the person into thinking they are real, it passes the Turing test.
Now, given a computer with sufficient exposures to enough social situations, and given a limited enough conversation in both time and depth, it would be reasonable to fool someone in this way. It would be even easier to create a computer program that didn't fool people but that people believed was a good facsimile of a real person.
When it comes to autism - having enough "scripts" and knowledge of different social situations is very similar. The autistic may never be able to pick up on social cues like a typical person, but they can be close. They can be close enough to cope - to fool some people, but be good enough others may detect an oddity but not much else.
Assuming it is not possible in most cases to really cure autism - only to give the ability to cope, ABA can work for social situations given enough training in enough situations. Eventually, autistics will learn for themselves what is expected simply by asking the right questions.
Before I continue - let me reiterate - I believe people are "cured" of autism by coping to a point where they can functionally beat the equivalent of the Turing test with neurotypical people 99% of the time. It happens, but it is probably very rare.
Let me contrast this to another disability. If someone is vision impaired and vision cannot be corrected with drugs or surgery, the next best thing is to give that person coping mechanisms. They learn to read with Braille. They may use a service dog or other method to get around. In the future, we may have the ability to use technology to provide a facsimile for vision (if you know Star Trek, think Geordi Laforge). However, in the end, the disability remains and the vision the blind person experiences will never be the same as that of the person who can see. The reality is that for most disabilities, coping is the best we can do because we can't cure them. If a blind person had eye implants with computerized motion to mimic a seeing person's eyes and converted light to images in the brain using implants so well the person could fly a plane, are they still blind? Yes, they are still blind. But they cope so well, it doesn't matter.
I don't think autism is any different. Some criticize ABA because it's not a real cure. I don't think there is a real cure, at least not in the medical sense of the word. I think some autistics have the capacity to become neurotypicals and a variety of treatments would work for them (in other words, if you are an autistic who can become neurotypical, several treatments could work with roughly the same effectiveness). For everyone else, all the treatments help them cope, and some do this better than others depending on the severity and types of deficits.
Now, a final unrelated point... many people with classic Kanner autism also have intellectual disability. There is a lot of disagreement with this statement and intelligence in autistics is very difficult to measure. But the research seems to state this is currently the prevalent view. If true, even if autism could be cured, intellectual disability (ID) would still remain. In this case I expect most parents of children with ID would want their children to participate the most fully in life they can. In this case, ABA would be warranted if it is an effective teaching method.
Some people call ABA repugnant because it smacks of animal training. But consider that a typical dog has the cognitive ability of a two year old. An adult functioning at a two year cognitive level is profoundly intellectually disabled. Anything they can learn to be more independent is a good thing. Yes, they may have the intelligence of a dog, but they are not a dog. They are human and deserve all of the respect that goes with that title. Using ABA does not rob them of this respect if it is effective!