You observe the behaviour. You decide whether you want to increase it or decrease it based on strict ethical criteria (I think it is something like your action should be good for the child AND good for society and you should be able to justify that) and then you start to collect information on it.
Sometimes this is easy. You can ask the child 'why do you do that?' Sometimes it will take a couple of weeks of filling out charts about when the behaviour happens, what is in the environment, what happens just before, what happens just after. Then you create a hypothesis and test it by changing one of the recorded variables and measuring what changes. If it isn't going the way you want after 3 goes you try something different. This does make is sound mechanical and a lot harder work than it is.
The emphasis is always that the child is right and behaving appropriately. If the child is walking out of the classroom, they are right to, because all things in their environment (including their internal environment) are telling them that that is what they should do. You don't try to 'force' anything else.
You might find after observations that the behaviour (i.e. walking out) only occurs during certain activities, when noise levels rise or when. You might notice that noise levels always rise at 'free-time', so you engineer noise levels to rise at a different time to see if it has the same effect.
Once you have established the cause you look into ways that you can address the behaviour. You don't bully a child into discomfort but set up a programme of intervention with very small steps. You would, most likely for a perceived sensory issue, involve an OT to give advice. You would probably then begin a desensitisation programme with the child moving closer and closer to the 'noise' from a distance. BUT, 3 goes (unless your data tells you the child grasps things at 2 goes or 6 goes) and then you change direction. No slogging dead horses because 'well it works for other children with this problem, and it's what we were trained to do, so it must just be that your child isn't capable'.
zzzzzz you express yourself very well. ABAers probably have an advantage in this kind argument because they have spent months if not years fighting for ABA using the best logic they can.
I like your notion of being coached through sex and probably you're right. But with enough coaching you know what to do and can eventually do it spontaneously, and once secure variations on a theme. Some people do employ sex therapists after all.
But you know, I have explained ABA to a HV once as 'extreme good parenting with no let up, never (at the beginning anyway) can you say 'ds get down from that wall, - I said get down, oh never mind, I have a beer and I'm sitting down, I'll address the behaviour tomorrow' like you can with NT children - so I do get your point. I mean, you CAN be inconsistent, but then you can lose the gains you've worked hard to achieve. I suppose that is certainly a negative.