If you do an A/B test you aren't actively disrupting me when I am doing things totally unrelated with what you are selling and impinging on my time. Telemarketing does.
With A/B testing at best you are trying to optimize your website (and indirectly and mostly unobtrusvely impinging the time and energy I choose to put into using your website). The difference is between trying to optimize layout in your store for when I do walk in vs calling me when I am playing with my children.
Telemarketers are active nuisances and in my (politically incorrect) opinion should be lined up against the nearest wall and shot out of hand.
What I'd pay money for is an Eliza like "voice bot", clever enough to waste telemarketers time to the point where the whole idea of telemarketing becomes economically infeasible.
"Here, practice your English."
Hilarity ensued.
However if you're going to "sell a grand piano" your margin is higher and thus you can employ someone who has better sales techniques.
I agree that it is the same game as A/B techniques but there are degrees of viciousness at play here and the classic "piano salesman" is well towards the devious end.
Except -- I wasn't pushing a piano at the cost of a family vacation. I was selling a $5 for three issues ($25 off cover) re-subscription to a computer magazine. Experiencing how hard it actually was to push this offer gave me quite a bit of faith in humanity. 5 our of 6 said no and hung up before I got through my scripted pitch.
There's a certain sentiment against sales here. But where is that anger when someone is talking about their latest A/B test here? An A/B test is a test to find out which of two alternatives is better to make a visitor do what we want that visitor to do. Since the need for the A/B test, we can conclude that the visitor doesn't take that action be default. Thus, we're doing this design to manipulate the visitor to take an action that's not in his best interest? Because if it was in the his best interest, he should be doing it anyway, without the optimization, right?