Tom Coates talks about using an inverse-learning curve (or learning-cliff-with-an-overhang) to defeat spammers and other abusers of internet systems:
“Essentially what many systems online need – the systems that are prone to abuse that is – are graphs of ‘difficulty of use’ that are exponential – start almost flat and then escalate heavily afterwards.”
Cf. with notions of “seamful design” for beautiful executions of this concept? Any existing examples?