Moo Tang Clan: moderating miscreants

Tuesday, May 20, 2008

moderating miscreants

If your goal is to encourage behaviour over a sustained period, then intermittent positive reinforcement works better than periodic positive reinforcement ... but which works better for negative reinforcement? That is, if there is some specific behaviour you wish to punish.

I'm told that random negative reinforcement works best (if you're attempting to instill terror).

The problem with random negative reinforcement in social systems is that the absence of punishment on others is used to rationalise away guilt: "he didn't get punished for doing X, therefore I shouldn't have. I shouldn't have been punished, therefore action X was not a wrong thing to do".

What if, instead of meting out punishment 1:1 in response to crime, but instead the total punishment due is accumulated into a growing debt and eventually meted out, in whole, by random selection, to some unlucky miscreant? For example: abusive name calling in the forums is punishable with a banning of 1 hour, but there's only a 10% chance you'll get punished .. after some time there have been 7 infractions of this type, none punished (so far), and then some poor sod goes off the handle. Instead of copping just a 1 hour banning, he cops the 1 hour ban plus the accumulated punishment debt of 7 hours.

This might sometimes be known as the "daddy is getting angry" model.

For this system to work (ie. not appear entirely capricious) it would be necessary to provide ongoing feedback to the community as to the current level of punishment debt. Other forms of community status/feedback might include a list of currently unpunished miscreants, or (more positively) a leaderboard of well behaved frequent posters.

Variations to this basic model are possible: the punishment debt could aggregate from all crimes instead of separate totals for each type being kept, the chance of the punishment being applied might increase as the debt increases, the chance of punishment might increase if you've previously escaped punishment, there could be some modulation of the upper limit according to the severity of the crime, there could be a gradual decay of the debt over time, and so on.

What might the effects be? If the community has been particularly naughty lately, it would be wise not to do even the littlest thing wrong, lest daddy's wrath descend on you. An acceptable (i.e. non-perfect) level of misbehaviour might be arrived at through community consensus. Some characters might take on the role of lightning rod/martyr, intentionally invoking the wrath of punishment so as to release the built up potential.

No comments: