Making Exceptions to General Rules

Sup­pose you make a gen­eral rule, ie. “I won’t eat any cook­ies”. Then you en­counter a situ­a­tion that le­gi­t­i­mately feels ex­cep­tional , “Th­ese are gen­er­ally con­sid­ered the best cook­ies in the en­tire state”. This tends to make peo­ple torn be­tween two threads of rea­son­ing:

1) Clearly the op­ti­mal strat­egy is to make an ex­cep­tion this one time and then fol­low the rule the rest of the time.

2) If you break the rule this one time, then you risk dis­man­tling the rule and end­ing up not fol­low­ing it at all.

How can we re­solve this? For a very, very long time I didn’t want to take op­tion 2) be­cause I felt that tak­ing a sub-op­ti­mal strat­egy was ir­ra­tional. It may seem silly, but I found this in­cred­ibly emo­tion­ally com­pel­ling and I had no idea of how to re­spond to this with logic. Surely ra­tio­nal­ity should never re­quire you to be ir­ra­tional? It took me liter­ally years to work out, but 1) is an in­cred­ibly mis­lead­ing way of fram­ing the situ­a­tion. In­stead, we should be think­ing:

3) If fu­ture you will always make the most ra­tio­nal de­ci­sion, the op­ti­mal strat­egy is go­ing to clearly be to make the ex­cep­tion. If there is a chance that it will cause you to fall off the path, ei­ther tem­porar­ily or per­ma­nently, then we need to ac­count for these prob­a­bil­ities in the ex­pected value calcu­la­tion.

As soon as I re­al­ised this was the more ac­cu­rate way of fram­ing 1), all of its rhetor­i­cal power dis­ap­peared. Who cares about what is best for a perfectly ra­tio­nal agent? That isn’t you. The other key benefit of this fram­ing is that it pushes you to think in terms of prob­a­bil­ity. Too of­ten I’ve thought, “I can make an ex­cep­tion with­out it be­com­ing a habit”. This is bad prac­tise. In­stead of think­ing in terms of a bi­nary “Can I?” or “Can’t I?”, we should be think­ing in terms of prob­a­bil­ities. Firstly, be­cause 0% prob­a­bil­ity is un­re­al­is­tic, sec­ondly be­cause mak­ing a prob­a­bil­ity es­ti­mate al­lows you to bet­ter cal­ibrate over time.

We can also up­date 2) to make it a more so­phis­ti­cated ar­gu­ment as well.

4) If you break the rule this one time, then you risk dis­man­tling the rule and end­ing up not fol­low­ing it at all. Fur­ther, hu­mans tend to heav­ily bi­ased to­wards be­liev­ing that their fu­ture selves will make the de­ci­sions that they want it to make. So much so, that at­tempt­ing the calcu­late this prob­a­bil­ity is hope­less. In­stead, you should only make an ex­cep­tion if the util­ity gain would be so much that you would be will­ing to lose the habit al­to­gether.

We still have two differ­ent ways of think­ing of the prob­lem, but at least they are more so­phis­ti­cated than when we started.

(This ar­ti­cle was in­spired by see­ing: The Soli­taire Prin­ci­ple. I wanted to ex­plain how I’ve pro­gressed on this is­sue and I also wanted to have a post to link peo­ple to which is much shorter)