Think Before You Speak (And Signal It)

In de­cid­ing whether to pay at­ten­tion to an idea, a big clue, if it were read­ily available, would be how many peo­ple have checked it over for cor­rect­ness, and for how long. Most new ideas that hu­man be­ings come up with are wrong, and if some­one just thought of some­thing five sec­onds ago and ex­cit­edly wants to tell you about it, prob­a­bly the only benefit of listen­ing is not offend­ing the per­son.

But it seems quite rare for this im­por­tant piece of meta­data to be straight­for­wardly de­clared, per­haps be­cause such dec­la­ra­tions can’t be trusted in gen­eral. In­stead, we usu­ally have to in­fer it from var­i­ous other clues, like the speaker’s per­son­al­ity (how long do they typ­i­cally think be­fore they speak?), for­mal­ity of the lan­guage em­ployed to ex­press the idea, the pres­ence of spel­ling and gram­mar mis­takes, the venue where the idea is pre­sented or pub­lished, etc.

Un­for­tu­nately, such in­fer­ences can be im­pre­cise or er­ror-prone. For ex­am­ple, the same speaker may some­times think a lot be­fore speak­ing, and other times think lit­tle be­fore speak­ing. Us­ing costly sig­nals like for­mal lan­guage is also waste­ful com­pared to ev­ery­one sim­ply tel­ling the truth (but can still be a sec­ond-best solu­tion in low-trust groups). In a com­mu­nity like ours, where most of us are striv­ing to build rep­u­ta­tions for be­ing (or at least try­ing to be) ra­tio­nal and co­op­er­a­tive, and there­fore there is a level of trust higher than usual, it might be worth ex­per­i­ment­ing with a norm of declar­ing how long we’ve thought about each new idea when pre­sent­ing it. This may be ei­ther in ad­di­tion to or as an al­ter­na­tive to other ways of com­mu­ni­cat­ing how con­fi­dent we are about our ideas.

To fol­low my own ad­vice, I’ll say that I’ve thought about this topic off and on for about two weeks, and then spent about three hours writ­ing and re­view­ing this post. I first started think­ing about it at the SIAI de­ci­sion the­ory work­shop, which was the first time I ever worked with a large group of peo­ple on a com­plex prob­lem in real time. I no­ticed that the var­i­ance in the amount of time differ­ent peo­ple spend think­ing through new ideas be­fore they speak is quite high. I was sur­prised to dis­cover, for ex­am­ple, that Gary Drescher has been work­ing on de­ci­sion the­ory for many years and has con­sid­ered and dis­carded about a dozen pos­si­ble solu­tions.

The trig­ger for ac­tu­ally writ­ing this post is yes­ter­day’s Over­com­ing Bias post Twin Con­spir­a­cies, which Robin seemed to have spent much less time think­ing through than usual, but which has no overt in­di­ca­tions of this. (An ob­vi­ous ob­jec­tion that he ap­par­ently failed to con­sider is, wouldn’t cor­po­ra­tions ac­tively re­cruit twins to be co-CEOs if they are so pro­duc­tive? Sev­eral OB com­menters also pointed this out.) A blog­ger may not want to spend days por­ing over ev­ery post, but why not make it eas­ier for the reader to dis­t­in­guish the se­ri­ous, care­fully thought out ideas from the throw­away ones?