Rationality detection

From Issuepedia
Jump to navigation Jump to search


Rationality detection is any technique whereby the presence or absence of a coherent rational thought process can be objectively determined.


There are many situations in which both parties to a disagreement claim to be reaching a rational decision. Although it is generally acknowledged that there is always the possibility that one party or the other is either lying or irrational, it seems to be generally assumed that the only way of resolving the dispute -- once all premises have been shared, so that both parties are working from the same data -- is to agree on a compromise position.

In cases where one party is wrong, this leaves both parties only half-wrong, which is a questionable gain at best. In cases where one party is lying, that party may not in fact be updating their belief at all (despite agreeing to do so), leaving us with a net loss of correctness.

Rationality being essentially the application of reason to the real world, however, it should be possible to follow each party's line of reasoning and thereby detect the point at which they diverge -- and hence to determine which party is in error.

The ability for others to examine the line of reasoning requires deliberate transparency on the part of each party.


Issuepedia proposes the following set of guidelines for determining if a party (referred to as "you") is making an honest attempt to be rational:

  • you must use only documented reasoning processes:
    • use the best known process(es) for a given class of problem
    • state clearly which particular process(es) you use
    • document any new processes you use
  • you must make every reasonable effort to verify that:
    • your inputs are reasonably accurate, and
    • there are no other reasoning processes which might be better suited to this class of problem, and
    • there are no significant flaws in your application of the reasoning processes you are using, and
    • there are no significant inputs you are ignoring

If your argument satisfies all of these requirements, it is at least provisionally rational. If it fails any one of them, then it's either not rational (and needs to be corrected or discarded) or else you are knowingly failing to make the necessary effort to allow verification, in which case your reasoning may be rational but nobody else is under any obligation to consider it so.

These guidelines are optimized to return fewer false positives (i.e. thinking that someone is rational when they actually are not) than false negatives (i.e. not believing that someone is rational when they actually are).


This page was adapted from a much-disparaged post on LessWrong. A few of the comments may suggest refinements to this page, but most were about the presentation.