Beware Ideal Screen Theories

Variable B "screens" variable A from variable C when learning the value of B makes A and C no longer dependent on one another; once you know B, A says nothing about C.   Screening is a useful concept, but we are often over eager to apply it.  For example:

Mood Swings – Since your internal state must pass through time, you know that in the absence of outside influences, your state today can only depend on your state two days ago via the intermediary of your state yesterday.  So if something bad happened to you two days ago, but yesterday you felt fine, you might conclude you are over it; that bad event can't hurt your mood today unless it causes some new outside influence on you.  Alas, your mood only summarizes a small part of your internal state.  What happened two days ago can pop up and bother you today, even if yesterday you were fine. 

Disagreement – When someone disagrees with you, you should wonder what they know that you do not. They might explain their reasons for their differing belief, i.e., their evidence and analysis, and you might hear and ponder those reasons and yet find that you still disagree.  In this case you might feel that the fact that they disagree no longer informs you on this topic; the reasons for their belief screen their belief from informing your belief.  And yes, if they could give you all their reasons, that would be enough.  But except in a few extremely formal contexts, this is not even remotely close to being true.  We are usually only aware of a small fraction of the relevant evidence and analysis that influences our beliefs.   Disagreement is problematic, even after you've exchanged reasons.

Evolved Betrayal – We take actions that influence people around us, and we wonder how blameworthy we are regarding those actions.  We know evolution shaped our minds to promote our selfish genetic interests relative to others, but we'd like to feel we can ignore that fact when we are consciously aware of positive intentions toward them.  If our conscious intentions toward others were our only evolution-influenced mental factors which change our behavior toward others, this would be correct; intentions would screen evolved selfishness from our behavior.  Alas, this seems quite unlikely.  Our minds are very complex, and a great many processes influence each choice we make, processes about which we are mostly unaware. 

For example, if we take an action that gives us selfish benefits, and if our minds saw clues with enough info to feasibly identify that selfish action, the fact that we had no conscious awareness of intending to achieve that selfish benefit should offer little reassurance.  It is a good bet that our mind was influenced by this selfish benefit, as well as by the impressions others might get from seeing such a selfish action.  You can hurt the ones you love, on "purpose."

GD Star Rating
Tagged as: ,
Trackback URL:
  • Eliezer Yudkowsky

    Understanding check: You mean that if someone is aware that love evolved for evolutionary reasons, and is also aware of the adaptation-executing-not-fitness-maximizing principle, it is imprudent for them to say:

    “Well, I know, my dear, that my love for you is an adaptation for the sake of reproductive fitness. But evolution can’t dynamically reconfigure me; and so all that matters from a psychological perspective is the fact that I love you; and so I can expect to carry out all the actions that would be associated with a true pure love for you, into the indefinite future.”

    Because there’s still leftover unscreened pattern that has to do with the evolutionary shaping, like:

    “We need to stay together for at least three years before we marry to find out if we’ll still like each other after the initial pair-bonding passion wears off, especially if we don’t plan on having children.”

  • Eliezer, yes, exactly.

  • Abigail

    “the fact that we had no conscious awareness of intending to achieve that selfish benefit should offer little reassurance.”

    For many of my actions, I can imagine bad motives, and good motives. “Things done wrong, and done to others’ harm, which once you took for exercise of virtue. Then fools’ approval stings, and honour stains” (TS Eliot, “Little Gidding”). I consider that too much navel-gazing will not result in true self-knowledge of motives. I may get to know myself better over time, but for the moment I just have to not-know.

    However, if one cannot bear to imagine that one is a bad person, or ever does bad acts, that will bias ones self-knowledge and self-observation.

  • The term “selfish” is awkward to use in this economics/ev. psych context, and I think it might be prudent to avoid it.

    Humans are sometimes “selfish”. This is an in-the-brain process involving computing means and ends and choosing an antisocial action that benefits the individual. Call this “brain selfishness”

    Anything which has undergone natural selection will behave “selfishly”, even if it does not have any brain or process akin to choosing. Call this “gene selfishness”.

    In the last paragraph, for example: “If we take an action that gives us selfish benefits…” does “selfish” mean “brain selfish” or “gene selfish”?

    Consider someone grasping something to prevent falling. Did the person see the object? Maybe. Did the person realize that they could grasp it in order to prevent falling? Maybe. Did they use means-ends thinking (“brain selfishness”) to achieve their goals by grasping it? Probably not. The part of our brain that does means-ends and social thinking is slower than that.

    I concede that we might unconsciously be doing antisocial means-ends thinking a lot of the time. But sometimes we just execute, without invoking the modules involved in “brain selfishness”.

  • Sideways

    Is this a restatement of the principles of correlation and causation? Feeling happy yesterday is correlated to feeling happy today; disagreement is correlated with doubt in your own position; conscious positive intentions are correlated to altruistic actions. There’s even a causal link in each of Robin’s examples.

    Of course, an outcome can (and usually does) have more than one cause. I would sum up Robin’s argument as: people are biased to pick their favorite contributing causes for a particular outcome, and ignore or underestimate others. I don’t think the concept of “screening” contributes anything over and above causation and correlation.

  • Edward

    ‘We know evolution shaped our minds to promote our selfish genetic interests relative to others’

    Yes, but it is also true that evolution has shaped our minds to promote co-operation in groups for the benefit, protection and survival of the group itself, and leadership behaviour that supports the group and others who are more vulnerable within it. Perhaps this supports the selfish genetic interest of the group.

    These two drives/instincts often come into conflict within an individual and the friction between them is one of the key sources of tension and thus growth and progress in human culture. These tensions might relate back to the urge to achieve Alfa status. Franz De Wall decribes primates seeking to achieve Alfa status and acting in predominantly selfish/agressive ways to achieve it. Once achieved, the Alfa must then act for the benefit of the group in order to successfully hold on to power. Even those primates who do not achieve alfa status will often act for the benefit of the group, and even sacrifice for it.

  • Sideways, the screening assumption is what lets people assume they can ignore other causes.

    Edward, evolution shaped our minds to help the group by helping individuals.

    Johnnicholas, our conscious doubts whether we are acting selfishly is about brain-selfishness.

  • talisman

    Incredible post all around. Each of the examples is interesting, but the abstraction of “beware ideal screen theories” is truly excellent. Thanks!