selection bias is when an entity (here an individual human or a group of humans such as a family, civilization or market) makes a choice that is fundamentally self-serving. in terms of the vulcan axiom of spock, the needs of the many outweigh the needs of the few or one. selection bias is generally a subconscious or intentionally ignored effect. after all, evolution has instilled individuals with a survival instinct. the major problem with the mechanism of survival is that it subverts complex structures.
the problems with selection bias can also be understood through the analogy of the pyramid. in the analogy of the pyramid, an analogy to the force of gravity is made. the basis, the lowest level, provides support for higher levels. the higher levels add weight to the lower levels. if the lower levels are structurally unsound, the structure collapses. if the upper levels are too heavy or not structurally sound, the structure likewise collapses.
there are other analogies, both for gravity and for other lesser understood forces. for instance, the arch could be used as another analogy for the force of gravity. the problem with analogies is that they are ready-made models that might not fit all the important variables. to understand an existing example, compare newton’s mechanics with einsteinian relativity. newton’s equations work well at the macro level but fall apart at the quantum level.
the five domains are an analysis tool similar to the analogy of the pyramid. yet, they unbind the system from the constraints of the physical reality of gravity. thus, the crushing power of pressure and heat, not to mention the unfathomable consequences of spacetime, are eliminated. the five domains (not counting the null domain) for selection bias are: individual survival instincts, small group survival decisions, the large group survival processes, the system survival homeostasis and — as in all five domain analogies — the fifth domain is currently unknown or unattainable.
in this framework, selection bias in one domain does not generally affect the abstractly mechanical rules in constituent domains. this is due to the rules of order of magnitude. what’s important in the lesser domain is the amount of material and its field effects, not its behavior. the small group domain functions rather well regardless of whether or not individuals in the individual domain practice detrimental selection bias. the localized effects are localized. however, the idea of selection bias can itself infect the small group domain as though it were a field effect or force of its own. if the small group domain begins to practice selection bias, it can cause itself to destabilize. there is little or no cause and effect relationship here as in the analogy of the pyramid. instead, an entire domain can become tainted.
incidentally, machines (here endowed with machine learning abilities), are nothing more than mechanical humans programmed to carry out human abilities efficiently (and likewise reproduce human biases, such as selection bias, efficiently).
