Consilience concepts

From Starcraft 2 AI
Jump to: navigation, search

Consilience

Consilience is the idea that other fields of thinking operate in similar ways. Also known as 'thinking in systems'. Other systems that work similarly can help you better understand the system you are working on.

Metaphor-based Heuristic algorithms

Meta concepts

  • "Uplift" - purposefully directing a lower-form of intelligence to become more intelligent. (Opposite of "Prime Directive".
    • "Progressor" is someone engaged in 'uplifting', "Pro-active or progressive evolution" as 'provolution' ('provolve').
    • "Cultural Uplift" as a proactive and progressive uplift via cultural evolution, without physical/biological changes (e.g. "Primate Learning Center" in Iowa. )

System concepts

Movement behavior

Behavior Change concepts

Conditioning concepts

  • Respondent conditioning ('classical'/'Pavlovian')
  • Operant (or instrumental) conditioning - a learning process through which the strength of a behavior is modified by reward or punishment.
    • Matching Law is a quantitative relationship that holds between the relative rates of response and the relative rates of reinforcement in concurrent schedules of reinforcement. For example, if two response alternatives A and B are offered to an organism, the ratio of response rates to A and B equals the ratio of reinforcements yielded by each response.
      • Equation: If R1 and R2 are the rate of responses on two schedules that yield obtained (as distinct from programmed) rates of reinforcement Rf1 and Rf2, the strict matching law holds that the relative response rate R1 / (R1 + R2) matches, that is, equals, the relative reinforcement rate Rf1 / (Rf1 + Rf2)
      • Deviations from matching: undermatching, overmatching and bias
        • Generalization of Matching Law: Log(R1/R2) = Log(b) + s * Log(Rf1/Rf2) -- The constants b and s are referred to as "bias" and "sensitivity" respectively. "Bias" reflects any tendency the subject may have to prefer one response over the other. "Sensitivity" reflects the degree to which the reinforcement ratio actually impacts the choice ratio. When this equation is plotted, the result is straight line; sensitivity changes the slope and bias changes the intercept of this line. The generalized matching law accounts for high proportions of the variance in most experiments on concurrent variable interval schedules in non-humans. Values of b often depend on details of the experiment set up, but values of s are consistently found to be around 0.8, whereas the value required for strict matching would be 1.0. (Also expressed as R1/R2 = b(Rf1/Rf2)^s)
      • Importance The matching law is theoretically important for several reasons. First, it offers a simple quantification of behavior that can be applied to a number of situations. Secondly, offers a lawful account of choice. As Herrnstein (1970) expressed it, under an operant analysis, choice is nothing but behavior set into the context of other behavior. The matching law thus challenges the idea that choice is an unpredictable outcome of free will, just as B.F. Skinner and others have argued.

Motivation concepts

  • Maslow's Hierarchy of Needs
  • "Motivating Operation" - behavior depends not only on the stimuli present in the current situation and the organism's past experience with those stimuli, but also on the organism's recent past history of deprivation, satiation, pain, or other such influences. Such a past history can have two effects: it can change the value of a consequence by making it more or less reinforcing, and/or it can change the probability of behaviors that have produced that consequence. A motivating operation differs from a discriminative stimulus (Sd). A discriminative stimulus signals the availability of reinforcement, while a motivating operation changes the effectiveness of a reinforcer.
    • 9 main unconditioned (i.e. not learned) motivating operations, have been identified in humans. Deprivation of food, water, sleep, activity, or oxygen; becoming too warm or too cold; and increase of a painful stimulus all function as establishing operations for related behaviors, and increase the effect of positive or negative reinforcement related to them.
    • Conditioned motivating operations that result from the learning history of the organism. Three kinds of conditioned operations have been identified: a surrogate, reflexive, and transitive. A surrogate MO has the same effect as the MO it was paired with when it was learned; a reflexive MO acts as a reinforcement when it is removed; a transitive MO make something else effective as reinforcement.

Decision-making concepts

Teaching concepts

  • Keller Plan -- A Personalized System of Instruction (PSI), with Written materials, Units of Content, Self-paced Instruction, Unit Mastery, Proctors.