Sunday 3 October 2010

Theoretical abstract animal AI game model


- - > this symbol means behavioral change from low to high variable value

  • “Happiness” variable:
Sadness (loss) - - > Joy (gain)
- change is based on result of actions, not initial intention. Based on that change - weakens/reinforces costs for behaviors
//models value of feedback, the “intent != result” principle
//models confidence, over & underestimation of the self

Affected by:
Pain
Loneliness
Boredom
etc

  • “Weight of latest events for changing costs” variable
Conservative (less affected by change) - - > liberal (too much biased towards new data)
//models adaptation to environment changes
//models sensing (gaining new data) VS intuition & knowledge (old data)

///

  • “Intellect” variable:
Emotional heuristics - - > Rational calculation, thinking in perspective
Fast with errors decision making - - > slow but thorough decision making
//models short term pain, long term gain situations

  • “Autonomy” variable
Obey with no responsibility - - > Free with much responsibility - - > Power, take freedom from others to self
//Models responsibility
//combined with bad attitude to others models power / dominance seeking

///

  • “Value of other people” variable for each known person:
Person has negative value (applies parasitism, aggression, enemy) - - > Person has zero value (egoism) - - > Person has positive value, but still less than 1 (all values are normalized, relative to mine) - - > Person has 1 and is equal to me (cooperation, morality) - - > Person is more valuable than self (martyrdom)
//models cooperation and morality, variable may be constant zero for non-living entities

  • “Danger / Benefit” evaluation:
Evaluation_for_action = My_Benefit – My_Cost +
            Sum_of_the_below_for_everybody_affected:
Value_of_that_person * (His_Benefit – His_Cost)  + (Expected_Help_from_this_person - Expected_Danger_from_this_person )

//The last line is basically a part of My_Cost, but I put it there to make it explicit                     
The action with best evaluation is chosen, possible actions:
Run away (fear)
Wait and do nothing (lazy, apathetic)
Help somebody
Attack Somebody
Self Defense
Call for help, attract attention
Trade
Lie / Cheat
Achieve some other goal, progress, resource

//models drive to achieve goals, to progress, to take initiative (fear, laziness, etc)
//takes into account that actions can be mutually beneficial (symbiotic) or mutually exclusive (parasitic) regardless of number of participants

  • “Default Cost for unknown inanimate entity” variable
Low (avoid unknowns) - - > high (exploration, curiosity)
//Models drive to understand the world

  • “Default Cost of unknown person” variable
Low (avoid unknowns) - - > high (social, charismatic)
//Models drive to socialize

//both of the 2 vars above are forms of learning, and that is form of self improvement.
//Variables are used instead of constants to model the risks in that learning (attitude towards the unknown)

///

  • Physical variables:
Health
Life Span
Speed
Damage
Cost (affects numbers)

Potential Problems: