1March 2004 © Kathleen M. Carley – CASOS – ISRI - CMU ActionsKathleen M. CarleyMarch 2004 © Kathleen M. Carley – CASOS – ISRI - CMUProblemPredict future directions for complex events When is a strike likely to occur When will a riot turn violent What type of attack is al-Qa’ida likely to try next Who will be assassinated nextExplain past events Why did X happen2March 2004 © Kathleen M. Carley – CASOS – ISRI - CMUMake this harder In particular predict and explain events where individual choice and social action are not filtered through the lens of votes or money Votes and money are “easy” Votes – preferences and combining rules Money – optimization framework and tradeoffs In both cases – analysis is facilitated byHaving a uniform scale for evaluationHaving numeric scales – direct comparison is possibleFeedback is relatively quickEvaluation criteria can be objective Consequently it is relatively straight forward to construct aggregation rulesMarch 2004 © Kathleen M. Carley – CASOS – ISRI - CMUExemplar workSee work of TsandholmUsing computational economics3March 2004 © Kathleen M. Carley – CASOS – ISRI - CMUAnd then there are physical actions …Physical movement entails other problemsRecognition of environmentAwareness of locationPhysical limitations…This is hard – but, to an extent, well definedMarch 2004 © Kathleen M. Carley – CASOS – ISRI - CMUExemplar workSee work by VelosoSee robo soccerSee Blocks-world and work by Pat Winston4March 2004 © Kathleen M. Carley – CASOS – ISRI - CMUMany types of actions are harder still (or less understood)Consider actions like support takeover bid laughOften no scale, let alone uniform scale for evaluation Numeric scales for comparing strength of actions at different times and by different actors is difficult scales may differ by individuals so not clear how to “combine” actions across individuals to get group effects – what are principles of aggregationThe impact of actions may be invisible occur at different time scalesEvaluation criteria are often subjectiveMarch 2004 © Kathleen M. Carley – CASOS – ISRI - CMUApproach 1: Story UnderstandingIdea - there are canonical actions Self-loco-motion, mechanical-loco-motion, speech act, ...Create a complete list of all such actionsDefine a template (schema, frame, …) for each action Self-loco-motionActor – valueActed upon – noneGoal – valueSpeed – valueDirection – valueType – valueDifferent actions can be characterized by filling different valuesExemplar work – Allan Collins5March 2004 © Kathleen M. Carley – CASOS – ISRI - CMUAction SchemasDefine characteristics of action – what happensIdentifies actor (acting)Identifies object (acted upon)Places qualifiers on specific instantiationMay place constraints on actionMay set action pre-requisitesMay set temporal constraintsMarch 2004 © Kathleen M. Carley – CASOS – ISRI - CMUActs as SchemasWorks well in small well defined domainsWorks when an ontology can be well defined and is shared by system usersThis gets complicated when: the impact of an action varies with who the acting and acted upon arethe same action can meet several goalsactions require multiple actors to coordinate actionactions are perpetrated on multiple others and the impact on each can be differentoutcomes are probabilisticThere are approaches – just it’s an unsolved problem6March 2004 © Kathleen M. Carley – CASOS – ISRI - CMUUsing schema’s to predict actionLocating behavioral or historical regularities for schemas is dauntingBuilding cross-cultural schemas may not be possibleDoesn’t explain individual differences in what actions are takenDoesn’t explain outcome differences due to the same actionMarch 2004 © Kathleen M. Carley – CASOS – ISRI - CMUApproach 2: Expert systemIdea – there are condition-action pairings that govern behaviorCreate a list of rules that link conditions to actionsComplications These rules may or may not be probabilistic May or may not be complex (multiple sub conditions may co-exist) May or may not require temporal reasoning (but that’s hard)Choices determine the type of inference engine usedFurther two approaches All actions are possible and rules eliminate possibilities Generate only those actions that match possibilities7March 2004 © Kathleen M. Carley – CASOS – ISRI - CMURules and ActionsWorks well in complex knowledge intensive domainThis gets complicated when The same conditions imply multiple actions Rules become contradictory Knowledge engineering techniques Can’t find rulesRules are actor dependentRules are object dependentRules require an emotional pre-requisiteDifferent scenarios generate seemingly contradictory rulesMarch 2004 © Kathleen M. Carley – CASOS – ISRI - CMUUsing rules to predict actionLocating behavioral or historical regularities for schemas is dauntingBuilding cross-cultural rules may not be possibleIndividual differences in what actions are taken can be handled by making actors objects in the expert systems statespace – Requires class of actors or look up tables for individualsDoesn’t explain outcome differences due to the same action8March 2004 © Kathleen M. Carley – CASOS – ISRI - CMUApproach 3: Goal based reasoningIdea – individuals take actions to meet goalsOptimization procedures are used to locate action most likely to achieve goalThis gets complicated when Actions are not readily placed in a numeric framework Goals are not easily articulated Actions cannot be directly related to goals Emotions, location, etc. alter the relation of actions to goal Actors are not “rational”March 2004 © Kathleen M. Carley – CASOS – ISRI - CMUGoal Based ApproachWorks reasonably well for things that can be monetarily characterizedGenerally produces behavior distinct from individual behaviorOptimization criteria need to be adjusted for cultures, individuals, etc. – so not that generalGets very complicated very quickly when: multi-criteria optimization is needed (multiple competing goals at multiple levels) goal change over time9March 2004 © Kathleen M. Carley – CASOS – ISRI -
or
We will never post anything without your permission.
Don't have an account? Sign up