Presentation on theme: "Welcome Welcome to the Dagstuhl seminar on Plan Recognition"— Presentation transcript:
1 Welcome Welcome to the Dagstuhl seminar on Plan Recognition Please upload titles for the talks you want to giveWe would like everyone to have an opportunity to give a short talkWe have some panel ideas, but these are open to reconsideration – contact meWe will be scheduling incrementallyScheduled through tomorrow…Schedules will be re-posted as updated…
2 Panel ideas Should there be a plan recognition competition? Rational versus fallible agentsActivity recognition, behavior recognition, plan recognition, goal recognitionOh, my!Full and partial observabilityGenerative versus plan library approaches
3 Schedule: Monday AM: Welcome and survey PM: Jerry Hobbs: discourse and plan recognitionShort talksGeorge FergusonMatthew StoneChris Baker: plan recognition and psychologyPanel: a plan recognition competition?Evening: get acquainted event
4 Schedule: Tuesday AM: PM: Kathy Laskey: probabilistic methods for PR Short talksFroduald KabanzaFrancis BissonGita SukthankarPM:Tom Dietterich: learning and plan recognitionDavid PattisonNate BlaylockPanel: Rational versus fallible agents?
5 Plan Recognition Historical Survey Henry KautzUniversity of RochesterRobert P. GoldmanSIFT, LLCOld school plan recognition…Dagstuhl, April 2011
6 Outline Dimensions of the plan recognition problem Historical survey of methodsChallenges
8 Keyhole, intended and adversarial plan recognition Observer non-intrusively watches the agentDetermine how an agent’s actions contribute to achieving possible or stipulated goalsModelWorldAgent’s beliefs
9 Keyhole, intended and adversarial plan recognition Intended recognitionAgent acts in order to signal his beliefs and desires to other agentsSpeech acts – inform, request, …Discourse conventions“The 3:15 train to Windsor?”“Gate 10” [Allen & Perrault]Symbolic actionsThe Statue of Liberty9/11?The agent may require a model of the observer.
10 Keyhole, intended and adversarial plan recognition Agent acts in order manipulate the observerDeception, bluffing, misdirection, etc. …Agent and observer will need sophisticated models of each other’s inferences
11 Ideal versus fallible agents Mistaken beliefsJohn drives to Reagan, but flight leaves from Dulles.The doctor bleeds the patient to cure disease.Cognitive errorsDistracted by the radio, John drives past the exit.Jill schedules a doctor’s appointment during her office hours.IrrationalityJohn furiously blows his horn at the car in front of him.
12 Output of plan recognition Activity recognitionSimply identify a known behavior patternGoalsRecognize the objective, but not the specific recipes usedPlansNext action the agent will take?Best action to aid or counter the agent?
13 Output of plan recognition: likelihood Most likely interpretation?Distribution over plans and goals?The above have subtly different strengths and weaknesses…Most critical plan or goal?
14 Richness of plans Are actions atomic? Or do they have parameters?Structure (e.g., cases)?Do plans have structure and parameters?Coreference?The patient of the plan will be the destination of step one and the patient of step two…Are there plan libraries at all?
15 Other dimensions Reliable versus unreliable observations “There’s a 80% chance John drove to Dulles.”Open versus closed worldsFixed plan library?Fixed set of goals?Fixed set of entities?Metric versus non-metric timeJohn enters a restaurant and leaves 1 hour later.John enters a restaurant and leaves 5 minutes later.Single versus multiple ongoing plans“White knights”Static versus evolving set of intentionsAbandoning goals: I was going to drive to the store, but the weather was too bad.Reacting to opportunities: I was going by the playroom on the way from the laundry, so I picked up the toys.
16 Dimensions Relation to agent Model of agent Goals and plans Observ-ationsInferModel (library)ProduceIntended(possibly) IrrationalStaticNoisyActivityIncomplete“The Answer”KeyholePartial knowledgePartialGoalBest answerAdversar-ialHomo Econom-icusDynamicCompletePlanDistribu-tionNext action
18 Earliest work Generally in service of language understanding Often narrative understandingUnderstanding indirect speech actsAllen & Perrault, “Analyzing Intention in Utterances,” AI, 1980Rich vein of work using plan recognition in dialog understanding and IUIWill be hearing more from George Ferguson later today!Methodologically: Mostly shared early enthusiasm for rule-based systems
19 Hypothesize & Revise The Plan Recognition Problem C. Schmidt, 1978 Based on psychological theories of human narrative understandingMention of objects suggest hypothesisPursue single hypothesis until matching failsThe Plan Recognition Problem C. Schmidt, 1978Related work from Yale AI Lab: Cullingford’s Script Applier Mechanism, Wilensky’s PAM, etc., 1978Charniak, Ms. Malaprop, 1978 – Frame-based and used TMS
20 Closed-world reasoning Infers the minimum set(s) of independent plans that entail the observationsObservations may be incompleteInfallible agentComplete plan libraryLimited to pasta preparationFlying Spaghetti MonsterA Formal Theory of Plan Recognition and its Implementation Henry Kautz, 1991
21 ParsingVilain use parsing results to characterize computational complexity of plan recognitionThere were earlier attempts to parse plansParsing techniques closely related to Closed-world reasoning (Built on Kautz and Allen)Find an explanation that covers all of the observationsParsing techniques deal poorly with partial ordering, worse with interleavingLeads to:Later work on stochastic parsing (Pynadath and Wellman)Attempts to exploit exotic parsing techniques (Geib)
22 Abduction Reason from effect to cause (C.S. Peirce) People: ExplanationDiagnosisPeople:CharniakHobbs et al., TACITUSLeads to interest in Bayes nets
23 Bayes Nets DAG-structured models of probability distributions Came into the fore for diagnostic applicationsChallenge: Static Bayes nets for complex domains can be extremely largeSprinklerRainingGrasswet
24 Bayes NetsKnowledge Based Model Construction: Dynamically build Bayes nets showing how plans explain actionsMultiple goalsAbstraction hierarchiesEquality reasoning for coreferencePoor treatment of time“Jack went to the liquor store.”Was he shopping?“A Bayesian Theory of Plan Recognition,” Charniak and Goldman, AIJ, 1993.“Interpretation as Abduction,” Hobbs, Stickel, Martin & Edwards, Proc. ACL, 1988.
25 More on Bayes net methods Laskey and her colleagues have worked on military domainsFurther developed KBMC techniques (e.g. query completeness); coreference, identity uncertaintyMany related techniquesE.g., Hobbs et al. cost-based abductionATMSes (d’Ambrosio, Provan, Charniak & Goldman)Horn logic (Poole)
26 Pending setsExplicitly models the agent’s “plan agenda” using Poole’s “probabilistic Horn abduction” rulesBridge between Bayes net and HMM frameworksHandles multiple concurrent interleaved plans & negative evidenceNumber of different possible pending sets can grow exponentiallyHappen(X,T+1) Pending(P,T), X in P, Pick(X,P,T+1).Pending(P’,T+1) Pending(P,T), Leaves(L), Progress(L, P, P’, T+1).A new model of plan recognition. Goldman, Geib, and Miller,1999“A probabilistic plan recognition algorithm based on plan tree grammars,” Geib and Goldman, AIJ, 2009.
27 Version Space AlgebraRecognizes novel plansComplete observationsA sound and fast goal recognizer Lesh & Etzioni, IJCAI 1995Programming by Demonstration Using Version Space Algebra Lau, Wolfman, Domingos, Weld.Related to later work on plan-recognition through planning
29 Evaluation Ground truth Prediction tasks Difficult to get labeled data Epistemic question --- do our proposed labelings correspond to any real ground truth?Prediction tasksNext action?Future action?Good choice of assistive action? Countermeasure?Can prediction act as proxy for ground truth?
30 Epistemic questionWhat is the status of the recipes that we postulate as explanations for actions?Are they taken as being real in some sense?Corresponding to mental contents?Identified regularities that really exist in the world?Data structures that just exist for our convenience
31 Computational difficulties Computational complexityTheoretical resultsPractical resultsChallenges from domainsSome domains inherently ambiguousAdversarial reasoningDo we need game-theoretic reasoningCooperative as well as adversarial
32 Plan libraries Engineered? Learned? Something in between? Learned ones often seem impoverishedEngineering seems impossible!
33 Learning Structural learning Parameter learning Learn the contents of plan libraries (in one form or another)Parameter learningAdjust parameters of known librariesBoth offer challenges related to those of evaluationPlan recognition may be done in service of learning, as well as the other way around.Infer goals to learn novel recipes
35 User modelsIn many domains, the behaviors exhibited are not just a function of the actions, goals and plans, but agent characteristics, as well.Developing clean ways to combine agent-dependent and – independent information is a challenge going forward.Often per-agent training is unacceptable.
36 Sensing In many cases it is difficult to sense the agents’ actions: Labeling actions in primitive sensor dataVisionNetwork packetsLinguistic utterancesHardware/software hybrid systemsE.g., oil refinery --- user can go out and use a wrench un-observedConventional softwareEven Horvitz et al. report difficulties “seeing” actions of Microsoft Office usersMixed streamsIndividual actions in network packet streams
37 Coreference and quantification In some domains we don’t have object identity and permanence and the number of agents simply handed to us.Story understandingMilitary situation interpretationIdentity hypotheses enter into plan recognition
38 Anomaly detectionOften appealed to as a solution for detecting some phenomenon that is difficult to model:Intrusion behavior in computer securityTerrorist behavior in tracking and camera dataDementia-induced behavior in tracking elderly subjectsAccuracy requires deep understanding of the models’ propertiesStationarity (often violated in computer security)“Size” and “shape” of normal behaviorsAs always, it’s hard to get something for nothing.
39 The Role of StateMany (but not all) plan recognition systems represent only the state of the planning agent.The state of the environment is modeled implicitly, if at all.
40 Groups Teamwork Role recognition Friendly: recognize teammates’ intentions to coordinate and aidHostile: recognize opponents’ intentions to hinder and obstructRole recognition
41 Hypothesis retrievalSome early work assumed that there were enough candidate hypotheses that retrieval could be an issue
42 Predictive and explanatory inference A lot of concern in early work about combining top-down and bottom-up inference
43 Actions with weak diagnostic power E.g., computer securityWe would desperately like to know the attacker’s motivationsBut what do we do withGet access to the targetGain administrator privileges on the target…