Very brief introduction to STRUCTURED ARGUMENTATION

Transcription

1 Very brief introduction to STRUCTURED ARGUMENTATION Massimiliano Giacomin DII - Dipartimento di Ingegneria dell Informazione Università degli Studi di Brescia (Italy)

2 STRUCTURED VS ABSTRACT ARGUMENTATION DEFINITION OF ARGUMENT STATUS ABSTRACT ARGUMENTATION DEFINITION OF ATTACK DEFINITION OF CONFLICT DEFINITION OF ARGUMENT STRUCTURED ARGUMENTATION UNDERLYING LANGUAGE (and consequence) 2

3 Outline Structured Argumentation: two illustrative approaches - Rule-based approaches (formal argumentation) - Argument schemes (semi-formal argumentation) Examples of Application: - Argumentation-based recommender systems - ARGMed: an ongoing project 3

4 Rule-based approaches (1) RULES OF INFERENCE OF TWO KINDS: A B: A B: deductive - indefeasible non-deductive - defeasible ARGUMENT DEFINITION a tree made up of rules of inference constructed from a set of premises to reach a conclusion A B C (0.7) D C A strength value may be associated to premises and defeasible rules, giving rise to argument strength A (0.7) B (0.9) C (0.7) D (0.9) C (0.8) See [J.Pollock, 1992], [G. Vreeswijk, 1997], 4

5 Rule-based approaches (2) Notion of conflict TWO BASIC KINDS OF CONFLICT RELATION Rebutting: an argument attacks another one by denying a [possibly intermediate] conclusion (premise, sub-conclusion, conclusion): i.e. conflict of the kind p VS p Undercutting: an argument attacks the applicability of a defeasible rule of inference (without denying its conclusion): i.e. conflict of the kind P C VS P C From P it is not valid to infer C 5

6 Rule-based approaches (3) EXAMPLE Propositional language SeeRed: I see an object red IsRed: The object is red SayNot: Tony says that the object is not red LRed: The object is illuminated by a red light Rules: SeeRed IsRed saynot IsRed LRed SeeRed IsRed Premises: SeeRed, SayNot, LRed SeeRed IsRed SayNot IsRed LRed (SeeRed IsRed) 6

7 Rule-based approaches (4) Notion of defeat (attack) [Pollock 92] An argument α defeats β iff: - α undercuts β, or - α rebuts β and α is not weaker than β (strength, specificity, preference ) EXAMPLE SeeRed IsRed SayNot IsRed LRed (SeeRed IsRed) 7

8 Rule-based approaches (5) EXAMPLE REBUTTING DEFEAT It s raining It s not raining Smith says it s raining Bob says it s not raining 8

9 Rule-based approaches (6) EXAMPLE REBUTTING DEFEAT It s raining It s not raining UNDERCUTTING DEFEAT Bob is unreliable Smith says it s raining Bob says it s not raining Bob is drunk 9

10 Argument schemes (1) ARGUMENT SCHEMES - correspond to recurring patterns of reasoning - have associated critical questions (possible reasons / conditions making the scheme not applicable) Example: Expert Testimony [WALTON 1996] E is expert on D E says P P is in D Therefore, P is the case Critical questions: Is E biased? Is P consistent with what other experts say? Is P consistent with known evidence? 10

11 Argument schemes (2) ARGUMENT SCHEMES IN A MEDICAL APPLICATION Viability Scheme Organ O of donor D is available No contraindications are known for donating O to recipient R Therefore, organ O is viable CRITICAL QUESTIONS: Does donor D have a contraindication for donating organ O? Nonviability Scheme Donor D of organ O has condition C C is a contraindication for donating O Therefore, organ O is nonviable [Tolchinsky et al, 2006] 11

12 Outline Structured Argumentation: two illustrative approaches - Rule-based approaches (formal argumentation) - Argument schemes (semi-formal argumentation) Examples of Application: - Argumentation-based recommender systems - ARGMed: an ongoing project 12

13 Argument-based user modelling Three main techniques 1. Stereotipical reasoning 2. Cognitive filtering 3. Collaborative filtering HOW TO INTEGRATE ALL OF THEM? Argument-based approach 1 DONNA Sesso: F Generi film: sent; Carat_Utente(U,Sesso,F) Stereotipo(U,Donna) Stereotipo(U,Donna) Piace_Carat_Prod(U,Genere,Sent) 2 Acquistato_Prod(U,P) Piace_Prod(U,P) Piace_Prod(U,P) Carat_Prod(P,Genere,C) Piace_Carat_Prod(U,Genere,C) 3 Piace_Prod(U2,P) Utente_Simile(U1,U2) Piace_Prod(U1,P) 13

14 EXAMPLE (a) DEDUCING USER PREFERENCES FROM ACTIONS / BASIC FACTS Carat_Utente (id01,sesso,f) Acquistato_P (id01,f55) Stereotipo (id01,donna) Piace_Prod (id01,f55) Carat_Prod (f55,genere,viol) Piace_Car_Prod (id01,genere,sent) Piace_Car_Prod (id01,genere,viol) [0.5] Piace_Car_Prod (id01,genere,viol) [0,7] 14

15 EXAMPLE (b) IDENTIFYING CONFLICTS Carat_Utente (id01,sesso,f) Acquistato_P (id01,f55) Stereotipo (id01,donna) Piace_Prod (id01,f55) Carat_Prod (f55,genere,viol) Piace_Car_Prod (id01,genere,sent) Piace_Car_Prod (id01,genere,viol) [0.5] Piace_Car_Prod (id01,genere,viol) [0,7] 15

16 EXAMPLE (c) COMPUTING ARGUMENT STATUS [JUSTIFIED] Carat_Utente (id01,sesso,f) Acquistato_P (id01,f55) [JUSTIFIED] [JUSTIFIED] Stereotipo (id01,donna) [JUSTIFIED] Piace_Prod (id01,f55) [JUSTIFIED] Carat_Prod (f55,genere,viol) Piace_Car_Prod (id01,genere,sent) [JUSTIFIED] Piace_Car_Prod (id01,genere,viol) [0.5] [UNJUSTIFIED] Piace_Car_Prod (id01,genere,viol) [0,7] [JUSTIFIED] 16

17 ArgMED: An ongoing project Medical Discussions in Meetings: - discussing clinical cases, diagnoses, methodologies - different points of views, based on evidences - contradictions 17

18 The problem Discussions are never documented (apart from final decisions) - information lost (analysis of symptoms, proposed and discarded hypotheses, alternative interpretation of examination results, reasons of the final decision) - difficult to come back to a discussion after some time Need for a tool to: - document a discussion, in order to recall it - analyze a discussion to identify weaknesses (alternative diagnosis, neglected treatment, side-effect come about, ) Two components - discussion documentation (simple graphical language) - discussion analysis through argumentation schemes 18

19 First component! 19

20 Second component (1) Discussion represented as a multi-level list! 20

21 component (2) AnalisiSecond discussione 21

22 Second component (3) Showing details of the identified arguments 22

23 Example From Alessandro Maffi s presenta8on 24/09/

24 Schemi per il dominio medico Argomento basato sulla conoscenza del dominio Premessa maggiore: I medici <M 1, M 2, M n > sono specializza8 in <S 1, S 2, S n > Premessa minore: <M 1, M 2, M n > asseriscono la proposizione <A> Conclusione: <A> può essere plausibilmente considerata vera CQ1: La proposizione <A> è consistente con quanto dicono altri esper8? CQ2: L affermazione fala da <M 1, M 2, M n > si basa su delle prove? Alessandro Maffi 24/09/

25 Schemi per il dominio medico Argomento a favore di un tra;amento Premessa maggiore: Il paziente è affelo dalla malama <M> Premessa minore: Il tralamento <T> è in grado di curare <M>. Conclusione: È opportuno procedere con il tralamento <T>. CQ1: Ci sono altre malame <M > cui dovrebbe essere data priorità? CQ2: Ci sono effem collaterali <E1, E2,...En> più indesiderabili di <M>? CQ3: Ci sono effem collaterali <E1, E2,...En> e c è un tralamento <T'> che cura <M> senza presentare gli effem collaterali <E1,E2,...En>? Alessandro Maffi 24/09/

26 La discussione completa Alessandro Maffi 24/09/

27 FINAL COMMENTS 27

28 Applications (1) Representing human discussions Medical discussions (ARGMed) Representing legal knowledge and reasoning Representing design decisions Argument mining on the web (e.g. argument-based trip-advisor) Driving human discussions E-democracy All previous applications Knowledge elicitation and exploitation System Autonomy Decision support systems Medical diagnosis 28

29 Applications (2) Recommender systems User modelling Giving suggestions Modelling scientific reasoning Argument mining from scientific papers Meta-analysis from medical papers and trials Agent and multi-agents systems Dialogue Automated negotiation All the forgotten ones and those still to think about 29

30 Final comments Argumentation can be applied to any situation where - incomplete / uncertain information is exploited - heterogeneous data source must be managed - subjectivity must be taken into account 30

31 Final comments Argumentation can be applied to almost any situation where - incomplete / uncertain information is exploited - heterogeneous data source must be managed - subjectivity must be taken into account 31

32 Final comments Argumentation can be applied to almost any situation where - incomplete / uncertain information is exploited - heterogeneous data source must be managed - subjectivity must be taken into account Any specific approach can be potentially exploited With the advantages of - natural explanation - dynamicity - dialogical side Still, many research issues: - relation between abstract / structured argumentation - reasoning at different levels of granularity, visualization, - diverse applications: inform theories!!! 32

33 Some Master Thesis proposals Working on ARGMed project - identification of argument schemes from medical literature - argument (and relationship) visualization - development, experiments and validation - rules elicitation Working on other application domains - analyzing examples and identifying an argument-based modelling Working on theory - extending structured argumentation with additional truth values - relationships between argument schemes, abstract argumentation, structured argumentation - summarization, reasoning at different granularity levels - relationship with further uncertainty techniques (e.g. probability) - algorithms and parallelization 33

34 MANY THANKS FOR YOUR KIND ATTENTION 34

35 Selected references (1) Landmark argumentation papers and books S. Toulmin, The Uses of Argument Cambridge University Press, R. P. Loui, Defeat Among Arguments: a System of Defeasible Inference, Computational Intelligence, vol. 3(3), J. Pollock, Defeasible Reasoning, Cognitive Science, vol. 11(4), G. Simari & R. P. Loui, A mathematical treatment of defeasible reasoning and its implementation, Artificial Intelligence, vol. 53(2-3),

36 Selected references (2) Argumentation surveys and books F. van Eemeren et al., Fundamentals of Argumentation Theory: a handbook of historical backgrounds and contemporary developments, L. Erlbaum Ass., C.I. Chesnevar, A.G. Maguitman, R.P. Loui, Logical models of argument, ACM Computing Surveys, vol. 32(4), H. Prakken & G.A.W. Vreeswijk, Logics for Defeasible Argumentation, in Handbook of Philosophical Logic, 2nd Edition, Kluwer Academic Publishers, H. Prakken, An overview of formal models of argumentation and their application in philosophy, Studies in Logic, vol. 4(1), pp 65-86,

37 Selected references (3) Argumentation semantics survey P. Baroni, M. Caminada, M. Giacomin An introduction to argumentation semantics, The Knowledge Engineering Review, vol. 26(4),2011. Books D. Walton, Fundamentals of critical argumentation, Cambridge University Press, P. Besnard & A. Hunter, Elements of Argumentation, MIT Press, Argumentation in Artificial Intelligence, edited by I. Rahwan and G. R. Simari, Springer, Dung s influential paper on abstract argumentation P.M. Dung, On the Acceptability of Arguments and Its Fundamental Role in Nonmonotonic Reasoning, Logic Programming, and n-person Games, Artificial Intelligence, vol. 77(2),

38 Selected references (4) Semantics P. Baroni, M. Giacomin, G. Guida, SCC-recursiveness: a general schema for argumentation semantics Artificial Intelligence, vol. 168(1-2), B. Verheij, Two approaches to dialectical argumentation:admissible sets and argumentation stages, Proc. of the 8th Dutch Conference on Artificial Intelligence, 1996 M. Caminada, Semi-Stable Semantics, Proc. of 1st International Conference on Computational Models of Arguments (COMMA 2006), 2006 P.M. Dung, P. Mancarella, F. Toni, A dialectic procedure for sceptical, assumption-based argumentation, Proc. of 1st International Conference on Computational Models of Arguments (COMMA 2006), 2006 S. Coste-Marquis, C. Devred, P. Marquis, "Prudent Semantics for Argumentation Frameworks", Proc. of 17th IEEE International Conference on Tools with Artificial Intelligence (ICTAI 2005),

39 Selected references (5) Semantics H. Jakobovits & D. Vermeir, "Robust Semantics for Argumentation Frameworks", Journal of Logic and Computation 9(2), 1999 P. Baroni, M. Giacomin, G. Guida, SCC-recursiveness: a general schema for argumentation semantics Artificial Intelligence, vol. 168(1-2), P. Baroni, M. Giacomin, Resolution-based argumentation semantics, Proc. of 2nd International Conference on Computational Models of Arguments (COMMA 2008), 2008 G.A. Bodanza, F.A. Tohmé, Two approaches to the problems of self-attacking arguments and general odd-length cycles of attack Journal of Applied Logic, to appear. P. Baroni, P. Dunne, M. Giacomin, Computational Properties of Resolution-based Grounded Semantics, IJCAI 2009, to appear. 39

40 Selected references (6) Algorithms, dialectical proof theories, dynamics P. Baroni, M. Giacomin, G. Guida,.SCC-recursiveness: a general schema for argumentation semantics. Artificial Intelligence, vol. 168(1-2), R. Baumann. Splitting an argumentation framework. Proc. of LPNMR R. Baumann, G. Brewka and R. Wong. Splitting Argumentation Frameworks: An Empirical Evaluation. Proc. of TAFA B. Liao, L. Jin, R. C. Koons. Dynamics of argumentation systems: A division-based method. Artificial Intelligence, vol. 175, P. Baroni, M. Giacomin, B. Liao. On topology-related properties of abstract argumentation semantics. A correction and extension to Dynamics of argumentation systems: A division-based method. Artificial Intelligence, vol. 212, S. Modgil, M. Caminada. Proof Theories and Algorithms for Abstract Argumentation Frameworks. Chapter 6 of Argumentation in Artificial Intelligence, G. Simari and I. Rahwan (Eds). Springer,

41 Selected references (7) General criteria for semantics evaluation and comparison M. Caminada & L. Amgoud, On the evaluation of argumentation formalisms, Artificial Intelligence, vol. 171(5-6), P. Baroni, M. Giacomin, G. Guida, On principle-based evaluation of extension-based argumentation semantics, Artificial Intelligence, vol. 171(10-15), P. Baroni, M. Giacomin, Skepticism relations for comparing argumentation semantics, International Journal of Approximate Reasoning, vol. 50(6), W. Dvorak & S. Woltran, On the Intertranslatability of Argumentation Semantics, Journal of Artificial Intelligence Research, 41, ,

42 Selected references (8) Other interesting investigations E. Oikarinen & S. Woltran, Characterizing strong equivalence for argumentation frameworks, Artificial Intelligence, vol. 175(14), S. Gaggl & S. Woltran, Strong Equivalence for Argumentation Semantics based on Conflict-free Sets, Proc. of 11th European Conference on Symbolic and Quantitative Approaches to Reasoning with Uncertainty (ECSQARU 2011), 38-49, R. Baumann, What Does it Take to Enforce an Argument? Minimal Change in abstract Argumentation, Proc. of ECAI

43 Selected references (9) Extensions of Dung s model L. Amgoud, C. Cayrol, A reasoning model based on the production of acceptable arguments. Annals of Mathematics and Artificial Intelligence 34, pp , Leila Amgoud, Claudette Cayrol, Marie-Christine Lagasquie-Schiex, P. Livet: On bipolarity in argumentation frameworks. Int. J. Intell. Syst. 23(10): , P. Baroni, F. Cerutti, M. Giacomin, G. Guida, AFRA: Argumentation Framework with Recursive Attacks. International Journal of Approximate Reasoning, vol. 52, iss. 1, pp , H. Barringer, D. M. Gabbay, J. V. Woods, Temporal dynamics of support and attack networks: From argumentation to zoology, in: D. Hutter, W. Stephan (Eds.), Mechanizing Mathematical Reasoning, Springer, 2005, Trevor J. M. Bench-Capon: Persuasion in Practical Argument Using Value-based Argumentation Frameworks. Journal of Logic and Computation 13(3), 2003,

44 Selected references (10) Extensions of Dung s model G. Boella, D. M. Gabbay, L. van der Torre, S. Villata, Meta-argumentation modelling I: Methodology and techniques, Studia Logica 93 (2/3) (2009) D. Gabbay, Semantics for higher level attacks in extended argumentation frames Part 1: Overview, Studia Logica 93 (2/3) (2009) D. Gabbay, Fibring argumentation frames, Studia Logica 93 (2/3) (2009) S. Modgil and T.J.M Bench-Capon. Integrating Object and Meta-level Value Based Argumentation. In: 2nd International Conference on Computational Models of Argument (COMMA 08), pp , Toulouse, France, May S. Modgil. Reasoning about preferences in argumentation frameworks. Artificial Intelligence, 173: , Nielsen, S. H., and S. Parsons, A Generalization of Dung s Abstract Framework for Argumentation: Arguing with Sets of Attacking Arguments, Argumentation in Multi- Agent Systems, LNCS Vol 4766, Springer, 2007, pp