1 / 62

Spring School on Argumentation in AI & Law Day 2 – lecture 2 Case-based Reasoning

Spring School on Argumentation in AI & Law Day 2 – lecture 2 Case-based Reasoning. Henry Prakken Guangzhou (China) 11 April 2018. Overview. Introduction to AI & Law Case-based reasoning: HYPO, CATO & beyond. Part 1: Introduction to AI & Law. Some history on symbolic AI (& Law).

tmoney
Download Presentation

Spring School on Argumentation in AI & Law Day 2 – lecture 2 Case-based Reasoning

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Spring School on Argumentation in AI & LawDay 2 – lecture 2Case-based Reasoning Henry Prakken Guangzhou (China) 11 April 2018

  2. Overview • Introduction to AI & Law • Case-based reasoning: HYPO, CATO & beyond

  3. Part 1: Introduction to AI & Law

  4. Some history on symbolic AI (& Law) •  1950 - 1970: modelling general intelligence • Newel & Simon’s General Problem Solver •  1970 - 1990: modelling expertise in limited domains • ‘Expert systems’, later ‘knowledge-based systems’ • Knowledge about a problem domain • Reasoning mechanisms for solving decision problems • E.g. MYCIN (diagnosis and treatment of infection diseases) • Since  1980: optimism about legal applications: • Model the rules in logic, reason logically

  5. “Vehicles are not allowed in the park” • Facts: evidence problems • Legal conditions: general terms • Legal rules: exceptions • Purpose of the rule • Principle “Vehicles are objects meant for normal transport”

  6. Legal reasoning is adversarial • Legal reasoning forms leave room for doubt • Legal cases involve clashes of interest Dispute • study constructingandattacking arguments

  7. Knowledge-based AI for decision support • Knowledge-based systems: • Knowledge about a problem domain • Reasoning mechanisms for solving decision problems • Legal decision problems • Deciding court cases • Or arguing for decisions • Processing legislation in public administration • Giving legal advice • On litigation • On documents • On legal constructions • … • Regulatory compliance • …

  8. Legal knowledge-based systems in practice • Quite a few rule-based systems in public administration • Don’t automate legal reasoning, but automate the logic of regulations • Proof of facts, classification and interpretation left to user • Until recently almost non-existent in court or advocacy • But recent developments, e.g. NeotaLogic

  9. Limitations of legal rule-based systems • No handling of exceptions and rule conflicts • No cases • No argumentation

  10. AI & Law research on legal argument • Legal reasoning as argumentation • Inference by constructing and comparing arguments and counterarguments • Applied to: • Determining the facts of a case (see tomorrow) • Legally classifying facts (today) • Applying legal rules to facts (see yesterday)

  11. AI & Law models of legal argument • Systems for rule-based reasoning with rule exceptions and conflicts ... (Prakken & Sartor, Hage & Verheij, ...) • Systems for case-based reasoning: HYPO, CATO, … • Combining rule and cases: CABARET, … • Follow-up work (mostly formal): Bench-Capon, Gordon, Horty, Prakken & Sartor, Verheij, ... • But not yet applied in practice • Commonsense, empathy, fairness, … • Two overview papers: • H. Prakken, Legal reasoning: computational models. In J.D. Wright (ed.): International Encyclopedia of the Social and Behavioural Sciences, 2nd edition. Elsevier Ltd, Oxford, 2015. • H. Prakken & G. Sartor, Law and logic: a review from an argumentation perspective. Artificial Intelligence 227 (2015): 214-245.

  12. New developments • Deep learning, natural-language processing, ubiquitous computing, big data • ML-techniques can now be applied at a large scale to unstructured data • IBM’s Watson: ‘knowledge-based’ system with access to huge amounts of unstructured information • Applied in medicine, financial trading, … • Now also in the law: ROSS (Un. Toronto) • http://www.rossintelligence.com/

  13. IBM’s Watson static.guim.co.uk

  14. ‘Data centric’ models (1) • Require no model of legal reasoning or decision making • But require `task models’ of what lawyers do in their daily work L.K. Branting, Data-centric and logic-based models for automated legal problem solving. Artificial Intelligence and Law, 25(1):5–27, 2017.

  15. ‘Data centric’ models (2) • At case level: • Predicting probability of success • lexmachina.com • lexpredict.com • premonition.com • Partly based on factors unrelated to the merit of a case (court, judge, jurisdiction, parties, attorneys, …) • ... L.K. Branting, Data-centric and logic-based models for automated legal problem solving. Artificial Intelligence and Law, 25(1):5–27, 2017.

  16. ‘Data centric’ models (3) • At document level: • Information extraction • Persons, organisations, claims, outcomes … • Automated summarisation • legalrobot.com • casetext.com • Parsing statutory tekst • … L.K. Branting, Data-centric and logic-based models for automated legal problem solving. Artificial Intelligence and Law, 25(1):5–27, 2017.

  17. ‘Data centric’ models (4) • At corpus level: • Network analysis • Case citations, regulations, … • Judicial database analysis • Improving court management • Error detection • … L.K. Branting, Data-centric and logic-based models for automated legal problem solving. Artificial Intelligence and Law, 25(1):5–27, 2017.

  18. Hybrid models • Argument mining • Mining (elements of) arguments • Guided by a model of argumentation • … • Watson’s debater function • Can this provide the input of advanced legal knowledge-based systems? K.D. Ashley, Artificial Intelligence and Legal Analytics. New Tools for Law Practice in the Digital Age. Cambridge University Press 2017.

  19. State of the art • Simple rule-based systems: • in routine use in public administration • No model of legal argument • Advanced knowledge-based systems for legal argument: • much theory and proof of concept • Not scalable • Data-centric applications • Many practical applications • Not suitable for legal argument / decision making • Hope for the future: hybrid approaches

  20. Part 2: Modelling legal classification and interpretation in factor-based domains

  21. Factor-based reasoning • In legal classification and interpretation there are often no clear rules (not even clear defeasible rules) • Often there only are factors: tentative reasons pro or con a conclusion • To draw a conclusion, the sets of all applicable factors pro and con should be compared • This is done in cases, which become precedents • New cases should be decided according to procedent (stare decisis) • Problem: new cases are often not identical to a precedent

  22. Running example factors: misuse of trade secrets • Some factors pro misuse of trade secrets: • F2 Bribe-Employee • F4 Agreed-Not-To-Disclose • F6 Security-Measures • F15 Unique-Product • F18 Identical-Products • F21 Knew-Info-Confidential • Some factors con misuse of trade secrets: • F1 Disclosure-In-Negotiations • F16 Info-Reverse-Engineerable • F23 Waiver-of-Confidentiality • F25 Info-Reverse-Engineered HYPO Ashley & Rissland 1985-1990 CATO Aleven & Ashley 1991-1997

  23. Statistical techniques, machine learning … • Only useful if: • The relevant factors are known • Many precedents are available • The precedents are largely consistent with each other • Useful for predicting outcomes • But no model of legal reasoning • Black box • No argumentation

  24. HYPO Ashley & Rissland 1987-1990 Kevin Ashley & Edwina Rissland • Representation language: • Cases: decision (p or d) + p-factors and d-factors • Current Fact Situation: factors • Arguments: • Citing (for its decision) a case on its similarities with CFS • Distinguishing a case on its differences with CFS • Taking into account which side is favoured by a factor

  25. HYPO Example C1 (p) C2 (d) CFS • p1 p2 p3 • d1 d2 • p1 p2 p3 • d1 d2 • p1 p4 • d2 p2 p3 p4 d2 d3

  26. HYPO Example C1 (p) C2 (d) CFS • p1 p2 p3 • d1 d2 • p1p2 p3 • d1 d2 • p1 p4 • d2 Distinguish! p2 p3 p4 d2d3 Distinguish!

  27. HYPO Example C1 (p) C2 (d) CFS • p1 p2 p3 • d1 d2 • p1 p2 p3 • d1 d2 • p1 p4 • d2 p2 p3 p4 d2 d3

  28. HYPO Example C1 (p) C2 (d) CFS • p1 p2 p3 • d1 d2 • p1 p2 p3 • d1 d2 • p1 p4 • d2 p2 p3 p4 d2 d3 Distinguish!

  29. HYPO’s argument game • Given: a case base and a current fact situation • Plaintiff starts with citation • A case decided for plaintiff and sharing pro-plaintiff factors with the CFS • Defendant: • cites all counterexamples (cases citable for defendant) • distinguishes citation in all possible ways • On pro-plaintiff factors of precedent lacking in CFS • On new pro-defendant factors in the CFS • Plaintiff distinguishes defendant’s counterexamples in all possible ways

  30. Citing precedent • Mason v Jack Daniels Distillery (Mason) – undecided. • F1 Disclosure-In-Negotiations (d) • F6 Security-Measures (p) • F15 Unique-Product (p) • F16 Info-Reverse-Engineerable (d) • F21 Knew-Info-Confidential (p) • Bryce and Associates v Gladstone (Bryce) – plaintiff • F1 Disclosure-In-Negotiations (d) • F4 Agreed-Not-To-Disclose (p) • F6 Security-Measures (p) • F18 Identical-Products (p) • F21 Knew-Info-Confidential (p)

  31. Citing precedent • Mason v Jack Daniels Distillery (Mason) – undecided. • F1 Disclosure-In-Negotiations (d) • F6 Security-Measures (p) • F15 Unique-Product (p) • F16 Info-Reverse-Engineerable (d) • F21 Knew-Info-Confidential (p) • Bryce and Associates v Gladstone (Bryce) – plaintiff • F1 Disclosure-In-Negotiations (d) • F4 Agreed-Not-To-Disclose (p) • F6 Security-Measures (p) • F18 Identical-Products (p) • F21 Knew-Info-Confidential (p) Plaintiff cites Bryce because of F6,F21

  32. Distinguishing precedent • Mason v Jack Daniels Distillery (Mason) – undecided. • F1 Disclosure-In-Negotiations (d) • F6 Security-Measures (p) • F15 Unique-Product (p) • F16 Info-Reverse-Engineerable (d) • F21 Knew-Info-Confidential (p) • Bryce and Associates v Gladstone (Bryce) – plaintiff • F1 Disclosure-In-Negotiations (d) • F4 Agreed-Not-To-Disclose (p) • F6 Security-Measures (p) • F18 Identical-Products (p) • F21 Knew-Info-Confidential (p) Plaintiff cites Bryce because of F6,F21 Defendant distinguishes Bryce because of F4,F18 and F16

  33. Counterexample • Mason v Jack Daniels Distillery – undecided. • F1 Disclosure-In-Negotiations (d) • F6 Security-Measures (p) • F15 Unique-Product (p) • F16 Info-Reverse-Engineerable (d) • F21 Knew-Info-Confidential (p) • Robinson v State of New Jersey – defendant. • F1 Disclosure-In-Negotiations (d) • F10 Secrets-Disclosed-Outsiders (d) • F18 Identical-Products (p) • F19 No-Security Measures (d) • F26 Deception (p) Defendant cites Robinson because of F1

  34. Distinguishing counterexample • Mason v Jack Daniels Distillery – undecided. • F1 Disclosure-In-Negotiations (d) • F6 Security-Measures (p) • F15 Unique-Product (p) • F16 Info-Reverse-Engineerable (d) • F21 Knew-Info-Confidential (p) • Robinson v State of New Jersey – defendant. • F1 Disclosure-In-Negotiations (d) • F10 Secrets-Disclosed-Outsiders (d) • F18 Identical-Products (p) • F19 No-Security Measures (d) • F26 Deception (p) Defendant cites Robinson because of F1 Plaintiff distinguishes Robinson because of F6,F15,F21 and F10,F19

  35. K.D. Ashley. Modeling Legal Argument: Reasoning with Cases and Hypotheticals. MIT Press, Cambridge, MA, 1990. Plaintiff: I should win because My case shares pro factors F6 and F21 with Bryce, which was won by plaintiff Defendant: Unlike the present case, Bryce had pro factors F4 and F18 Defendant: UnlikeBryce, the present case has con factor F16 Defendant: I should win becausemy case shares con factor F1 with Robinson, which was won bydefendant Plaintiff: Unlike the present case, Robinson had con factors F10 and F19 Plaintiff: Unlike Robinson, the present case has pro factors F6, F15 and F21

  36. Adam Wyner Trevor Bench-Capon & Katie Atkinson A logical account of case-based reasoning in factor-based domains H. Prakken, A. Wyner, T. Bench-Capon & K. Atkinson, A formalisation of argumentation schemes for legal case-based reasoning in ASPIC+. Journal of Logic and Computation 25 (2015): 1141-1166.

  37. Basic scheme for reasoning with two-valued factors AS1: ThePro-factors of current are P TheCon-factors of current are C P arepreferred over C Current should be decided Pro ThePro-factors of current are P TheCon-factors of current are C C arepreferred over P Current should be decided Con

  38. Preferences from precedents (1) AS2: ThePro-factors of precedent are P TheCon-factors of precedent are C precedent was decided Pro P arepreferred over C Limitation 1: the current case will often not exactly match a precedent

  39. A fortiori reasoning with two-valued factors AS3: P arepreferred over C P+arepreferred over C- Limitation 2: not all differences with a precedent will make a current case stronger P+ = P plus zero or more additional pro-factors C- = C minus zero or more con factors

  40. Vincent Aleven 1991-1997 (snapshot of)CATO Factor Hierarchy Misuse of Trade Secret (p) F120: Info legitimately obtained elsewhere (d) F101: Info Trade Secret (p) F104: Info valuable (p) F102: Efforts to maintain secrecy (p) F4: Agreed not to disclose (p) F1: Disclosures in negotiations (d) F6: Security measures (p) F15: Unique product (p)

  41. V. Aleven. Using background knowledge in case-based legal reasoning: a computational model and an intelligent learning environment. ArtificialIntelligence 150:183-237, 2003. Distinguishing Misuse of Trade Secret (p) F120: Info legitimately obtained elsewhere (d) F101: Info Trade Secret (p) F104: Info valuable (p) F102: Efforts to maintain secrecy (p) F4: Agreed not to disclose (p) F1: Disclosures in negotiations (d) F6: Security measures (p) F15: Unique product (p)

  42. Emphasising distinctions Misuse of Trade Secret (p) F120: Info legitimately obtained elsewhere (d) F101: Info Trade Secret (p) F104: Info valuable (p) F102: Efforts to maintain secrecy (p) F4: Agreed not to disclose (p) F1: Disclosures in negotiations (d) F6: Security measures (p) F15: Unique product (p)

  43. Downplaying distinctions Misuse of Trade Secret (p) F120: Info legitimately obtained elsewhere (d) F101: Info Trade Secret (p) F104: Info valuable (p) F102: Efforts to maintain secrecy (p) F4: Agreed not to disclose (p) F1: Disclosures in negotiations (d) F6: Security measures (p) F15: Unique product (p)

  44. Exploiting factor hierarchies (1):current misses pro factor AS4: P1 arepreferred over C P2 substitutesP1 P2 arepreferred over C Def1: Factor set P2 substitutesfactor set P1 iff For all factors p1 in P1 that are not in P2 there exists a factor p2 in P2 that substitutesp1 Def2: Factor p2 substitutes factor p1 iff p1 instantiates abstract factor p3 and p2 instantiates abstract factor p3

  45. Current should be decided Pro ThePro-factors of Current are {F6,F21} {F6,F21} > {F1} TheCon-factors of Current are {F1} {F4,F21} > {F1} {F6,F21} substitutes {F4,F21} ThePro-factors of Precedent are {F4,F21} Precedent was decided Pro TheCon-factors of Precedent are {F1} F6 substitutes F4 F4 instantiates F102 F6 instantiates F102

  46. From two-valued to many-valued factors (dimensions) • Dimensions can have a value from an ordered range of values • Numbers • Anything else that can be ordered • Notation: (dimension,value) or (d,v) • Dimensions have polarities: < con pro 0,1,2,…. .…, 500, ….... Primary school, secondary school, Bsc, Msc, Dr

  47. Example dimensions in HYPO • Number of disclosees (0,1,….) • Competetive advantage (none, weak, moderate, strong) < pro con 0 1 2 3 4 5, …....

  48. Example dimensions in HYPO • Number of disclosees (0,1,….) • Competetive advantage (none, weak, moderate, strong) < con pro none weak moderate strong

  49. A fortiori reasoning with dimensions AS6: P1 arepreferred over C1 P2 are at least as strong as P1 C1 are at least as strong as C2 P2 arepreferred over C2 • Def5: • Set P2 of dimension-value pairs pro is at least as strong as set P1 of dimension-value pairs proiff • For all pairs (d,v1)in P1 there exists a pair (d,v2)in P2 such that v1 ≤ v2 • Set C1 of dimension-value pairs con is at least as strong as set C2 of dimension-value pairs con iff • For all pairs (d,v2)in C2 there exists a pair (d,v1)in C1 such that v1 ≤ v2

  50. Example with dimensions (1) • Precedent – defendant • F1 Disclosure-In-Negotiations (d) • F21 Knew-Info-Confidential (p) • Fx Competetive-advantage = strong (p) • Fy Number of disclosees = 10 (d) • New case – undecided • F1 Disclosure-In-Negotiations (d) • F21 Knew-Info-Confidential (p) • Fx’ Competetive-advantage = moderate (p) • Fy’ Number of disclosees = 6 (d) {F21, Fx} < {F1,Fy} because of precedent Defendant wants to argue that {F21, Fx’} < {F1,Fy’}

More Related