commonsense reasoning and argumentation 13/14 hc 14: dialogue systems for argumentation (2)

Commonsense Reasoning and Argumentation 17/18

HC 14:Legal argumentation (2)

Concluding remarks

Henry Prakken4 April 2018

(snapshot of)CATO Factor

Hierarchy

F101: Info Trade Secret (p)

F102: Efforts to maintain secrecy (p)

F104: Info valuable (p)

F4: Agreed not to disclose (p)

F1: Disclosuresin negotiations (d)

F6: Securitymeasures (p)

F15: Unique product (p)

Misuse of Trade Secret (p)

F120: Info legitimatelyobtained elsewhere (d)

Vincent Aleven 1991-1997

Distinguishing










V. Aleven. Using background knowledge in case-based legal reasoning: a computational model and an intelligent learning environment. Artificial Intelligence 150:183-237, 2003.

Emphasising distinctions










Downplaying distinctions










Exploiting factor hierarchies (1):current misses pro factor

AS4:P1 are preferred over CP2 substitutes P1P2 are preferred over C

Def1:Factor set P2 substitutes factor set P1 iff

For all factors p1 in P1 that are not in P2 there exists a factor p2 in P2 that substitutes p1

Def2:Factor p2 substitutes factor p1 iff

p1 instantiates abstract factor p3 andp2 instantiates abstract factor p3

Current should be decided Pro

The Pro-factors of Current are {F6,F21}

The Con-factors of Current are {F1}

{F6,F21} > {F1}

{F4,F21} > {F1}

{F6,F21} substitutes {F4,F21}

F4 instantiates F102

F6 instantiates

F102

F6 substitutes F4

The Pro-factors of Precedent

are {F4,F21}

The Con-factors of Precedent are {F1}

Precedent was

decided Pro

With alternative to

U2.2.1

Distinguishing


F102: Efforts to maintain Secrecy (p)



F1: DisclosuresIn negotiations (d)





Emphasising distinctions










Downplaying distinctions










Exploiting factor hierarchies (2):current has additional con factor

AS5:P are preferred over CP cancels C+P are preferred over C+

Def3:Factor set P cancels factor set C iff

For all factors c1 in C+ that are not in C there exists a factor p1 in P such that p1 cancels c1

Def4:Factor p1 cancels factor c1 iff

p1 instantiates abstract factor p2 andc1 is con abstract factor p2 andp1 is preferred over c1

Example of cancellation Precedent – plaintiff

F15 Unique-Product (p) F21 Knew-Info-Confidential (p) F120 Info-Legitimately obtained elsewhere (d)

New case – undecided F1 Disclosure-In-Negotiations (d) F4 Agreed-Not-To-Disclose (p) F15 Unique-Product (p) F21 Knew-Info-Confidential (p) F120 Info-Legitimately obtained elsewhere (d)

{F15,F21} > {F120} because of precedent

Plaintiff wants to argue that {F4,F15,F21} >

{F1,F120}

Current should be decided Pro

The Pro-factors of Current are

{F4,F15,F21}

The Con-factors of Current are {F1,F120}

{F4,F15, F21} > {F1,F120}

{F15,F21} > {F120}

{F4,F15,F21} cancels {F1,F120}

F4 instantiates

F102

F1 is con F102

F4 cancels F1The Pro-factors of Precedent

are {F15,F21}

The Con-factors of Precedent

are {F120}

Precedent was

decided Pro

{F4,F15,F21} > {F120}

F4 > F1

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

<

Example dimensions in HYPO

Number of disclosees (0,1,….) Competetive advantage (none, weak,

moderate, strong)

pro con

0 1 2 3 4 5, …....

<

Example dimensions in HYPO

Number of disclosees (0,1,….) Competetive advantage (none, weak,

moderate, strong)

con pro

none weak moderate strong

<

A fortiori reasoning with dimensions

AS6:P1 are preferred over C1P2 are at least as strong as P1C1 are at least as strong as C2P2 are preferred over C2

Def5:Set P2 of dimension-value pairs pro is at least as strong as set P1 of dimension-value pairs pro iff

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

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’}

Example with dimensions (2)

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’}

What if the previous schemes do not apply?

Which decisions are allowed by a body of precedents? Precedential constraint

Where do preferences then come from?

Precedential constraint:consistency of preferences

A preference relation < on factor sets is consistent if and only if there are no factor sets X and Y such that both X < Y and Y < X.

Preferences from precedents

AS2:The Pro-factors of precedent are PThe Con-factors of precedent are Cprecedent was decided ProP are preferred over C

A fortiori reasoning with two-valued factors

AS3:P are preferred over CP+ are preferred over C-

P+ = P plus zero or more additional pro-factorsC- = C minus zero or more con factors

Precedential constraint:allowed and forced decisions

Let < be determined by: A set S of precedents The preferences derivable from it by

AS2 (preferences from precedent) AS3 (a fortiori for two-valued factors)

Assume < is consistent Then a decision pro in a new case C is:

allowed by S iff adding C with decision pro to S leaves < consistent.

forced iff allowed and adding C with decision con to S makes < inconsistent

Following, distinguishing and overruling precedents

Let Prec have pro factors P and con factors C and decision pro.

Let Curr have pro factors Pcurr such that P is included in Pcurr. Following Prec = deciding Curr pro Distinguishing Prec = deciding Curr con

where deciding Curr either pro or con is allowed

Overruling Prec = deciding Curr con where deciding Curr pro is forced

Example (1) Precedent – plaintiff

F1 Disclosure-In-Negotiations (d) F6 Security-Measures (p) F21 Knew-Info-Confidential (p) F23 Waiver-of-Confidentiality (d)

New Case – undecided F21 Knew-Info-Confidential (p) F23 Waiver-of-Confidentiality (d) F25 Info-Reverse-Engineered (d)

{F6,F21} > {F1,F23}

Pro = {F21} > {F23,F25}

Con = {F21} < {F23,F25}

Both pro and con allowed

Deciding pro follows precedentDeciding con distinguishes precedent

Example (2) Precedent 1 – plaintiff

F1 Disclosure-In-Negotiations (d) F6 Security-Measures (p) F21 Knew-Info-Confidential (p) F23 Waiver-of-Confidentiality (d)

Precedent 2 – defendant F6 Security-Measures (p) F21 Knew-Info-Confidential (p) F25 Info-Reverse-Engineered (d)

New Case – undecided F21 Knew-Info-Confidential (p) F23 Waiver-of-Confidentiality (d) F25 Info-Reverse-Engineered (d)

{F6,F21} > {F1,F23}

{F6,F21} < {F25}

Pro = {F21} > {F23,F25}

Con = {F21} < {F23,F25}

Only con allowed

Deciding pro overrules precedent 2Deciding con follows precedent 2

What if the previous schemes do not apply?

Which decisions are allowed by a body of precedents? Precedential constraint

Where do preferences then come from?

Basic scheme for value-based reasoning with

precedents

Deciding current pro promotes set of values V1 Deciding current con promotes set of values V2 V1 is preferred over V2Therefore, current should be decided pro

Deciding case pro when it contains P promotes value V

D.H. Berman and C.D. Hafner. Representing teleological structure in case-based legal reasoning: the missing link. In Proceedings of the Fourth International Conference on Articial Intelligence and Law, pages 50-59, New York, 1993. ACM Press.

Scheme for inferring value orderings from cases

Deciding precedent pro promotes set of values V1 Deciding precedent con promotes set of values V2precedent was decided pro Therefore, V1 is preferred over V2

Wild animals example

Pierson v Post: Plaintiff is hunting a fox on open land. Defendant kills the fox.

Keeble v Hickersgill: Plaintiff is a professional hunter. Lures ducks to his pond. Defendant scares the ducks away

Young v Hitchens: Plaintiff is a professional fisherman. Spreads his nets. Defendant gets inside the nets and catches the fish.

Slide by Trevor Bench-Capon

Pierson – defendant NotDefLiv: Defendant not pursuing livelihood (p) NotPlLiv: Plaintiff not pursuing livelihood (d) NotOwnLand: Plaintiff not on own land (d) NotCaught: Plaintiff had not caught animal (d)

Keeble – plaintiff NotDefLiv: Defendant not pursuing livelihood (p) PlLiv: Plaintiff pursuing livelihood (p) OwnLand: Plaintiff on own land (p) NotCaught: Plaintiff had not caught animal (d)

Young – (defendant) DefLiv: Defendant pursuing livelihood (d) PlLiv: Plaintiff pursuing livelihood (p) NotOwnLand: Plaintiff not on own land (d) NotCaught: Plaintiff had not caught animal (d)

Factors in the wild animals cases

Con = {PlLiv} < {NotOwnLand,NotCaught,Def

Liv}

Pro = {PlLiv} > {NotOwnLand,NotCaught,Def

Liv}

{NotDefLiv} < {NotPlLiv,NotOwnLand

, NotCaught}

{NotDefLiv,PlLiv, OwnLand} > {NotCaught}

Values Cval: Certainty and avoidance of litigation Eval: Economic benefit for society Pval: respecting Property

From factors to values: Deciding pro when case contains PlLiv promotes Eval Deciding pro when case contains OwnLand promotes Pval Deciding pro when case contains Caught promotes Pval Deciding con when case contains NotCaught promotes Cval Deciding con when case contains DefLiv promotes Eval

Values in the wild animals cases

Pierson – defendant NotDefLiv: Defendant not pursuing livelihood (p) NotPlLiv: Plaintiff not pursuing livelihood (d) NotOwnLand: Plaintiff not on own land (d) NotCaught: Plaintiff had not caught animal (d) Cval

Keeble – plaintiff NotDefLiv: Defendant not pursuing livelihood (p) PlLiv: Plaintiff pursuing livelihood (p) Eval OwnLand: Plaintiff on own land (p) Pval NotCaught: Plaintiff had not caught animal (d) Cval

Young – (defendant) DefLiv: Defendant pursuing livelihood (d) Eval PlLiv: Plaintiff pursuing livelihood (p) Eval NotOwnLand: Plaintiff not on own land (d) NotCaught: Plaintiff had not caught animal (d) Cval

Values in the wild animals cases

{} < {Cval}

{Eval,Pval} > {Cval}

Pro = {Eval} > {Eval,Cval}

Con = {Eval} < {Eval,Cval}

Further refinements Promotion and demotion of values Degrees of promotion or demotion

Absolute or marginal Probability of promotion or

demotion

Law making as decision making

Olga Monge v. Beebe Rubber Company (1974)

Facts (1): Olga Monge, employed for an indefinite period of time

(“at will”), was fired by her foreman for no reason Law:

Every employment contract that specifies no duration is terminable at will by either party

If an employment contract is terminable at will, then the employee can be fired for any reason or no reason at all

Decision: Firing Olga Monge was a breach of contract

Facts (2): Monge claimed that she was fired since she had

refused to go out with the foreman

H. Prakken, Formalising debates about law-making proposals as practical reasoning. In M. Araszkiewicz & K. Pleszka (eds.): Logic in

the Theory and Practice of Lawmaking. Legisprudence Library Series, Springer 2015, 301-321.

Quotes from Monge In all employment contracts, whether at will or for a

definite term, the employer's interest in running his business as he sees fit must be balanced against the interest of the employee in maintaining his employment, and the public's interest in maintaining a proper balance between the two.

We hold that a termination by the employer of a contract of employment at will which is motivated by bad faith or malice or based on retaliation is not in the best interest of the economic system or the public good and constitutes a breach of the employment contract.

Monge as decision making

We should adopt the Old Rule as the valid

rule

The old rule makes that

employers can run their business

as they see fit

Employers being able to run their business as they see fit promotes individual liberty

We should adopt the New Rule as the valid

rule

The new rule makes that good

employees cannot be fired in malice

Good employees not being able to

be fired in malice promotes

the economic system and public good

Short for “Every employment contract that

specifies no duration is terminable at will

by either party”

Short for “Every employment contract that specifies no

duration is terminable at will by either party unless the

employer terminates the contract in bad faith, malice,

or retaliation”

Rule R

We should adopt rule R as the valid one

IF we should adopt rule R as the valid one THEN rule R

Supporting and using legal rules

Conditions of rule R

Conclusion of rule R

Every employment contract that specifies no duration is terminable at will by either

party

We should adopt the Old Rule as the valid one


Supporting and using legal rules in the Monge case (1)

Monge was employed at will

Monge’s contract can be terminated at will by Monge’s employer

Further arguments for and against this premise

“Every employment contract that specifies no duration is terminable at will by either party, unless the

employer terminates the contract in bad faith, malice, or

retaliation

We should adopt the New Rule as the valid one


Supporting and using legal rules in the Monge case (2)

Monge was employed at will

Monge’s contract cannot be terminated

at will by Monge’s employer

Further arguments for and against this premise

Monge was fired in malice

Monge as practical reasoning

We should adopt the Old Rule as the valid

rule

The old rule makes that

employers can run their business

as they see fit

Employers being able to run their business as they see fit promotes individual liberty

We should adopt the New Rule as the valid

rule

The new rule makes that good

employees cannot be fired in malice

Good employees not being able to

be fired in malice promotes

the economic system and public good

Basic scheme for value-based reasoning with precedents

Proposal 1 promotes set of values V1 Proposal 2 promotes set of values V2 V1 is preferred over V2Therefore, Proposal 1 should be adopted

How is this determined?

Basic scheme instantiated for Monge

Adopting the old rule promotes {Individual liberty} Adopting the new rule promotes {Economic system, public good}{Economic system, public good} is preferred over {Individual liberty}Therefore, the new rule should be adopted

Conclusions Modelling legal reasoning and decision making

as argumentation is a fruitful approach AI (& Law) has formal and computational tools

available But practical applications often face the

knowledge acquisition bottleneck

Topics for final Msc project (1)

Theoretical research on inference Defeasible vs. probabilistic reasoning

E.g. add probabilistic strengths to ASPIC+ A theory of preferences in argumentation Optimising argument games by utilising the

structure of arguments Dynamics in structured argumentation

Theoretical research on dialogue Dialogue protocols Agent strategies and personalities Argumentation in negotiation or deliberation Relation with game theory and judgement aggregation …

Topics for final Msc project (2)

Implementation Argument games, embedding theorem provers in

ASPIC+, dialogue protocols, dialogical agents … Apply a logic or implementation to a certain

problem Medical, legal, …

Experiment Effects of agent strategies or personalities, …

Simulation environment available! Recognising agent strategies or personalities Investigating whether humans reason according to

argumentation semantics

commonsense reasoning and argumentation 13/14 hc 14: dialogue systems for argumentation (2)

Documents