argumentation in artificial intelligence henry prakken lissabon, portugal december 11, 2009

46
Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Upload: lesley-paul

Post on 17-Dec-2015

213 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Argumentation in Artificial Intelligence

Henry PrakkenLissabon, Portugal

December 11, 2009

Page 2: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Why do agents need argumentation?

For their internal reasoning To draw conclusions given conflicting arguments

For their interaction with other agents To persuade given a conflict of opinion

Page 3: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Toulmin’s argument scheme

ClaimDataSo

Warrant

Since

Backing

On account of

Rebuttal

Unless

Page 4: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Toulmin’s argument scheme

João is catholicJoão is

Portuguese

So

Most Portugueseare catholic

Since

statistics

On account of

João often visits aprotestant church

Unless

Page 5: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Toulmin’s argument scheme

João is PortugueseJoão was Born

In Portugal

So

Who is born inPortugal isPortuguese

Since

Portuguese law

On account of

João adopted another

nationality

Unless

Page 6: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

From Toulmin to modern argumentation theory

Toulmin’s (1958) main contributions: Arguments can be defeated Validity of arguments is procedural (and

field-dependent?) This led to the idea of argument(ation)

schemes. An argument is acceptable if: it instantiates an argument scheme the critical questions asked in dialogue can

be answered

Page 7: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Argument(ation) schemes: general form

But also critical questions

Premise 1, … , Premise nTherefore (presumably), conclusion

Page 8: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Expert testimony(Walton 1996)

Critical questions: Is E really expert on D? Did E really say that P? Is P really within D? Is E biased? Is P consistent with what other experts say? Is P consistent with known evidence?

E is expert on DE says that PP is within D Therefore (presumably), P is the case

Page 9: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Witness testimony

Critical questions: Is W sincere? Does W’s memory function properly? Did W’s senses function properly?

W says PW was in the position to observe PTherefore (presumably), P

Page 10: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Arguments from consequences

Critical questions: Does A also have bad consequences? Are there other ways to bring about G? ...

Action A brings about G, G is goodTherefore (presumably), A should be done

Page 11: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Three layers in argumentation Logic layer

Fixed theory Procedural layer

Dynamic theory Strategic layer

Dynamic theory

My point: even logic is partly dialectic

Page 12: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

We should lower taxes

Lower taxes increase productivity

Increased productivity is good

Page 13: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

We should lower taxes

Lower taxes increase productivity

Increased productivity is good

We should not lower taxes

Lower taxes increase inequality

Increased inequality is bad

Page 14: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

We should lower taxes

Lower taxes increase productivity

Increased productivity is good

We should not lower taxes

Lower taxes increase inequality

Increased inequality is bad

Lower taxes do not increase productivity

USA lowered taxes but productivity decreased

Page 15: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

We should lower taxes

Lower taxes increase productivity

Increased productivity is good

We should not lower taxes

Lower taxes increase inequality

Increased inequality is bad

Lower taxes do not increase productivity

Prof. P says that …

USA lowered taxes but productivity decreased

Page 16: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

We should lower taxes

Lower taxes increase productivity

Increased productivity is good

We should not lower taxes

Lower taxes increase inequality

Increased inequality is bad

Lower taxes do not increase productivity

Prof. P says that …

Prof. P has political ambitions

People with political ambitions are not objective

Prof. P is not objective

USA lowered taxes but productivity decreased

Page 17: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

We should lower taxes

Lower taxes increase productivity

Increased productivity is good

We should not lower taxes

Lower taxes increase inequality

Increased inequality is bad

Lower taxes do not increase productivity

Prof. P says that …

Prof. P has political ambitions

People with political ambitions are not objective

Prof. P is not objective

USA lowered taxes but productivity decreased

Page 18: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

We should lower taxes

Lower taxes increase productivity

Increased productivity is good

We should not lower taxes

Lower taxes increase inequality

Increased inequality is bad

Lower taxes do not increase productivity

Prof. P says that …

Prof. P has political ambitions

People with political ambitions are not objective

Prof. P is not objective

Increased inequality is good

Increased inequality stimulates competition

Competition is good

USA lowered taxes but productivity decreased

Page 19: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

We should lower taxes

Lower taxes increase productivity

Increased productivity is good

We should not lower taxes

Lower taxes increase inequality

Increased inequality is bad

Lower taxes do not increase productivity

Prof. P says that …

Prof. P has political ambitions

People with political ambitions are not objective

Prof. P is not objective

Increased inequality is good

Increased inequality stimulates competition

Competition is good

USA lowered taxes but productivity decreased

Page 20: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

A B

C D E

1. An argument is In iff all arguments defeating it are Out.2. An argument is Out iff it is defeated by an argument that is In.

Dung 1995

Grounded semantics minimises node colouring Preferred semantics maximises node colouring

Page 21: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Argument game for grounded semantics

Rules of the game: Proponent starts with an argument Then each player defeats the previous move of the

other player Proponent moves strict defeaters, opponent moves

defeaters Proponent does not repeat his moves

A player wins iff the other player cannot move

Result: A is in the grounded extension iff proponent has a winning strategy in a game about A.

Page 22: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

A defeat graph

A

B

C

D

E

F

Page 23: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

A game tree

P: AA

B

C

D

E

F

move

Page 24: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

A game tree

P: AA

B

C

D

E

F

O: F

move

Page 25: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

A game tree

P: AA

B

C

D

E

F

O: F

P: E

move

Page 26: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

A game tree

P: A

O: B

A

B

C

D

E

F

O: F

P: E

move

Page 27: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

A game tree

P: A

O: B

P: C

A

B

C

D

E

F

O: F

P: E

move

Page 28: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

A game tree

P: A

O: B

P: C

O: D

A

B

C

D

E

F

O: F

P: E

move

Page 29: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

A game tree

P: A

O: B

P: C P: E

O: D

A

B

C

D

E

F

O: F

P: E

move

Page 30: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

We should lower taxes

Lower taxes increase productivity

Increased productivity is good

We should not lower taxes

Lower taxes increase inequality

Increased inequality is bad

Lower taxes do not increase productivity

Prof. P says that …

Prof. P has political ambitions

People with political ambitions are not objective

Prof. P is not objective

Increased inequality is good

Increased inequality stimulates competition

Competition is good

USA lowered taxes but productivity decreased

Page 31: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Structured arguments

Argument structure: Trees where

Nodes are formulas of a logical language L Links are applications of inference rules

Rs = Strict rules (1, ..., 1 ); or Rd= Defeasible rules (1, ..., 1 )

Reasoning starts from a knowledge base K L

Defeat: attack on conclusion, premise or defeasible inference, + preferencesArgument acceptability: Dung (1995)

Page 32: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Argument(ation) schemes: general form

Defeasible inference rules! But also critical questions

Negative answers are counterarguments

Premise 1, … , Premise nTherefore (presumably), conclusion

Page 33: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Expert testimony(Walton 1996)

Critical questions: Is E really expert on D? Did E really say that P? Is P really within D? Is E biased? Is P consistent with what other experts say? Is P consistent with known evidence?

E is expert on DE says that PP is within D Therefore (presumably), P is the case

Page 34: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

We should lower taxes

Lower taxes increase productivity

Increased productivity is good

We should not lower taxes

Lower taxes increase inequality

Increased inequality is bad

Lower taxes do not increase productivity

Prof. P says that …

Prof. P has political ambitions

People with political ambitions are not objective

Prof. P is not objective

Increased inequality is good

Increased inequality stimulates competition

Competition is good

USA lowered taxes but productivity decreased

Page 35: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Three layers in argumentation Logic layer

Fixed theory Procedural layer

Dynamic theory Strategic layer

Dynamic theory

Page 36: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Interaction Argument games verify status of

argument (or statement) given a single theory (knowledge base)

But real argumentation dialogues have Distributed information Dynamics Real players!

Page 37: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

A ‘real’ argumentation dialogue I claim that we should lower taxes Why? Since lower taxes increase productivity, which is

good I disagree. We should not lower taxes, since that

would increase inequality, which is bad. Besides, lower taxes will not increase

productivity Why not? Since the USA recently lowered their taxes but

productivity decreased. OK, I admit that lower taxes do not always

increase productivity; I retract my claim.

Page 38: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Dialogue systems (according to Carlson 1983)

Dialogue systems define the conditions under which an utterance is appropriate

An utterance is appropriate if it furthers the goal of the dialogue in which it is made

Appropriateness defined not at speech act level but at dialogue level

Dialogue game approach Protocol should promote the goal of the dialogue

Page 39: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Dialogue game systems A communication language

Well-formed utterances Rules for when an utterance is

allowed Protocol

Turntaking rules Termination rules

Page 40: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Dialogical aspects of argument schemes

Some critical questions ask “why this premise?”

Other critical questions ask “is there no exception?” But burden of proof is on respondent to

show that there are exceptions!

Dialogue systems should allow for counterarguments

Page 41: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Need for other speech acts(and for rhetoric)

Paul: r

Olga: s

p qr ps r

Knowledge bases Inference rules

P1: q since p

Page 42: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Need for other speech acts(and for rhetoric)

Paul: r

Olga: s

Knowledge bases Inference rules

P1: q since p

O1: why p?

p qr ps r

Page 43: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Need for other speech acts(and for rhetoric)

Paul: r

Olga: s

Knowledge bases Inference rules

P1: q since p

O1: why p?

P2: p since r

p qr ps r

Page 44: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Need for other speech acts(and for rhetoric)

Paul: r

Olga: s

Knowledge bases Inference rules

P1: q since p

O1: why p?

O2: r since s

P2: p since r

p qr ps r

Page 45: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Some properties that can be studied

Correspondence with participants’ beliefs If union of beliefs implies p, can/will

agreement on p result? If participants agree on p, does union of

beliefs imply p? Disregarding vs. assuming agent

strategies/tactics In general it will be hard to enforce agreement

Page 46: Argumentation in Artificial Intelligence Henry Prakken Lissabon, Portugal December 11, 2009

Conclusions Argumentation theory can benefit from AI

Formalisation Computer models Computer tools

AI can benefit from argumentation theory Concepts Theories …