The Long and Winding Road 40 Years of Argumentation Trevor Bench-Capon University of Liverpool Comma 2012, Vienna Overview A review of my 40 years of engagement with argumentation: influences, lessons learned, and suggestions. 1972-78: Philosophy Maths is about proof, but Philosophy is about argument
Proofs are constructed, understood, admired Arguments are developed, engaged with, worried about Proofs are audience independent, context free Arguments have a time, a place and an audience Epistemology: Argument from Illusion, Transcendental Argument, Private Language Argument Philosophy of Religion: Ontological, Teleological, Cosmological Arguments Argument From Illusion
Consider a Stick in a Jar of Water What I see is bent But the stick is straight So I dont see what is there So I never experience the real world So I can never know anything about the real world The real world might not exist I might be the only person in the universe I might not even exist myself, just my perceptions Not convincing, but whats wrong with it? Arguments Dont Go Away
Arguments dont have a canonical form: each generation reinvents, reexpresses them, and falls under their spell again: The Argument from Illusion Appears again and again: different illusions, but the same doubts Two views of Philosophy Bertrand Russell said: The point of philosophy is to start with philosophy something so simple as not to seem as game worth stating, and to end with something so paradoxical that no one will believe it.
Wittgenstein said philosophy The aim of philosophy is to show the fly as obsession the way out of the fly bottle Arguments can amuse, or can entrap 1978-84: Civil Service Role of a UK Civil Servant (preThatcher) Given a policy question:
Find arguments for every option Allow politicians to choose Accept their choice Justify their choice to parliament Justify their choice to the public Lessons There are no good arguments: there are arguments that people accept Politicians will choose between arguments according to their aims, interests and aspirations greater equality is a pro reason for some, and a con reason for others The same policy needs different arguments for different people. We should cut taxes to: Let people keep what they earn (employers)
To increase prosperity, in which all will share (workers) 1984-87: Imperial College Logic programming was developed by mathematicians Program is a theory, a collection of axioms Computation is deductive proof Alternatively: Program is a collection of beliefs, conjectures and prejudices (possibly incompatible) Computation produces a set of arguments, perhaps both for and against, which can be taken or left
Closed World Assumption Logic Programming interprets failure as negation Completion of the Database makes it logically sound (if difficult Knowledge Based Systems are justto achieve except for Expert definitions)
Argument from Opinion? Argument from Ignorance makes it acceptable (subject to felicity conditions (critical questions)) Application to Law The logic program can been seen as following standard legal procedure Represent statutes, case law and (perhaps conflicting) opinion; Generate arguments for both sides of a case and Choose between them Bench-Capon and Sergot (1985). We need a representation in computer intelligible terms of what
it is that makes an argument persuasive Ive been more or less searching for this ever since 1987-2012: Liverpool Explanation If users are to choose between arguments they need them to be presented effectively Enthymemes, assumptions, shared background, ignoring rare exceptions Toulmins argumentation scheme gave a structure: for navigation basis for dialogue
Toulmins Scheme But does the rebuttal attack the Different kinds of attack: qualifier, compare critical questions of informal logic the conclusion and undercutters versus rebuttals in abstract the warrant or the backing? argument Roles for premises John is old if man(X) age(X,A) A > 75,
not has drunk the elixir of youth In current systems, (e.g. Carneades) Needs the sortal usually presupposed Notcommon really antoinequality Not really a statement distinguish Exception dogs get old
more quickly Trivial once A is bound Cant be false Assumptions: (John is a man) Not worth mentioning buildings
get old more slowly Really contains the knowledge Just there to bind A (John is aged over 75) Ordinary Premises: Unless the question arises Men are (normally) oldhad
when they are 75 of youth) Exceptions: (John drunk the elixir 1995 Dung - Context Dung published his work on Argumentation Frameworks in 1993 (IJCAI) and 1995 (AIJ). The most significant idea for this community Most important point (for me) is that
Arguments are evaluated in the context of other, related, arguments Arguments cannot be considered in isolation. They are accepted relative to the context created by other arguments Both attackers and defenders Sets of arguments stand (or fall) together 2000 onwards: Audiences Preferred semantics can give us multiple preferred extensions credulous (as well as sceptical) acceptance But how do we choose between them? The choice is not arbitrary:
think of politicians and judges Perelman: The acceptability of the argument depends upon its audience Attack and Defeat Dung gave us credulous and sceptical acceptance every attack succeeds, so there is no room for personal expression In politics, in law, in morals, in life, we need to recognise that we are not compelled by arguments: we choose to accept them An attack may fail: we need to distinguish attack from defeat
subjective and objective acceptance Unsuccessful attack Preferences: States preferences Cayrol and Amgoud. A preference order on arguments. Framework has a single audience Values: Relates preferences Arguments are related to a (smallish) set of values. Audiences order the values (and hence the arguments). Framework has several audiences
Explicit reasons for Extended Argumentation: (and against)preferences Modgil. Arguments attack attacks as well as arguments. Preferences are arguments defeating certain attacks. Audiences are themselves arguments. Relation to law Laws are made for a purpose to promote certain social values. Stare decisis must give way to Revenge is no longer an acceptable changing
values Thurgood Marshall value Marshall arguing against capital punishment: (Supreme Court Justice) Furman v Georgia (1972) Can explain variation between different and different Revenge times restored as a
proper American value Gregg v Georgia (1976) jurisdictions Can provide the basis for reasoning about which argument to accept Argumentation Schemes Logic programming showed us the need for argumentation schemes (expert opinion, ignorance) Toulmin showed us the need to distinguish different roles for premises (assumptions, presumptions, exceptions) Argumentation schemes provide a way of exploring these issues
Practical Reasoning Scheme I have used most: In the current circumstances We should do this action Which will have these consequences And so realise this goal Which will promote this value Encapsulates
current beliefs, knowledge of possible plans, a causal theory, behavioural assumptions about others, theory of values, preferences, side effects etc. A lot of reasoning hidden in this scheme Models Basis for many argumentation systems is a standard rule base Argumentation schemes as defeasible rules Same old problems of knowledge based systems since the 1980s: lack robustness hard to maintain hard to reuse, extend and share knowledge
The answer to these problems was model based reasoning, reasoning from first principles, qualitative reasoning Models for Practical Reasoning Alternating Action based Transition Systems make an effective basis for models to support practical reasoning States can represent current beliefs and possible consequences Transitions represent joint actions of the relevant agents Label transitions with the values they promote and demote
See my demo later today Modelling Real Arguments It is all too easy to got lost in a formalism Plenty of interest there, but it can be misleading Some problems are artefacts of the representation Often the answer is to represent the problem correctly, not to seek a technical fix Make the problems concrete before looking for a solution Flies back in the fly bottles Real arguments can provide a sanity check; inspiration more interest
Real Arguments Are Out There Since 1984 I have worked in AI and Law No shortage of arguments in the Legal Domain Supreme Court Arguments The wild animals cases Fourth Amendment Cases Capital Punishment cases Lots of interest in argument mapping for understanding and analysis Need to model them computationally too Why Argumentation?
Logic provides an idealisation of reasoning, a straight, well maintained highway But most interesting problems (outside mathematics) do not fit this very well Logic is a motorway, neat, tidy and effective Argumentation is a country lane, slow and frustrating but interesting Interesting problems One of more of the following features: Have reasons for believing both sides Are decided relative to a context Are decided relative to the audience (and it is normal for different people to decide
differently) Conclusions are chosen not simply recognised Are about values, aspirations and interests, not about facts Is it true? Is often the least interesting question that can be asked. Is it interesting, helpful, beautiful are often much more important. Computational Argumentation Of course, computational argumentation, has its own mathematics. For abstract argumentation, we need formal methods and proofs Plenty of that here
But dont lose sight of why we do argumentation Dont lose all sight of real arguments, those made by real people in real situations Like people, it is the peculiarities and specific Dont abstract all the interest away from them features that make arguments worth studying Thanks To all former and current friends and colleagues, that I have argued with over the years
Once METH binds one of two things happens… Option 1: The DAT is reversed, dumping DA out into the synpase. Option 2: Vesicles containing DA dump their DA into the cytosol, building up in the presynaptic neuron, and making a...
START FINISH Ace extension and Bone Building Bessemer Building, Aston Webb Building and Goldsmiths Beit Building Biochemistry Building Sir Alexander Fleming Building Blackett Building Chemistry Building Chemistry RCS1 Skempton Building College House Electrical Engineering Building Flowers Building Huxley Building Imperial...
Options for Integrating Quotations: #3 = Use a sentence. Use a full sentence of your own to introduce a full sentence by the quoted author. You must use a colon to introduce the quotation in this case. Jacob Needleman says...
Reconsideration (Recon) First level of appeal in most states, except for these areas: Alabama, Alaska, California (Los Angeles North and Los Angeles WB ), Colorado, Louisiana, Michigan, Missouri, New Hampshire, New York, and Pennsylvania. Request for Recon needed to initiate...
The Tudor dynasty is shown here in the context of the history of the Mint at the Tower of London. Historical notes: The Tudors and the Mint -- England's main Mint was at the Tower of London throughout the Tudor...
Fastening . Fasteners Devices that hold two, or more, things together. Ex: nails, staples, bolts, screws, rivets and dowels. Unfortunately, the holes made in the structure, by the fastener, actually weaken the structure
Ready to download the document? Go ahead and hit continue!