Dialogical Models of Explanation

1
Dialogical Models of Explanation

• ExaCt Vancouver July 21-22, 2007 Douglas Walton

2
Example of a Hamblin Dialogue
3
Dialogue Typology
4
Locutions and Speech Acts

• Statements and questions are locutions.
• Making an assertion is a speech act.
• Asking a question is a speech act.
• Asking for an explanation is an even more
  specific speech act.
• Putting forward an argument is a speech act.
• Offering an explanation is a speech act.

5
Speech Act Moves in a Dialogue
6
Evaluating Argumentation

• Take the text of discourse as your evidence.
• Is the selected speech act an argument, a report
  or an explanation?
• If an argument, what are the premises and
  conclusions?
• Does it fit an argumentation scheme?
• Apply the scheme to the argument

7
Reasoning, Argument and Explanation
Reasoning can be used for differing purposes, for
example in explanations and arguments. Reasoning
is a process of inference in passing from certain
propositions known or assumed to be true to other
propositions in a sequence (Walton, 1990).
Abductive reasoning is inference to the best
explanation (Josephsons, 1994). Practical
reasoning seeks out a prudential line of conduct
for an agent in a particular situation, while
theoretical reasoning seeks evidence that counts
for or against the truth of a proposition
(Walton, 1990).
8
What is an Argument?
An argument is a social and verbal means of
trying to resolve, or at least contend with, a
conflict or difference that has arisen between
two parties engaged in a dialogue (Walton 1990,
p. 411). According to this definition, an
argument necessarily involves a claim that is
advanced by one of the parties, typically an
opinion that the one party has put forward as
true, and that the other party questions.
9
Asking Questions

• The speech act of asking a question is different
  from the speech act of putting forward an
  argument.
• Questions dont make assertions.
• But questions can be loaded.
• So asking a question may not be entirely harmless
  or free from assertive content.

10
What is an Explanation?
The new dialectical theory (Walton, 2004) models
an explanation as a dialogue between two agents
in which one agent is presumed by a second agent
to understand something, and the second agent
asks a question meant to enable him to come to
understand it as well. The model articulates the
view of Scriven (2002, p. 49) Explanation is
literally and logically the process of filling in
gaps in understanding, and to do this we must
start out with some understanding of something.
11
How to Tell the Difference
Test to judge whether a given text of discourse
contains an argument or an explanation. Take the
statement that is the thing to be proved or
explained, and ask yourself the following
question. Is it taken as an accepted fact, or
something that is in doubt? If the former, its
an explanation. If the latter, its an argument.
The Goal of Dialogue is Different The purpose
of an argument is to get the hearer to come to
accept something that is doubtful or unsettled.
The purpose of an explanation is to get him to
understand something that he already accepts as a
fact.
12
Dialogue Model of Explanation

• Dialogue Conditions explainee asks question of a
  specific form asking about what is assumed to be
  a known fact S.
• Understanding Conditions explainee does not
  understand S, but assumes that explainer
  understands S.
• Success Conditions explainer by what she says
  brings the explainee to understand S.

13
Explanation in a Sequence of Dialogue
14
Rules for CE Dialogue System

• Opening when explainee makes an explanation
  request for S (accepted fact).
• Locution Rules defines different speech acts
  (kinds of moves) that are allowed.
• Dialogue Rules show which move must follow each
  previous kind of move.
• Success Rules show when transfer of
  understanding has been achieved.
• Closing when explainee says In understand it
  or explainer says I cant explain it.

15
Typical Profile of Explanation Dialogue
16
Problems for Future Work

• How can we test whether understanding has
  successfully been transferred?
• How can we evaluate whether one given explanation
  is better than another?
• What is the structure of explanations of human
  (and artificial agent) actions?
• What tools do we have for visualizing the logical
  structure of an explanation?

17
What is the test whether understanding has been
successfully transferred?

• Scriven (1972, p. 32) Suggested an answer to this
  question in his remark quoted below.1
• How is it that we test comprehension or
  understanding of a theory? We ask the subject
  questions about it, questions of a particular
  kind. They must not merely request recovery of
  information that has been explicitly presented
  (that would test mere knowledge, as in knowing
  the time or knowing the age of the universe).
  They must instead test the capacity to answer new
  questions.
• This remark suggests that the test is the
  explainees capacity to answer new questions,
  shown in a dialogue.
• But what kind of dialogue is it?

18
Examination Dialogue

• The examiner puts questions to the examinee,
  keeps track of the examinees answers, and probes
  into them critically.
• Examination dialogue is classified by Dunne,
  Doutre and Bench-Capon (2004), and Walton (2006)
  as a species of information-seeking type of
  dialog that can often shift to a persuasion
  dialog in which the questioner critically probes
  into the tenability of the respondents
  collective replies (Dunne, Doutre and
  Bench-Capon, 2004, p. 1560).
• In this way, the formal structure of examination
  as a dialogue model can be applied to central
  features of the kind of cases of examination
  commonly found in trials in law.

19
Shift from Explanation to Examination

• We can test for the success (failure) of an
  explanation by asking the explainee questions
  about new situations that are similar to S or are
  extrapolated from S.
• For example, if the explainee can draw a new
  inference from S that is reasonable, that is
  evidence he has understood S correctly.

20
Example of a Dialectical Shift

• Two agents have a joint intention to hang a
  picture. One has the picture and a hammer, and
  knows where the other can get a nail. They have a
  deliberation dialogue but cant agree on who
  should do which task. They then shift to a
  negotiation dialogue in which the one agent
  proposes that he will hang the picture if the
  other agent will go and get the nail (Parsons and
  Jennings, 1997).

21
Evaluating Competing Explanations

• Car crash where passenger claimed driver lost
  control, and driver claimed passenger suddenly
  pulled handbrake (Dutch Supreme Court Case cited
  by Prakken and Renooij, 2001).
• Evidence skid marks, crashed car, handbrake
  found in pulled position, expert witness says
  pulling handbrakes can cause wheels to lock.
• Analysis of Prakken and Renooij the drivers
  account explains the factual evidence and
  contradicts less of it than the passengers
  account.
• For summary, see (Walton, 2004a, 170-175)

22
Action Explanations

• Explanations of actions, of the kind especially
  common in history and law, is based on
  goal-directed reasoning.
• Goal-directed or means-end reasoning is called
  practical reasoning.
• Practical reasoning has a special argumentation
  scheme. Indeed, it has two of them.

23
Instrumental Scheme for Practical Reasoning

• I have a goal G.
• Bringing about A is necessary (or sufficient) for
  me to bring about G.
• Therefore, I should (practically ought to) bring
  about A.

24
Scheme for Value-based Practical Reasoning

• I have a goal G.
• G is supported by my set of values, V.
• Bringing about A is necessary (or sufficient) for
  me to bring about G.
• Therefore, I should (practically ought to) bring
  about A.

25
The Scalpicin Example
26
Araucaria
Araucaria is a software tool for analyzing
arguments. It aids a user in reconstructing and
diagramming an argument using a simple
point-and-click interface. The software also
supports argumentation schemes, and provides a
user-customizable set of schemes with which to
analyze arguments. Once arguments have been
analyzed they can be saved in a portable format
called "AML", the Argument Markup Language, which
is based on XML.
http//www.computing.dundee.ac.uk/staff/creed/arau
caria/
27
Some References
P.E. Dunne, S. Doutre, and T.J.M. Bench-Capon,
Discovering Inconsistency through Examination
Dialogues, Proceedings IJCAI-05, Edinburgh,
2005, 1560-1561. S. Parsons and N. R. Jennings,
Negotiation through Argumentation A Preliminary
Report, Proceedings of the Second International
Conference on Multi-Agents Systems, ed. Mario
Tokoro, AAAI Press, Menlo Park, California, 1997,
267-274. H. Prakken and S. Renooij,
Reconstructing Causal Reasoning about Evidence
A Case Study, Legal Knowledge and Information
Systems, ed. B. Verheij et al., Amsterdam, IOS
Press, 131-142. M. Scriven, The Concept of
Comprehension from Semantics to Software,
Language Comprehension and the Acquisition of
Knowledge, ed. J.B. Carroll and R.O. Freedle,
Washington, W. H. Winston Sons, 1972, 31-39. M.
Scriven, The Limits of Explication,
Argumentation, 16, 2002, 47-57. D. Walton, A New
Dialectical Theory of Explanation, Philosophical
Explorations, 7, 2004, 71-89. D. Walton,
Abductive Reasoning, University of Alabama Press,
2004a. D. Walton, Examination Dialogue An
Argumentation Framework for Critically
Questioning an Expert Opinion, Journal of
Pragmatics, 38, 2006, 745-777.