KNOWLEDGE AND ACTION IN RATIONAL DELIBERATION

1
KNOWLEDGE AND ACTION IN RATIONAL DELIBERATION

• Douglas Walton
• CRRAR
• 7th Conference on Analytic Philosophy in China,
  Shanghai, Oct. 30, 2011

2
Bounded Procedural Rationality

• In this paper it is shown how knowledge can lead
  to a rational decision for action or inaction
  based on argumentation process called
  deliberation.
• The viewpoint adopted is one of bounded
  procedural rationality based on a notion of
  defeasible knowledge.
• The problem confronted is that decision-making
  about real-world problems needs to be made under
  conditions of uncertainty, and even apparent
  inconsistency where there is both pro and contra
  evidence for a conclusion to be decided.

3
Argumentation Methods

• It is just in this kind of case that methods of
  argumentation are especially useful.
• Argumentation can be defined as a procedure to
  identify, analyze and evaluate the arguments on
  both sides of a claim, and to use the evidence
  that is collected by this procedure to determine
  whether to accept the claim or not.
• It is also a part of argumentation methodology
  that setting a burden of proof on each side by
  determining what kind of arguments are relevant,
  and what standards of proof should be required,
  is an essential requirements of the procedure.

4
First Example

• Lets consider the case of the student who is
  writing an essay. He is collecting all kinds of
  knowledge from books and periodicals, but he has
  a strict deadline for finishing the assignment.
• This problem is to determine when he should stop
  searching for new knowledge and attempt to write
  the essay. The longer he delays writing in order
  to search for new knowledge, the better the essay
  will be. But if he delays too long, he will not
  have enough time to properly write the essay, and
  the result will be that the essay will not be
  very good.
• The general problem in many comparable cases of
  this kind is one of when to terminate the process
  of deliberation and close off the collecting of
  new knowledge.

5
Second Example

• Another kind of example can also be cited. In
  July 2010, scientists found a way of altering the
  DNA of mosquitoes that shortens their lifespan so
  that malarial parasites to not have enough time
  to grow to maturity. This discovery gives the
  scientists the possibility of releasing a
  malaria-proof mosquito into the wild, thereby
  eliminating mosquitoes that can cause malaria.
  Right now malaria kills about one million people
  every year.
• However, there is a problem. Altering the DNA of
  mosquitoes might make them better carriers of
  other diseases.
• This proposal cannot be carried out for another
  ten years, and even then, there may be no way to
  know what the consequences are.
• Once the malaria-proof mosquito is produced by
  the scientists, even though we can study the
  problem and collect knowledge about it, we will
  never know what all the side-effects will be
  until we release the new mosquitoes into the
  wild.
• R. M. Schneidermann, God lives in a Lab in
  Arizona, Newsweek, August 9, 2010, 8.

6
Aquinas Poses the Problem

• Aquinas asked the question, May deliberation go
  on endlessly? and answered it in his Summa
  Theologiae in Question 14, Article 6 quoted from
  (Blackfriars Edition, 155)
•  
• 1. Yes, apparently, for it is about the
  particular things which are the concern of
  practical knowledge. These are infinite.
  Accordingly no term is to be set to the inquiry
  of deliberation about them.
• 2. Further, we have to weigh up not only what has
  to be done, but also how to clear away the
  obstacles. Now any number of objections to any
  particular course of action can be put up and
  knocked down in our mind. Therefore there is no
  stop to our questioning about how to deal with
  them.
• 3. Moreover, the inquiry instituted by
  demonstrative science does not lead back
  indefinitely, but arrives at self-evident
  principles which are altogether certain. Such
  certainty, however, cannot be found in contingent
  and individual facts, which are variable and
  uncertain. Deliberation, therefore, goes on
  endlessly.

7
The Closed World Assumption

• The closed world assumption is the inference
  drawn that any positive fact not specified in a
  given database may be assumed to be false, on the
  basis that all of the relevant knowledge has been
  specified (Reiter, 1987).
• Consider the familiar example (Reiter 1980, 69)
  of scanning an airline monitor. Lets say that no
  direct flight is listed from Windsor to Shanghai.
  The closed world assumption is that all the
  relevant data on flights leaving from Windsor at
  this time are listed on the monitor on the
  airport website.
• So if a direct Windsor to Shanghai flight is not
  listed, it is reasonable to draw the conclusion
  that no such flight is available. In this
  situation, the closed world assumption is
  reasonable to invoke, because we have good reason
  to assume that the knowledge base is complete.

8
Another Example

• The official listing of baseball statistics about
  hits, home runs and so forth, is known to not
  only be complete for the major-league baseball
  teams, but also highly reliable, because there
  are many fans who are passionate about keeping
  baseball statistics, and who would immediately
  challenge any error they might find in the
  baseball statistics knowledge base.
• So if we were to look in the database and see
  that some information about some home runs
  alleged to be hit in 1936 was not in it, we could
  very confidently invoke the closed world
  assumption to draw the conclusion that there were
  no such home runs hit that year.

9
Scheme for Practical Reasoning

• Major Premise I have a goal G.
• Minor Premise Carrying out action A is a means
  to realize G.
• Conclusion I ought (practically speaking) to
  carry out action A.
• The first-person singular pronoun I in the
  scheme for practical reasoning above represents
  an agent. An agent is an entity that has goals
  and knowledge about its circumstances, can take
  action in its circumstances based on this
  knowledge, and can also see the consequences of
  its actions so that it can correct them through
  feedback.

10
Critical Questions Matching Scheme

• CQ1 What other goals do I have that should be
  considered that might conflict with G?
• CQ2 What alternative actions to my bringing
  about A that would also bring about G should be
  considered?
• CQ3 Among bringing about A and these alternative
  actions, which is arguably the most efficient?
• CQ4 What grounds are there for arguing that it
  is practically possible for me to bring about A?
• CQ5 What consequences of my bringing about A
  should also be taken into account?

11
How to Use Critical Questions

• These five basic critical questions for practical
  reasoning are not complete.
• As shown in (Walton 1990), each of these five
  critical questions has critical sub-questions.
• The five basic critical questions are meant as
  devices to help a critic or student of critical
  thinking find weak points in an argument of this
  type that can be challenged or cast into doubt.

12
Value-based Practical Reasoning

• Value-based practical reasoning (Atkinson,
  Bench-Capon and McBurney, 2006) is made up of two
  more basic schemes, the one for practical
  reasoning and the one for argument from values
  (Bench-Capon, 2003). The argumentation scheme for
  value-based practical reasoning has this form
  (Atkinson, Bench-Capon and McBurney, 2006).
•  
• Scheme for Value-based Practical Reasoning
•  
• In the current circumstances R
• we should perform action A
• to achieve New Circumstances S
• which will realize some goal G
• which will promote some value V.

13
Questioning versus Arguing

• Asking the critical question C5 needs to be
  backed up with some evidence of what the negative
  consequences are before an argument based on
  practical reasoning is refuted.
• When this happens, the asking of this critical
  question can also be analyzed as a
  counter-argument.
• Argument from negative consequences cites the
  consequences of a proposed course of action as a
  reason against taking that course of action.
• This argument also has a positive form in which
  positive consequences of an action are cited as a
  reason for carrying out the action.

14
Argument from Consequences

• Scheme for Argument from Positive Consequences
•  
• Major Premise Its having good consequences is a
  reason for doing something.
• Minor Premise If A is brought about, good
  consequences will plausibly occur.
• Conclusion A should be brought about.
•  
• Scheme for Argument from Negative Consequences
•  
• Major Premise Its having good consequences is a
  reason for not doing something.
• Minor Premise If A is brought about, then bad
  consequences will occur.
• Conclusion A should not be brought about.

15
The Carneades System

• The Carneades Argumentation System uses
  argumentation schemes and critical questions for
  argument analysis and evaluation (Gordon, 2010).
• Carneades is a computational model consisting of
  mathematical structures and functions on them
  (Gordon, Prakken and Walton, 2007).
• Carneades models the structure of arguments, the
  acceptability of statements, and burdens of
  proof.
• Carneades has an open source graphical user
  interface (http//carneades.github.com/ ).

16
Tweety Example
17
Wiki Example 1
18
Problem with Critical Questions

• It would be very nice if the five critical
  questions for practical reasoning could be
  represented as additional premises of the
  argumentation scheme. Then we could represent the
  critical questions as implicit assumptions of the
  argument when we analyze the argument using an
  argument diagram of the standard kind.
• However, there is a problem. With some critical
  questions, simply asking the question is enough
  to defeat the original argument, whereas with
  other critical questions, merely asking the
  critical question is not enough to defeat the
  argument. In order to defeat the argument some
  evidence has to be given to back up the critical
  question.

19
Solution to the Problem

• To solve this problem, the Carneades system
  offers two ways of responding the asking of a
  critical question by distinguishing three types
  of premises in argumentation scheme, called
  ordinary premises, assumptions and exceptions
  (Walton and Gordon, 2009).
• Ordinary premises are explicitly stated premises
  of the argumentation scheme. They are assumed to
  hold tentatively, but if challenged they may have
  to be given up.
• Assumptions, like ordinary premises, are assumed
  to be true.
• Exceptions are assumed not to hold, and therefore
  they do not defeat an argument unless backed up
  by evidence to support them.

20
Carneades Map of Practical Reasoning
21
Evaluating Arguments

• Based on this method of representing the critical
  questions of an argumentation scheme by
  representing them as different kinds of premises
  of scheme, Carneades has a computational method
  for evaluating arguments (Gordon and Walton,
  2006).
• At each stage of the argumentation process, an
  effective method (decision procedure) is used for
  testing whether the conclusion of an argument is
  acceptable or not, given knowledge about whether
  its premises are acceptable or not.
• The assumptions represent undisputed facts, the
  current consensus of the participants, or the
  commitments or beliefs of some agent, depending
  on the task.
• The evaluation of the given argument may depend
  on the proof standard applicable to the
  proposition at issue, and on the dialogue
  procedure the argument is embedded in.

22
Dialogue Models

• Dialogue models have rules on how participants
  should ideally speak and respond in order to
  achieve a common conversational goal.
• Dialogue models of argumentation (Walton and
  Krabbe, 1995) have proved their usefulness in
  argumentation studies, artificial intelligence,
  and multi-agent systems (Bench-Capon, 2003
  Prakken, 2005).
• Walton and Krabbe (1995) identified six primary
  types of dialogue information-seeking dialogue,
  inquiry, deliberation, persuasion dialogue,
  negotiation and eristic (quarrelsome) dialogue.

23
Formal Dialogue Systems

• A dialogue is generally a group activity with
  multiple participants, but in the simplest case
  there are only two parties called the proponent
  and the respondent.
• A dialogue is defined in the Carneades model as
  an ordered 3-tuple ltO, A, Cgt where O is the
  opening stage, A is the argumentation stage, and
  C is the closing stage (Gordon and Walton, 2009,
  5).
• Dialogue protocols regulate the types of moves
  that are allowed and how a participant must
  respond to a previous move made by the other
  party (Walton and Krabbe, 1995).
• So far, Carneades has not provided protocols for
  deliberation dialogues, but there is a formal
  model.

24
Deliberation Dialogue

• The initial situation of deliberation is the need
  for action arising out of a choice between
  alternative competing courses of action.
• The collective goal of this type of dialogue is
  for the participants to collectively decide on
  what is the best available proposal for action
  that has been put forward for the group at the
  proposal stage, once that stage has been reached.
  
• Once that stage has been reached, the
  participants evaluate the proposals in a process
  in which each party puts forward its own
  proposals and critically evaluates the competing
  proposals put forward by others.
• There is also a prior inform stage where the
  facts collected and shared among the
  participants.
• In a successful deliberation, the strengths and
  weaknesses of each proposal are brought out by
  the discussion, and this evidence is used to
  judge which proposal is the one that should be
  selected to move forward with.

25
8 Stages of Deliberation Dialogue

• Open In this stage a governing question is
  raised about what is to be done. A governing
  question, like Where shall we go for dinner this
  evening? is posed.
• Inform This stage includes discussion of
  desirable goals, values, constraints on possible
  actions, evaluation criteria for proposals, and
  determination of relevant facts.
• Propose Proposals cite possible action-options
  relevant to the governing question
• Consider This stage concerns commenting on the
  proposals from various perspectives.
• Revise Goals, constraints, perspectives, and
  action-options are revised in light of comments
  presented and information gathering as well as
  fact-checking.
• Recommend A proposal for action is recommended
  for acceptance or non-acceptance by each
  participant.
• Confirm The participants can confirm acceptance
  of the recommended proposal according to some
  procedure they have agreed on.
• Close The termination of the dialogue takes
  place.
• (Hitchcock, McBurney and Parsons, 2007)

26
Judging a Deliberation Dialogue
27
Scientific Inquiry and Truth

• According to the traditional view the conclusion
  of an inquiry has to be drawn by the deductive
  chain of reasoning from a set of premises that
  are absolutely certain.
• Any scientific inquiry might lead to a conclusion
  which might need to be revised in the future
  (Peirce, 1931, 2.75). Popper called this
  falsifiability.
• Peirce wrote that many things are substantially
  certain (Peirce 1931, 1.152), but that this is
  different from the kind of absolute certainty
  that implies truth.
• On his view truth is an important motivation for
  scientists to have as the ultimate goal of
  scientific research, but he argued that that
  truth can only be arrived at beyond all doubt by
  an inquiry that would take an infinite amount of
  time.

28
Fallibilism of Peirce and Popper

• On their view, knowledge does not imply truth. In
  other words, it is not a requirement for
  proposition to be part of knowledge that it be
  true, at least in any sense requiring that it
  will not turn out to be false in the future.
• On this view, called fallibilism, scientific
  knowledge is defeasible, meaning that even though
  a proposition is accepted as knowledge, it might
  be defeated in the future by enough evidence
  casting doubt on it, or even showing that it is
  false, so that it needs to be retracted.

29
Defeasible Knowledge

• Peirce described the process of inquiry as one in
  which different participants set out with
  conflicting views, but are led through a process
  of marshalling and testing evidence to accepting
  the same conclusion. This convergence takes place
  as a successful inquiry moves to completion.
• According to (Walton, 2010) Peircean belief is
  characterized as a settled state we do not wish
  to change. Once fixed, it is something we cling
  tenaciously to. It is an indication of a habit,
  and a matter of degree.
• It puts us into a condition so we act in a
  certain way in the future, and it guides our
  desires and shapes our actions.
• On this view, defeasible knowledge is a species
  of belief that is fixed firmly by a scientific
  discipline through process of inquiry that tests
  the belief as a hypothesis against all the pro
  and contra evidence that can be collected and is
  relevant to proving or disproving it.

30
Proof Standards

• The following four standards of proof are used in
  the Carneades Argumentation System (Gordon and
  Walton, 2009).
• Scintilla of Evidence (SE) is met if there is at
  least one applicable argument for a claim.
• Preponderance of the Evidence (PE) is met if SE
  is satisfied and the maximum weight assigned to
  an applicable pro argument (for the claim) is
  greater than the maximum weight of an applicable
  con argument (against the claim).
• Clear and Convincing Evidence (CCE), is met if PE
  is satisfied, the maximum weight of applicable
  pro arguments exceeds some threshold a, and the
  difference between the maximum weight of the
  applicable pro arguments and the maximum weight
  of the applicable con arguments exceeds some
  threshold ß.
• Beyond Reasonable Doubt (BRD) is met if CCE is
  satisfied and the maximum weight of the
  applicable con arguments is less than some
  threshold ?.

31
The Mosquitoes Example

• The collection of knowledge phase will only be
  reached at some point after ten years. At this
  point there will have to be deliberations that
  many organizations will take part in, including
  the World Health Organization, which will need to
  develop rules for testing genetically modified
  mosquitoes.
• However, even at this point, it is possible to
  see how the argumentation structure of the
  deliberation in this case takes a pro and contra
  argument form based on argumentation schemes.

32
Carneades Map of Mosquitoes Argumentation
33
Evaluation of Mosquitoes Case

• This method of moving forward with an evaluating
  deliberation requires taking fully into account
  evidence obtained from scientific inquiry into
  the circumstances of the case.
• On this model, the factual basis of evidence from
  the scientific inquiry is part of what is
  required to assess the depth of assessment of the
  proposals made in the deliberation dialogue.
• The dialogue should only be closed when this
  depth of assessment by argumentation has met the
  standard of evidence set for this deliberation
  dialogue.
• At some point the cost in lives due to malaria
  will require that the decision be made one way or
  the other, provided that the alternative of doing
  nothing continues to result in highly significant
  loss of human lives.

34
Conclusion

• The paper has shown how a rational decision on
  what to do depends on an evaluation of the pro
  and contra arguments for each proposal, once all
  the proposals have been stated.
• It has also shown that this decision depends on
  how well informed these proposals are, based on
  the scientific and factual evidence concerning
  the circumstances of the case.
• To provide a method for making these decisions,
  the paper has utilized formal dialectical models
  of deliberation dialogue and inquiry dialogue,
  showing how the latter type of dialogue is
  embedded in the former.

35
3 Problems for Further Research

• The first is to devise computational
  argumentation tools to measure the depth of
  argumentation behind a proposal that has been
  discussed in a deliberation dialogue by the
  closing stage.
• The second is to show how knowledge is
  transferred from inquiry dialogue to deliberation
  dialogue, typically using the argumentation
  scheme for argument from expert opinion.
• The third is to apply the methods of this paper
  to a more detailed example of deliberation using
  knowledge obtained from scientific inquiry.

36
References

• Aquinas, T. (1970), Summa Theologiae, vol. 17.
  Trans. Thomas Gilby, Blackfriars Edition, New
  York, McGraw-Hill.
• Atkinson, K., Bench-Capon, T. and McBurney, P.
  (2006). Computational Representation of Practical
  Argument, Synthese, 152, 157-206.
• Bench-Capon, T. (2003). Persuasion in Practical
  Argument Using Value-basedArgumentation
  Frameworks, Journal of Logic and Computation, 13,
  429-448.
• Cooke, E. (2006). Peirces Pragmatic Theory of
  Inquiry Fallibilism and Indeterminacy. London
  Continuum.
• Gordon, T. F. (2010). The Carneades Argumentation
  Support System, Dialectics, Dialogue and
  Argumentation, ed. C. Reed and C. W. Tindale,
  London College Publications.
• Gordon, T. F. and Walton, D. (2009). Proof
  Burdens and Standards. Argumentation and
  Artificial Intelligence, ed. I. Rahwan and G.
  Simari. Berlin Springer, 239-260.
• Gordon, T. F., Prakken, H. and Walton, D. (2007).
  The Carneades Model of Argument and Burden of
  Proof, Artificial Intelligence, 171, 875-896.
• Hitchcock, D. McBurney, P. and Parsons, S.
  (2007). The Eightfold Way of Deliberation
  Dialogue, International Journal of Intelligent
  Systems, 22(1), 9513.
• Peirce, C. S. (1984), Writings of Charles S.
  Peirce A Chronological Edition, vol. 2. Ed. E.
  C. Moore. Bloomington Indiana University Press.
• Popper, K. (1972). Objective Knowledge An
  Evolutionary Approach. Oxford Oxford University
  Press, 1972.
• Prakken, H. (2005). Coherence and Flexibility in
  Dialogue Games for Argumentation. Journal of
  Logic and Computation 15, 1009-1040.
• Reiter, R. (1980). A Logic for Default Reasoning,
  Artificial Intelligence, 13, 81-132.
• Reiter, R. (1987). Nonmonotonic Reasoning. Annual
  Review of Computer Science 2 147-186.
• Walton, D. (2010). A Dialogue Model of Belief,
  Argument and Computation, 1, 23-46.
• Walton, D. (2007). Dialog Theory for Critical
  Argumentation. Amsterdam Benjamins.
• Walton, D. and Krabbe, E. C. W. (1995).
  Commitment in Dialogue Basic Concepts of
  Interpersonal Reasoning. Albany State University
  of New York Press.
• Walton, D. and Gordon, T. F. (2005). Critical
  Questions in Computational Models of Legal
  Argument, Argumentation in Artificial
  Intelligence and Law, IAAIL Workshop Series, ed.
  Dunne, P. E. and T. Bench-Capon, Nijmegen, Wolf
  Legal Publishers, 103-111.
• Walton, D., Reed, C. and Macagno, F. (2008).
  Argumentation Schemes. Cambridge Cambridge
  University Press.