Bereich für Partnerlogo
Auf eine Ausgewogenheit zwischen den gleichberechtigten
Logos ist unbedingt zu achten. Gegebenenfalls muss die
Größe des Partnerlogos angepasst werden.
Die Größe des Logos der Universität Konstanz darf nicht
verändert werden.
Computational Linguistics
Week 3: Chris Reed‘s talk (Lab session)
Annette Hautli-Janisz, Prof. Dr.
Passau, 18 May 2022
Universität Konstanz
Argument Technology in the wild
information retrieval
[Link]
2 Universität Konstanz
Now
A very short introduction to argumentation theory.
Based on:
Frans van Eemeren et al. 2007. The Handbook of Argumentation Theory,
Chapter 1. Springer.
Douglas Walton. 2009. Argumentation Theory: A Very Short Introduction.
In: Simari G., Rahwan I. (eds) Argumentation in Artificial Intelligence.
Springer, Boston, MA.
3 Universität Konstanz
The beginning
Deductive logic, starting in Ancient Greece (Aristotle et al.):
The process of reasoning from one or more statements (premises) to reach
a logical conclusion.
Classic example:
All men are mortal. (first premise)
Socrates is a man. (second premise)
Therefore, Socrates is mortal. (conclusion)
4 Universität Konstanz
The beginning
Deductive logic, starting in Ancient Greece (Aristotle et al.):
The process of reasoning from one or more statements (premises) to reach
a logical conclusion.
Classic example:
P!Q (first premise)
P (second premise)
Q (conclusion)
5 Universität Konstanz
The beginning
Deductive logic, starting in Ancient Greece (Aristotle et al.):
The process of reasoning from one or more statements (premises) to reach
a logical conclusion.
Classic example:
P!Q (first premise)
P (second premise)
Q (conclusion)
Modus ponens: “P implies Q and P are both asserted to be true, therefore
Q must be true” (primary deductive rule of inference)
6 Universität Konstanz
The beginning
Deductive logic, starting in Ancient Greece (Aristotle et al.):
The process of reasoning from one or more statements (premises) to reach
a logical conclusion.
Classic example from Theophrastus:
P!Q (first premise)
Q (second premise)
P (conclusion)
Modus tollens: “P implies Q is asserted, it is not the case that Q, therefore
it’s not the case that P”
7 Universität Konstanz
The beginning
The limits of deductive logic
• Valid deductive arguments guarantee that the conclusion is true iff the
premises are.
• How is this guarantee possible?
• There is absolutely no more information in the conclusion than was in
the premises (usually less).
Anyone who is 2m tall is more than 1m tall.
Fred is 2m tall.
So, Fred is more than 1m tall.
• We have thrown some available information away – cost of ‘the
guarantee’.
8 [Link] Universität Konstanz
The beginning
In contrast: Non-deductive (inductive) logic (also starting in Ancient
Greece)
We can produce arguments that increase (or amplify) our information only
by giving up the deductive guarantee, e.g., polling.
• Contrast:
• asking everyone some question and recording their answers
• asking some of the population that question and recording their
answers
• In the second case, what do we know about what the whole
population thinks?
• No guarantee – but a lot less work!
9 [Link] Universität Konstanz
The beginning
Non-deductive (inductive) logic
In a good inductive argument, the truth of the premises provides some
degree of support for the truth of the conclusion, where this degree-of-
support might be measured via some numerical scale. (Stanford Encyclopedia of Philosophy)
Example:
90% of freshmen are between 18 and 22 years of age.
John is a freshman.
John is between 18 and 22 years of age.
Strong argument, but the truth of the premise does not guarantee the truth
of the conclusion.
10 [Link] Universität Konstanz
The beginning
Non-deductive (inductive) logic
Inductive reasoning is a method of
reasoning in which the premises
are viewed as supplying some
evidence for the truth of the
conclusion. While the conclusion
Copi, I. M.; Cohen, C.; Flage, D.
E. (2006). Essentials of Logic of a deductive argument is
(Second ed.). Upper Saddle certain, the truth of the conclusion
River, NJ: Pearson Education. of an inductive argument may be
probable, based upon the
evidence given.
11 Universität Konstanz
Fallacies
Starting with Aristotle and going on for the next 2000 years: Identify
the errors (fallacies) of reasoning
Copi (1961): Set of core
fallacies stemming from
Aristotle and John Locke (1690) A fallacy is a form of argument
that seems to be correct but which
proves, upon examination, not to
be so.
12 Universität Konstanz
Fallacies
Example fallacies:
The fallacy of equivocation: argument which exploits the ambiguity of a
term or phrase which has occurred at least twice in an argument
The end of life is death.
Happiness is the end of life.
So, death is happiness.
What is the ambiguous term and what are its meanings?
! A non sequitur (lat. ‘it does not follow’)
Source: Stanford Encyclopedic of Philosophy
13 Universität Konstanz
Fallacies
Example fallacies:
The fallacy of amphiboly: a fallacy of ambiguity where the ambiguity is due
to indeterminate syntactic structure
The police were told to stop drinking on campus after midnight.
So, now they are able to respond to emergencies much better than
before.
Source: Stanford Encyclopedic of Philosophy
14 Universität Konstanz
From the 1970s onwards
Informal logic (aka ‘argumentation’):
New, practical approach to critical thinking beyond deductive logic.
Practical methods for analyzing and evaluating arguments and apply them
to real, naturally occurring examples of argumentation.
These methods are in a process of rapid evolution
• computer science
• argumentation models in the field of artificial intelligence (multi-agent
systems, legal reasoning)
15 Universität Konstanz
Informal logic (aka ‘argumentation’)
Four tasks:
1. Identification
a. Identify the premises and conclusion of an argument in the text
b. Does the argument fit an argumentation scheme? (more below)
2. Analysis
Find implicit premises or conclusions that need to made explicit in
order to properly evaluate an argument (enthymemes, more below)
3. Evaluation
Determine whether an argument is strong or weak
4. Invention
Construct new arguments in order to prove a specific conclusion
16 Universität Konstanz
Informal logic (aka ‘argumentation’)
1a. Identifying premises and conclusions
p1: Switzerland is the perfect place for living.
p2: People are kind and honest.
p3: According to my experience, people are boring and suspicious towards
foreigners.
p4: Salaries are adequate to the cost of living.
p5: Swiss people do not have a common culture. People speak three
different languages, and a culture is defined by its language.
17 Universität Konstanz
Informal logic (aka ‘argumentation’)
1a. Identifying premises and conclusions
Argumentation diagramming
[Link]
18 Universität Konstanz
Informal logic (aka ‘argumentation’)
1a. Identifying premises and conclusions
Argumentation diagramming
Convergent argument
[Link]
19 Universität Konstanz
Informal logic (aka ‘argumentation’)
1a. Identifying premises and conclusions
Argumentation diagramming
Convergent argument
Linked argument
[Link]
20 Universität Konstanz
Informal logic (aka ‘argumentation’)
Four tasks:
1. Identification
a. Identify the premises and conclusion of an argument in the text
b. Does the argument fit an argumentation scheme?
2. Analysis
Find implicit premises or conclusions that need to made explicit in
order to properly evaluate an argument (enthymemes, more below)
3. Evaluation
Determine whether an argument is strong or weak
4. Invention
Construct new arguments in order to prove a specific conclusion
21 Universität Konstanz
Informal logic (aka ‘argumentation’)
1b. Argumentation schemes
Argumentation schemes are abstract argument forms commonly used in
everyday conversational argumentation, legal and scientific argumentation
(not the deductive argument forms)
Some of the most common schemes (Walton 1996 presents 25 schemes):
• Argument from expert opinion
• Argument from popular opinion
• Argument from example
• …
22 Universität Konstanz
Informal logic (aka ‘argumentation’)
1b. Argumentation schemes
Example: Argument from expert opinion
A is a proposition
E is an expert
D is a domain of knowledge
Major premise: Source E is an expert in subject domain S containing a
proposition A.
Minor premise: E asserts that proposition A is true (false)
Conclusion: A is true (false)
AND: Set of critical questions to critically probe into an argument
23 Universität Konstanz
Informal logic (aka ‘argumentation’)
1b. Argumentation schemes
Example: Argument from expert opinion
24 Universität Konstanz
Informal logic (aka ‘argumentation’)
1b. Argumentation schemes
Example: Argument from expert opinion
Critical questions (Walton, Reed and Macagno 2008)
CQ1: Expertise Question. How credible is E as an expert source?
CQ2: Field Question. Is E an expert in the field that A is in?
CQ3: Opinion Question. What did E assert that implies A?
CQ4: Trustworthiness Question. Is E personally reliable as a source?
CQ5: Consistency Question. Is A consistent with what other experts assert?
CQ6: Backup Evidence Question. Is E's assertion based on evidence?
25 Universität Konstanz
Informal logic (aka ‘argumentation’)
1b. Argumentation schemes
Example: Argument from expert opinion
26 Universität Konstanz
Informal logic (aka ‘argumentation’)
Four tasks:
1. Identification
a. Identify the premises and conclusion of an argument in the text
b. Does the argument fit an argumentation scheme? (more below)
2. Analysis
Find implicit premises or conclusions that need to made explicit in
order to properly evaluate an argument (enthymemes, more below)
3. Evaluation
Determine whether an argument is strong or weak
4. Invention
Construct new arguments in order to prove a specific conclusion
27 Universität Konstanz
Informal logic (aka ‘argumentation’)
2. Find implicit premises and conclusions
Enthymeme (from enthymema,
lit. ‘thought’, ‘consideration’)
An enthymeme is an argument
with an implicit premise or
conclusion that needs to be
made explicit before the
argument can be properly
understood or evaluated.
28 Universität Konstanz
Informal logic (aka ‘argumentation’)
2. Find implicit premises and conclusions
Reconsider: Argument from expert opinion
29 Universität Konstanz
Informal logic (aka ‘argumentation’)
2. Find implicit premises and conclusions
Reconsider: Argument from expert opinion
30 Universität Konstanz
Informal logic (aka ‘argumentation’)
Enthymemes: Examples
1. We cannot trust Katie, because she lied last week.
2. The gun has the defendant’s fingerprints on the trigger. He is clearly
guilty!
3. “The battle of Iraq is one victory in a war on terror that began on
September the 11th, 2001, and still goes on … With those attacks, the
terrorists and their supporters declared war on the United States. And
war is what they got.” (George Bush, New York Times Interview, May 2,
2003)
31 Universität Konstanz
Argue with a chatbot!
32 Universität Konstanz
Quick glance: Developing datasets for argument
mining
Argument mining is one of the grand challenges in NLP (more of this in the
lecture on 7 July on ‘Discourse processing and argument mining’).
Common approach: supervised machine learning.
! Gold standard annotation of argumentation.
In case of dialogical argumentation: Inference Anchoring Theory, labeling in
OVA 3 ([Link]
Example:
Bob: The government will inevitably lower the tax rate.
Wilma: Why?
Bob: Lower taxes stimulate the economy.
33 Universität Konstanz
Argument Technology in the wild
information retrieval
[Link]
34 Universität Konstanz
Bereich für Partnerlogo
Auf eine Ausgewogenheit zwischen den gleichberechtigten
Logos ist unbedingt zu achten. Gegebenenfalls muss die
Größe des Partnerlogos angepasst werden.
Die Größe des Logos der Universität Konstanz darf nicht
verändert werden.
Thank you.
Questions?
Comments?
Annette Hautli-Janisz, Prof. Dr.
cornlp-teaching@[Link]
Universität Konstanz
