Argument Analysis is a method for identifying the claims, evidence, premises, inferences, and objections contained in written or spoken communication and clarifying how these elements are related. It is used to examine why a particular conclusion is supported, whether the reasons offered are sufficient, and whether alternative interpretations or counterarguments exist.
For example, suppose the following proposal is made in a corporate research and development department:
“Although we have increased the amount of technological information we collect, we have not generated enough new R&D themes. Therefore, we need to reconsider not only how we collect information but also how we explore potential research themes.”
This statement contains both an observation about an existing problem and a proposal derived from that observation.
However, is the shortage of new R&D themes caused by the information collection process? Or could it be related to evaluation criteria, organizational structures, technological capabilities, or business constraints?
Even if it is established that increasing information collection has not generated enough new themes, does that fact alone justify changing the exploration process?
Argument Analysis does not begin by declaring such a statement correct or incorrect. Instead, it first clarifies what is presented as fact, which assumptions are being made, and how the conclusion is derived from those assumptions.
It then examines the reliability of the evidence, the appropriateness of the inference, possible objections, and the conditions under which the conclusion would hold.
The purpose of Argument Analysis is not to identify errors in someone else’s reasoning simply to defeat them in a debate. It is to clarify what supports a claim and where uncertainty or disagreement remains.
What Is an Argument?
An argument is a set of statements in which one or more reasons or premises are offered to support a claim or conclusion.
Unlike the everyday meaning of argument as a disagreement or quarrel, the term does not necessarily involve people expressing opposing views.
An argument may be contained in a single person’s written explanation.
The two most fundamental components of an argument are the Premises and the Conclusion.
Consider the following example:
“All products of type A must undergo inspection before shipment.”
“Product X belongs to type A.”
“Therefore, Product X must undergo inspection before shipment.”
The first two statements are premises, and the final statement is the conclusion.
In this case, if the premises are true, the conclusion must also be true.
Now consider another argument:
“The increase in Equipment A’s temperature coincided with an increase in the defect rate.”
“Therefore, the temperature increase may have contributed to the higher defect rate.”
Even if the premise is correct, the conclusion is not necessarily correct.
For example, both the temperature increase and the higher defect rate might have been caused by another factor.
Arguments therefore include both inferences in which the conclusion necessarily follows from the premises and inferences in which the premises provide only partial support for the conclusion.
The History and Theoretical Background of Argument Analysis
The study of arguments has a long history.
In ancient Greece, Aristotle systematically investigated syllogistic reasoning. In works such as Topics and Rhetoric, he also examined forms of reasoning used in practical discussion and persuasion.
Logic subsequently developed within philosophy and mathematics, with particular attention to the formal validity of inference.
However, many arguments used in everyday life do not involve conclusions that follow with deductive certainty.
People frequently make judgments based on expert opinions, previous experience, statistical trends, analogous examples, and anticipated consequences.
During the twentieth century, several approaches were developed to analyze these forms of practical reasoning.
Representative examples include Stephen Toulmin’s model of argumentation, the argumentation schemes studied by Douglas Walton and colleagues, and Pragma-Dialectics developed by Frans H. van Eemeren and Rob Grootendorst.
Computer science has also investigated methods for formally representing relations among arguments, including attacks and support.
Argument Mining, which uses natural-language processing to extract argumentative structures from text, is another related research field.
| Theory or Method | Main Concern |
|---|---|
| Formal Logic | Whether a conclusion follows logically from its premises |
| Informal Logic | Whether reasoning in practical arguments is appropriate |
| Toulmin Model | How claims, grounds, warrants, and other elements are connected |
| Argumentation Schemes | Common patterns of reasoning and questions for evaluating them |
| Pragma-Dialectics | How disagreements can be addressed through reasonable discussion |
| Computational Argumentation | How relations such as support and attack can be represented and evaluated computationally |
| Argument Mining | How argumentative components and relationships can be extracted from natural language |
These approaches are related but do not constitute a single unified theory.
This page focuses on principles useful for practical Argument Analysis and their potential applications in AI-assisted thinking.
The Basic Components of an Argument
The first task in Argument Analysis is to distinguish a claim from the information offered to support it.
Consider the following statement:
“Defective products have become more frequent on Equipment A. Measurements confirm that its positioning accuracy has deteriorated. Therefore, equipment adjustment or replacement should be considered.”
Several elements can be identified.
| Element | Content |
|---|---|
| Claim or Conclusion | Equipment adjustment or replacement should be considered |
| Evidence 1 | The occurrence of defective products has increased |
| Evidence 2 | The equipment’s positioning accuracy has deteriorated |
| Implicit Assumption | The deterioration in positioning accuracy is related to the defects |
| Additional Assumption | Adjustment or replacement may reduce the occurrence of defects |
| Unresolved Questions | Are other causes involved? What are the likely benefits and costs of the proposed measures? |
This analysis reveals assumptions that were not explicitly stated.
For example, confirming that positioning accuracy has deteriorated is not the same as establishing that this deterioration is the principal cause of the increased defect rate.
Furthermore, even if the cause is identified, equipment replacement may not be the best corrective measure.
Argument Analysis helps distinguish observed facts, inferences, and decisions.
Identifying Premises and Conclusions
In Argument Analysis, identifying the conclusion is often a useful starting point.
A conclusion does not necessarily appear at the end of an argument.
For example, the following statements express essentially the same argument in different sequences:
“Equipment A should be replaced because failures are becoming more frequent and repair expenses are increasing.”
“Equipment A has experienced increasingly frequent failures and rising repair expenses. Therefore, it should be replaced.”
“Shouldn’t we consider replacing Equipment A? Increasing failures and repair expenses have become serious problems.”
In each case, equipment failure and repair costs are presented as reasons supporting replacement.
However, the order and wording differ.
Real arguments may also contain ambiguous conclusions or multiple conclusions.
For example:
“Equipment A needs to be replaced.”
“Equipment A should be replaced during the current fiscal year.”
These are not identical claims.
The first concerns whether replacement is necessary.
The second introduces an additional judgment about timing.
The second conclusion therefore requires additional supporting reasons.
Argument Analysis requires careful identification of precisely what is being claimed.
Identifying Implicit Assumptions — The Toulmin Model
Not every premise in an argument is explicitly stated.
Speakers often omit assumptions they regard as obvious or believe to be shared with their audience.
The Toulmin Model provides a useful framework for examining such assumptions.
In his 1958 book The Uses of Argument, Stephen Toulmin proposed a model for analyzing arguments as they occur in practical situations.
Its six commonly recognized components are as follows.
| Component | Function |
|---|---|
| Claim | The conclusion being supported |
| Grounds / Data | Facts or information supporting the claim |
| Warrant | An assumption or principle that justifies moving from the grounds to the claim |
| Backing | Evidence or theory supporting the warrant |
| Qualifier | The degree of certainty or conditions under which the claim holds |
| Rebuttal | Exceptions or circumstances in which the claim may not hold |
For example, consider the following claim:
“The recent increase in defects is probably related to equipment deterioration, so replacement should be considered.”
The argument can be organized using the Toulmin Model.
| Component | Example |
|---|---|
| Claim | Equipment replacement should be considered |
| Grounds | Positioning accuracy has deteriorated while the defect rate has increased |
| Warrant | Deterioration in positioning accuracy can cause dimensional defects |
| Backing | Dimensional measurements, equipment inspection records, processing conditions |
| Qualifier | Provided that equipment deterioration is confirmed as a significant cause |
| Rebuttal | Materials or processing conditions may be the main cause; adjustment may be sufficient |
This representation reveals issues that are not apparent in the simpler argument, “Defects have increased, so the equipment should be replaced.”
The objective is not to force every argument into six predefined fields.
It is to identify which assumptions are necessary to move from the evidence to the conclusion.
Argument Structures — More Than One Supporting Reason
Real arguments often include multiple premises and intermediate conclusions.
Several basic structural patterns can help explain how these elements are connected.
1. Serial Arguments
A serial argument occurs when one conclusion becomes a premise supporting another conclusion.
For example:
“The positioning accuracy of Equipment A has deteriorated.”
↓
“The deterioration may be contributing to an increase in dimensional defects.”
↓
“Equipment adjustment or replacement should be considered.”
Here, an intermediate conclusion supports the next stage of reasoning.
If the intermediate inference is not adequately supported, the final conclusion may also be weakened.
2. Convergent Arguments
A convergent argument contains several reasons that independently support the same conclusion.
For example, a proposal to replace equipment might be supported by three considerations:
- The frequency of failures is increasing.
- Maintenance costs are rising.
- Replacement parts are becoming difficult to obtain.
Each reason may provide some support for the conclusion even if the other reasons are not established.
However, having multiple pieces of evidence does not necessarily mean that they represent independent sources.
For example, several reports may all rely on the same original investigation.
Such reports do not necessarily constitute independent confirmation.
3. Linked Arguments
In a linked argument, two or more premises work together to support a conclusion.
For example:
“The positioning accuracy of Equipment A has deteriorated.”
“Deterioration in positioning accuracy has been confirmed to cause dimensional defects in this particular product.”
The first premise alone is insufficient to establish that the equipment is responsible for the defects.
The second premise alone does not show that Equipment A is actually causing the problem.
Together, they provide support for the conclusion that deterioration in Equipment A’s positioning accuracy is contributing to the defects.
This illustrates the importance of distinguishing independently supporting premises from premises that function together as a single inferential unit.
Argument Maps
An Argument Map is a visual representation of claims, supporting reasons, objections, and their relationships.
Claims and premises are typically represented as nodes connected by lines or arrows indicating support or opposition.
For example, an argument about replacing equipment could be represented as follows.
| Element | Content | Relationship |
|---|---|---|
| Claim A | Replacement of Equipment A should be considered | Central claim |
| Reason B | Failures are becoming more frequent | Supports A |
| Reason C | Repair expenses are increasing | Supports A |
| Reason D | Replacement parts are becoming difficult to obtain | Supports A |
| Objection E | Adjustment or partial repair may be sufficient | Weakens the case for replacement |
| Objection F | Replacement costs may exceed the expected benefits | Challenges A |
| Investigation G | Compare the costs and benefits of replacement and repair | Additional analysis needed |
This representation makes it easier to see which reasons support a claim and where objections exist.
However, Argument Maps are not identical to Logic Trees.
Logic Trees are primarily used to decompose subjects, problems, or issues into constituent parts.
Argument Maps represent how claims are justified, challenged, or otherwise related through reasoning.
Moreover, an Argument Map is not necessarily a simple hierarchy.
One premise may support multiple conclusions, and several arguments may support or attack one another.
Deductive and Non-Deductive Arguments
When evaluating arguments, it is important to identify the type of inference connecting the premises with the conclusion.
Representative forms of reasoning include Deduction, Induction, Abduction, and Analogy.
| Type | Characteristics | Evaluation Criteria |
|---|---|---|
| Deduction | If the premises are true and the inference is valid, the conclusion must be true | Logical validity and truth of premises |
| Induction | Infers general patterns or tendencies from observations | Sampling quality and strength of generalization |
| Abduction | Develops hypotheses explaining observed phenomena | Explanatory power, alternatives, and testability |
| Analogy | Infers possibilities from correspondences with other examples | Relational correspondence, differences, and applicability conditions |
For example, a statistical investigation may identify an association between a particular factor and the defect rate.
However, statistical association alone does not establish causation.
Conversely, when regulatory requirements are clearly defined and a product satisfies the specified conditions, a deductive determination may be possible.
Argument Analysis must therefore use evaluation criteria appropriate to the kind of inference involved.
It is not appropriate to evaluate every practical argument as though it were a deductive proof.
Evaluating Arguments — Reliability, Relevance, and Sufficiency
When evaluating an argument, it is useful to distinguish at least three aspects.
First, are the premises acceptable?
Second, are those premises relevant to the conclusion?
Third, do they provide sufficient support for the conclusion?
In Informal Logic, these considerations are commonly organized as Acceptability, Relevance, and Sufficiency.
| Criterion | Question |
|---|---|
| Acceptability | Are the premises based on facts, credible reports, or reasonable assumptions? |
| Relevance | Do the reasons actually bear on the conclusion? |
| Sufficiency | Is the available evidence sufficient to support the conclusion? |
For example, the fact that failures are becoming more frequent on Equipment A does not necessarily establish that replacement is required.
If the failures are minor and can be resolved through inexpensive adjustment, the case for replacement may be weak.
Conversely, if failures create serious safety risks, the argument for replacement may be considerably stronger.
Evaluating arguments therefore requires examining actual conditions, not merely their formal structures.
Argumentation Schemes — Common Patterns of Reasoning
Many practical arguments rely on patterns of reasoning that are reasonable under certain conditions but do not guarantee their conclusions deductively.
Douglas Walton, Christopher Reed, and Fabrizio Macagno examined numerous such patterns in their 2008 book Argumentation Schemes.
These patterns are called Argumentation Schemes.
| Argumentation Scheme | Typical Reasoning |
|---|---|
| Argument from Expert Opinion | A relevant expert supports a claim, providing a reason to regard it as credible |
| Argument from Cause to Effect | If A causes B, the occurrence of A may support an expectation that B will occur |
| Argument from Analogy | A relationship established in one situation may apply to another sufficiently similar situation |
| Argument from Consequences | An action is supported because it is expected to produce desirable consequences |
| Practical Reasoning | Action A is recommended because it is necessary or useful for achieving a particular goal |
These arguments do not necessarily produce deductively certain conclusions.
However, that does not mean they are automatically fallacious.
For example, relying on relevant expert knowledge may be reasonable when making specialized judgments.
Nevertheless, experts can be mistaken, and their judgments may sometimes extend beyond their areas of expertise.
This leads to an important feature of argumentation schemes: Critical Questions.
Evaluating Arguments Through Critical Questions
Each argumentation scheme has characteristic assumptions, limitations, and possible weaknesses.
These can be examined through questions appropriate to the particular form of reasoning.
For example, arguments from expert opinion can be evaluated using the following questions.
| Critical Question | What to Examine |
|---|---|
| Expertise | Is the speaker an expert in the field relevant to the claim? |
| Evidential Basis | What information supports the expert’s judgment? |
| Other Experts | Is there substantial disagreement among relevant experts? |
| Conflicts of Interest | Could the expert’s judgment be influenced by relevant interests? |
| Applicability | Does the opinion apply to the specific problem and conditions being considered? |
The objective is neither to accept a claim unconditionally because an expert supports it nor to reject it automatically as an appeal to authority.
The important question is under what conditions expert opinion provides a reasonable basis for judgment.
Similarly, arguments from analogy can be examined by asking which relationships correspond between the source and target, and whether important differences undermine the inference.
Practical reasoning can be examined by asking whether alternatives exist, whether the proposed action is likely to achieve the desired outcome, and whether undesirable consequences may occur.
Identifying the argumentation scheme helps make the questions requiring investigation more specific.
Different Types of Objections
Argument Analysis considers not only supporting reasons but also the structure of objections and counterarguments.
However, objections can target different parts of an argument.
Consider the following:
“The increase in Equipment A’s temperature coincided with an increase in the defect rate. Therefore, the temperature increase is probably responsible for the higher defect rate.”
Several kinds of objections are possible.
| Target of Objection | Example |
|---|---|
| Premise | Production records show that the temperature and defect-rate increases did not occur during the same period |
| Inference | Coincidence in timing does not necessarily establish causation |
| Alternative Hypothesis | The material lot was changed during the same period |
| Scope of Conclusion | Even if temperature was a contributing factor, it may not have been the principal cause |
These objections are not equivalent.
Challenging the truth of a premise is different from challenging the inference connecting the premise to the conclusion.
Likewise, proposing an alternative cause does not necessarily establish that the original causal hypothesis is false.
Multiple factors may operate simultaneously.
Argument Analysis should clarify precisely what an objection challenges and what it does not establish.
Logical Fallacies
Argument Analysis often involves examining whether reasoning contains common errors.
These are frequently studied as Logical Fallacies.
Representative examples include the following.
| Fallacy or Problem | Description | Example |
|---|---|---|
| Hasty Generalization | Drawing a broad conclusion from too few examples | One company failed to implement the technology, so the technology is unusable |
| False Dilemma | Presenting only two alternatives when others exist | We must either retain the existing system unchanged or abandon it completely |
| Circular Reasoning | Assuming the conclusion within the premises | This method is superior because it is the best method |
| Ad Hominem | Attacking a speaker instead of adequately addressing the argument | The person lacks practical experience, so the claim must be wrong |
| Straw Man | Distorting an opponent’s claim to make it easier to attack | Interpreting a proposal to reconsider research methods as a demand to abolish all information collection |
| Confusing Correlation with Causation | Treating an observed association as sufficient proof of a causal relationship | Profit increased after AI adoption, so AI must have caused the increase |
However, assigning fallacy labels should not become the primary purpose of Argument Analysis.
For example, while attacking a speaker instead of addressing an argument may be fallacious, examining a source’s expertise or conflicts of interest may be appropriate when evaluating the credibility of testimony.
Similarly, drawing conclusions from limited evidence is not always unreasonable. Practical decisions sometimes have to be made when complete information is unavailable.
The important issue is whether a particular inference is appropriate under the relevant conditions.
Practical Example — A Proposal for R&D Theme Exploration
Consider the following argument about R&D theme exploration:
“Even though we collect large quantities of academic papers, patents, and other technical information, we are not discovering enough new R&D themes. Therefore, we should introduce a method for exploring technological principles across different domains in addition to our conventional information collection process.”
We can analyze this argument in several stages.
1. Separate the Claim from Its Supporting Reasons
| Element | Content |
|---|---|
| Claim | A cross-domain technological exploration method should be introduced |
| Grounds | Existing information collection has not generated enough new R&D themes |
| Implicit Assumption 1 | The current approach may be insufficient for discovering certain kinds of themes |
| Implicit Assumption 2 | Cross-domain exploration may generate different candidate themes |
| Implicit Assumption 3 | The value of introducing the additional approach may justify its cost |
| Verification Required | Comparison with existing methods, quality of candidates, feasibility, and cost-effectiveness |
This analysis shows that the conclusion does not necessarily follow from the observation alone.
Even if insufficient new R&D themes are being generated, the principal problem may not be the exploration method.
For example, the organization might have overly restrictive evaluation criteria or management might be unwilling to invest in unfamiliar technological fields.
2. Examine Alternative Explanations
| Hypothesis | Possible Explanation |
|---|---|
| Hypothesis A | Information is being collected and analyzed, but cross-domain hypothesis formation is insufficient |
| Hypothesis B | New themes are identified but rejected during evaluation |
| Hypothesis C | The range of permitted R&D fields is too narrowly restricted |
| Hypothesis D | The organization lacks the technological or organizational capabilities required to pursue newly identified themes |
These hypotheses are not necessarily mutually exclusive.
Several factors may contribute simultaneously.
It would therefore be premature to conclude that changing the exploration method alone will solve the problem.
3. Transform the Argument into a Testable Proposal
A more concrete proposal might be:
“While maintaining the existing information collection and analysis process, we could apply cross-domain exploration to a limited number of technical challenges and compare the resulting candidates with those produced through conventional methods in terms of variety, novelty, and technical feasibility.”
This proposal makes it possible to investigate whether cross-domain exploration adds value.
However, exploration outcomes may depend on how problems are formulated and how candidate themes are evaluated.
Appropriate comparison conditions would therefore be necessary.
Argument Analysis can thus be used not merely to support or reject a claim but also to transform an ambiguous argument into a concrete investigation or experiment.
A Practical Approach to Argument Analysis
There is no single mandatory procedure suitable for every written argument or discussion.
However, distinguishing the following activities can make Argument Analysis more practical.
| Stage | Activity | Main Question |
|---|---|---|
| 1. Identify the Claim | Clarify what conclusion is being proposed | Is the conclusion sufficiently specific? |
| 2. Extract the Grounds | Identify information supporting the conclusion | Are facts and opinions being distinguished? |
| 3. Analyze the Inference | Examine how the grounds lead to the conclusion | Are there unjustified inferential steps? |
| 4. Identify Implicit Assumptions | Examine the premises and conditions that have been omitted | Are these assumptions reasonable? |
| 5. Evaluate the Evidence | Check sources, credibility, and relevance | Is the evidence sufficient? |
| 6. Examine Objections | Consider alternative explanations and exceptions | Which parts of the argument are weakened by objections? |
| 7. Reconsider the Conclusion | Revise its strength and scope | What conclusion is actually supported? |
The objective is not to classify every statement as simply true or false.
For example, evaluating the claim that “a particular technology is promising” requires clarifying what promising means.
A technology may offer high performance while having limited market demand.
A market may be growing while the company lacks a sustainable competitive advantage.
Clarifying the conditions under which a conclusion holds is therefore an important part of Argument Analysis.
Argument Analysis and Logical Thinking
Argument Analysis is closely related to Logical Thinking, but they are not synonymous.
Logical Thinking is a broad approach to reasoning, organizing information, drawing conclusions, and communicating explanations coherently.
Argument Analysis focuses more specifically on the structure and evaluation of claims, evidence, premises, inferences, and objections.
| Perspective | Logical Thinking | Argument Analysis |
|---|---|---|
| Main Subject | Reasoning, problems, information, and explanations | Claims, premises, evidence, inferences, and objections |
| Main Purpose | Think and communicate logically | Clarify and evaluate argumentative structures |
| Representative Methods | MECE, Logic Trees, Pyramid Principle | Toulmin Model, Argument Maps, Argumentation Schemes |
| Main Output | Organized reasoning or explanations | Argument structures, weaknesses, and verification questions |
For example, the Pyramid Principle may help organize a logical explanation by arranging conclusions and supporting reasons.
Argument Analysis can then be used to examine whether the explanation contains unjustified inferences.
However, a logically well-supported argument is not necessarily understood by its intended audience.
Differences in background knowledge, interests, and shared assumptions must also be considered.
Argument Analysis and Critical Thinking
Argument Analysis is an important component of Critical Thinking.
Critical Thinking involves examining evidence, premises, reasoning, alternative explanations, and other considerations rather than accepting information uncritically.
Argument Analysis focuses particularly on making the relationships between claims and supporting reasons explicit.
However, Critical Thinking also includes issues that extend beyond argumentative structure, such as source reliability, cognitive biases, uncertainty, and decision criteria.
For example, a scientific paper may present a logically coherent conclusion, but its findings may still be unreliable because of problems with measurement methods or sampling.
Similarly, if a paper cites data that do not actually exist, analyzing the formal structure of its argument may not reveal the problem.
Argument Analysis does not replace fact-checking or scientific evaluation.
Argument Analysis, Discussion, and Consensus Building
Argument Analysis is also useful when several people exchange different opinions.
For example, suppose some managers support investing in a particular business while others oppose the proposal.
Supporters may emphasize market growth.
Opponents may emphasize uncertainty about investment recovery.
The disagreement does not necessarily result from an error in either side’s reasoning.
Participants may assign different importance to the same information or apply different risk tolerances and decision criteria.
Argument Analysis can organize the claims and supporting reasons offered by each side.
However, it does not automatically determine which values or priorities should prevail.
Pragma-Dialectics is one theoretical approach to studying argumentation in the context of disagreement.
It examines rules and procedures through which differences of opinion can be addressed through reasonable discussion.
An important concern is understanding opposing positions accurately and evaluating disputed claims through reasons rather than simply insisting on one’s preferred conclusion.
Argument Analysis can therefore help clarify the actual points of disagreement instead of treating discussion merely as a contest with winners and losers.
Computational Argumentation — Representing Attacks and Support
Computer science has developed formal approaches to representing relationships among arguments.
A representative contribution is the Abstract Argumentation Framework introduced by Phan Minh Dung in 1995.
This framework treats a collection of arguments and the attack relations among them as its fundamental elements.
For example, Argument B may attack Argument A, while Argument C attacks B.
The important idea is that it becomes possible to investigate which arguments may be accepted under different formal criteria, based on the structure of their attack relations.
However, an Abstract Argumentation Framework does not automatically evaluate the factual truth of each argument or the reliability of its evidence.
Frameworks incorporating support relations have also been developed, but attack relations are central to Dung’s original abstract framework.
These studies provide a foundation for computationally representing situations in which multiple arguments challenge one another.
Argument Mining
Argument Mining is a research field concerned with automatically identifying argumentative components and relationships in natural-language text.
For example, it may attempt to extract the following information from news articles, academic papers, reports, or debate transcripts:
- What claims are being made?
- Which statements provide support for those claims?
- Which arguments challenge other arguments?
- Which assumptions are being used implicitly?
- Which arguments concern the same issue?
Argument Mining research has investigated methods for detecting argumentative passages, classifying their roles, and identifying relationships among them.
Marco Lippi and Paolo Torroni reviewed important methods, applications, and research challenges in their 2016 survey, “Argumentation Mining: State of the Art and Emerging Trends.”
However, extracting a statement that appears to be a claim is different from verifying whether that claim is supported by correct evidence.
Natural-language statements may also perform multiple argumentative functions.
Whether a particular sentence is a premise or a conclusion may depend on the broader context.
Argument Mining can therefore require more complex semantic analysis than ordinary sentence classification.
LLM-Assisted Argument Analysis
Large Language Models can help extract claims and supporting reasons from natural-language documents.
They can also suggest implicit assumptions and possible counterarguments.
For example, an AI system might analyze a business report and organize its claims, grounds, warrants, conclusions, and objections into a table.
It might also extract multiple arguments from a document and represent their supporting and opposing relationships in an Argument Map.
| AI-Assisted Function | Main Operation |
|---|---|
| Claim Extraction | Identify conclusions and proposals in text |
| Evidence Extraction | Identify statements or data offered in support of claims |
| Implicit Assumption Identification | Suggest premises needed to connect evidence with conclusions |
| Inference Classification | Identify deduction, induction, abduction, analogy, and other patterns |
| Argument Mapping | Organize support and opposition relationships |
| Alternative Hypothesis Generation | Suggest other explanations consistent with the available facts |
| Critical Question Generation | Produce questions appropriate to the argumentation scheme |
| Evidence Alignment | Connect claims with original documents and data |
| Explanation Reconstruction | Rewrite explanations to clarify their reasoning and support |
However, LLM-assisted Argument Analysis requires caution.
For example, an assumption inferred by an LLM may not be what the original author intended.
Likewise, an AI-generated objection may misinterpret the original claim.
Rewriting an argument into more persuasive language may also make uncertain reasoning appear better supported than it actually is.
It is therefore important to distinguish information explicitly present in the original document from interpretations and suggestions generated by AI.
How Should AI-Assisted Argument Analysis Be Recorded?
When Argument Analysis is used repeatedly, structured storage of the results may be useful.
For example, the following fields could be maintained.
| Field | Description |
|---|---|
| Argument ID | Identifier for an individual argument |
| Claim | The conclusion being proposed |
| Grounds | Information supporting the claim |
| Source | Original document or underlying data |
| Warrant / Implicit Assumption | Conditions connecting the grounds with the claim |
| Inference Type | Deduction, induction, abduction, analogy, etc. |
| Support Relations | Which other claims the argument supports |
| Attack Relations | Which claims or inferences it challenges |
| Unresolved Questions | Issues requiring further investigation |
| Verification Status | Confirmed, unverified, potentially refuted, etc. |
It may also be useful to distinguish AI-generated inferences from claims entered by human users.
For example, a claim directly extracted from an original document could be stored as a Source-Extracted Claim.
An assumption inferred by AI could be labeled as an Inferred Assumption.
A counterargument generated by AI could be labeled as a Generated Counterargument Candidate.
Such distinctions reduce the risk of treating AI-generated interpretations as established facts from the original source.
They could also support re-evaluation of earlier arguments when new evidence becomes available.
Argument Analysis and Knowledge Graphs
Argument Maps can be represented through network structures resembling Concept Maps and Knowledge Graphs.
However, the relationships represented by these methods differ.
For example, a Knowledge Graph might contain the relationship:
“Material M has high-temperature resistance.”
An Argument Map, by contrast, would represent which experimental findings support this claim and under what conditions objections or counterevidence exist.
| Representation | Main Relationships |
|---|---|
| Concept Map | Semantic relationships between concepts |
| Knowledge Graph | Structured relationships involving entities and facts |
| Argument Map | Claims, evidence, support, opposition, and inference |
Suppose one academic paper reports that Material M remains stable under high-temperature conditions.
Another paper reports deterioration under particular environmental conditions.
If only general relationships between concepts are stored, the difference between the two findings may become unclear.
Argument Analysis can represent the conditions and evidence supporting each claim.
This may be useful when comparing statements extracted from multiple documents and identifying apparent contradictions or unresolved questions.
However, different claims are not necessarily contradictory.
If their conditions or対象 differ, both may be correct.
Limitations of Argument Analysis
Argument Analysis is useful, but it does not automatically produce correct decisions.
First, an argument may be structurally coherent while relying on false information. Argument Analysis does not replace fact-checking or experimentation.
Second, claims and assumptions are not always explicit. Different analysts may interpret the same statement differently.
Third, what counts as sufficient evidence depends on the nature of the problem. Mathematical proofs, R&D hypotheses, and business decisions require different levels and forms of justification.
Fourth, formalizing an argument may remove relevant context or obscure the speaker’s intentions.
Fifth, organizing extensive supporting reasons and objections does not automatically determine which decision should be adopted. Value judgments, risk tolerance, costs, and organizational constraints may all matter.
Sixth, using Argument Analysis exclusively to find errors in other people’s reasoning may make discussion and collaboration more difficult.
The same analytical standards should also be applied to one’s own arguments.
Clarifying the structure of an argument is one condition for improving judgment, but it does not guarantee that the conclusion is correct.
Argument Analysis in ThinkNavi
ThinkNavi treats Argument Analysis as a practical method for supporting Logical Thinking and Critical Thinking.
MECE and Logic Trees are useful for organizing information and decomposing problems.
Argument Analysis examines how the organized information supports particular conclusions.
For example, claims about a technology could be extracted from multiple academic papers and technical reports, together with the evidence and assumptions on which those claims depend.
For ThinkNavi Free, possible implementations include worksheets and Argument Maps for manually organizing claims, grounds, warrants, and counterarguments.
Potential LLM-assisted functions include extracting argumentative structures from documents, proposing implicit assumptions, identifying argumentation schemes, generating critical questions, and organizing conflicting claims.
When multiple documents are involved, the system could compare different claims about the same issue and investigate the evidence supporting each position.
For example, academic papers, patents, corporate announcements, and industry reports may express different views about the prospects of a particular technology.
Their claims and supporting reasons could be organized separately for analysis.
For systems such as ConceptMiner R&D, which generate hypotheses from technical information, it is particularly important to distinguish the evidence supporting a hypothesis from the parts that remain speculative.
Relationships among technical elements, hypotheses, source documents, objections, and verification results could be represented in a structured form.
This would make it possible to trace how hypotheses were generated and how extensively they have been evaluated.
However, the support relations identified through Argument Analysis do not establish technological feasibility.
Specialized investigation and experimentation are still required.
Furthermore, support and attack relations in Argument Analysis differ from similarity relationships generated in feature spaces through methods such as Growing Neural Gas (GNG).
If these approaches are combined, the meaning of each type of relationship must be preserved.
These are potential directions for future development and do not imply that all such functions are currently available.
The purpose of computational Argument Analysis should not be to replace human discussion entirely with formal rules.
Instead, it should make supporting reasons, implicit assumptions, and possible objections visible so that judgments can be improved as additional information becomes available.
Argument Analysis is a method for clarifying not only what is being claimed but also why it is believed. It can help examine our own reasoning and that of others, identify uncertainty, and determine what should be investigated next.
Related Topics
- Logical Thinking — Building Sound Arguments, Communicating Reasoning, and Achieving Mutual Understanding
- Critical Thinking | Examining Claims and Evidence
- Induction, Deduction, and Abduction — Discovering Patterns, Drawing Conclusions, and Generating New Hypothese
- The Scientific Method and Falsifiability — Generating Hypotheses and Testing Them Critically
- Problem Structuring — Breaking Down Problems, Understanding Relationships, and Finding Paths to Solutions
- Logic Trees and Issue Trees — Breaking Down Problems and Identifying the Questions That Need Answers
- MECEMECE — Logical Thinking Through Mutually Exclusive and Collectively Exhaustive Classification, and Its Limitations
- Root Cause Analysis (RCA) — Understanding Why Problems Occur and Preventing Recurrence
- Concept Maps — Representing Relationships Between Concepts and Understanding Knowledge Structures
- Analogy — Connecting Knowledge Across Domains to Generate New Ideas and Hypotheses
- Structure-Mapping Theory | Understanding Relational Correspondences and Generating New Inferences Across Domains
- Epistemology | What Can We Know, and How?
- Cognitive Biases
- Decision Matrices
- Design Thinking — From Human-Centered Problem Discovery to Ideation, Prototyping, and Testing
- Concept Exploration
References and Further Reading
- Stanford Encyclopedia of Philosophy — Argument and Argumentation
- Stanford Encyclopedia of Philosophy — Informal Logic
- Stephen E. Toulmin, The Uses of Argument, Cambridge University Press, 1958.
- Purdue Online Writing Lab — Toulmin Argument
- Douglas Walton, Christopher Reed, and Fabrizio Macagno, Argumentation Schemes, Cambridge University Press, 2008.
- Frans H. van Eemeren and Rob Grootendorst, A Systematic Theory of Argumentation: The Pragma-Dialectical Approach, Cambridge University Press, 2004.
- Ralph H. Johnson and J. Anthony Blair, Logical Self-Defense, McGraw-Hill Ryerson, 1977.
- Trudy Govier, A Practical Study of Argument, Wadsworth, 2010.
- Douglas N. Walton, Fundamentals of Critical Argumentation, Cambridge University Press, 2006.
- Phan Minh Dung, “On the Acceptability of Arguments and Its Fundamental Role in Nonmonotonic Reasoning, Logic Programming and N-Person Games,” Artificial Intelligence, 77(2), 1995, pp. 321–357.
- Philippe Besnard and Anthony Hunter, Elements of Argumentation, MIT Press, 2008.
- Marco Lippi and Paolo Torroni, “Argumentation Mining: State of the Art and Emerging Trends,” ACM Transactions on Internet Technology, 16(2), 2016.
- Chris Reed and Glenn Rowe, “Araucaria: Software for Argument Analysis, Diagramming and Representation,” International Journal on Artificial Intelligence Tools, 13(4), 2004, pp. 961–979.
- Purdue Online Writing Lab — Organizing Your Argument