Term 7 · Module 3 of 8

Constructing and Evaluating Arguments

Critical Thinking

Practise this module in Claude, ChatGPT or Gemini

Set it up free
Mind map

Building Persuasive Business Arguments

Stories vs. arguments

A business narrative (the SCR framework: situation, complication, resolution; mapped to context, conflict, choice, consequences) is not a pile of facts. It is a carefully constructed account that steers an audience towards a particular way of seeing a problem and therefore a particular solution. The resolution stage is where description turns into recommendation, and that is where storytelling becomes an argument.

  • A founder does not just say a market exists; she argues it is worth entering.
  • A consultant argues for one strategy over another.
  • A product manager argues a feature deserves funding.
  • Even an email asking a manager to approve a proposal is an argument.

Stories capture attention; arguments justify decisions. Business almost always needs both.

Why this matters in the age of AI

ChatGPT, Gemini, Claude and Copilot produce polished, well-organised reports, memos, copy and investment summaries in seconds, and can even recommend actions. Technology changed the speed of production, not the standards for judging. Beneath polished writing, always ask:

  1. What is the claim?
  2. What evidence supports it?
  3. What reasoning connects the evidence to the conclusion?
  4. What assumptions make it look solid?

Case: Netflix in 2006

In hindsight the Netflix streaming story looks inevitable (Netflix saw the future; Blockbuster did not; streaming replaced DVDs). In 2006 it was not:

  • DVD-by-mail was profitable, growing steadily and well understood.
  • Streaming was expensive, unreliable for many households, and broadband was not yet widespread.
  • Investing in streaming meant pouring resources into a business that was not making money while risking the profitable one.

"Streaming is the future" is neither evidence nor an argument. The board needed:

Part Netflix content
Claim Netflix should invest aggressively in streaming technology
Evidence Rising broadband adoption, falling data transmission costs, changing media consumption patterns, projections of future consumer behaviour
Reasoning Why do these trends justify such a costly investment now (2006) rather than waiting another 5 years? This logical bridge is what persuaded the board

The argument came first; the heroes-and-villains story came later, after the decision succeeded. Successful decisions are retold as stories with clear heroes, villains, turning points and neat endings. A future manager must not only communicate clearly but justify conclusions and recommendations.

Key takeaways

  • Resolution is where narrative becomes argument; stories attract attention, arguments justify decisions.
  • AI changed production speed, not the standards for judging arguments.
  • Always ask: claim, evidence, reasoning, assumptions.
  • "X is the future" is not an argument; the reasoning bridge is.
  • Hindsight stories hide the uncertainty the original argument had to overcome.

Rhetorical Appeals: Ethos, Pathos, Logos

Some arguments feel convincing though weak; others with good evidence fail. Data rarely speaks for itself: audiences must decide whether data is trustworthy, whether it matters, and whether it justifies the recommendation. So persuasion begins before anyone has fully evaluated the logic.

Aristotle (ancient Athens; citizens argued their own cases in court and debated policy in the assembly) observed that people listen because they trust the speaker, care about the issue, and find the reasoning makes sense. His treatise on rhetoric names three modes of persuasion. Every pitch, campaign or launch still uses some mix.

Appeal Question it answers Concerns
Ethos Why should I trust you? Credibility, trust
Pathos Why does this matter? Emotion, stakes
Logos Why should I believe this conclusion follows from the evidence? Logic, evidence linked to conclusion

Rhetorical Appeals: Ethos, Pathos, Logos

Ethos: credibility

Audiences make a quick judgment about the speaker before evaluating evidence, and that judgment influences the decision. Same investment advice from a friend with no financial background versus an analyst with about 20 years' industry experience: same evidence, different source, and most would trust the analyst.

  • Credibility = experience, expertise, reputation, transparency.
  • It can be earned through behaviour; admitting uncertainty or mistakes can raise trust (a company acknowledging mistakes is often more trustworthy than one claiming it never errs).

Patagonia, "Don't Buy This Jacket": on Black Friday (associated with consumer spending) the company ran a full-page newspaper ad with this headline. It explained that every jacket carries hidden environmental costs (water, energy, raw materials) and that unnecessary consumption damages the environment. The ad had little technical evidence and few statistics. It persuaded because it strengthened ethos: even at the risk of lower sales, it stayed consistent with its environmental values.

Pathos: emotion

Emotion is almost always part of decision-making (even choosing coffee or chai). Business teaching often treats numbers as enough, but data rarely speaks for itself. Employees rarely support organisational change because a spreadsheet says so; they support it when they understand what is at stake.

  • Pathos is not manipulation; it helps the audience see why an issue matters or deserves attention.

Logos: logic

Good logos connects the data logically to a recommendation and explains why the facts support the conclusion. Piling on charts, graphics and statistics without a convincing argument can overwhelm the audience.

Case: Amazon Prime combining all three

Prime asked customers to pay an annual fee upfront for benefits not yet experienced; not an obvious proposition.

Appeal How Amazon used it
Ethos By the mid-2000s Amazon had a reputation for reliable, on-time delivery, fair returns and continual improvement; this made the new promise believable
Logos A simple economic calculation: order often and the subscription quickly pays for itself with unlimited fast shipping; pay once, save repeatedly. Deliberately rational
Pathos Prime sold convenience, not shipping: no hesitating over delivery charges, no waiting days for standard shipping; a smoother, less frustrating experience (subtle but powerful)

Success was not inevitable. It worked because credibility, logic and emotion were combined in the right balance; none alone would have sufficed. The appeals are interconnected, not separate ingredients.

Generative AI and the appeals

An AI market-entry report arrives polished, with statistics, narrative and spreadsheets.

  • Ethos: do you trust it merely because it sounds confident? Have you independently verified sources?
  • Pathos: AI writes emotionally engaging language very well, but that does not make a recommendation more accurate.
  • Logos: does the reasoning genuinely connect evidence to conclusion, or does it just place plausible facts beside an equally plausible recommendation?

The three questions for any product launch, pitch or AI presentation: Why trust the speaker (ethos)? Why care (pathos)? Why believe the conclusion follows (logos)? They reveal how an argument persuades, not whether it is correct. Persuasion and proof are not the same.

Key takeaways

  • Ethos = credibility, pathos = emotion/stakes, logos = logic; strong persuasion balances all three.
  • Credibility is earned via expertise, reputation, transparency and consistency with values (Patagonia).
  • Pathos is legitimate when it helps an audience see what is at stake; it is not manipulation.
  • Amazon Prime: ethos (reliability), logos (pays for itself), pathos (convenience).
  • AI fluency can fake expertise; persuasion is not proof.

Building Blocks of an Argument

Every argument has three parts; if one is missing it is much weaker.

  1. 1Evidence: why believe it?
  2. 2Reasoning: the logical bridge
  3. 3Claim: what to believe or do

Building Blocks of an Argument

Claim

The claim is the conclusion the speaker wants the audience to believe or act on (e.g. "we should launch this product", "enter the Indonesian market", "this campaign was successful"). It may be implied (ads rarely say "buy this" or "invest in this fund"), but if the audience must guess the recommendation, the argument is harder to follow.

  • Test: What exactly is the author asking me to believe or do? If unclear, the argument is probably unclear.

Evidence

Evidence is information offered in support of the claim: financial data, customer research, market reports, experiments, surveys. It is not limited to numbers: qualitative evidence (interviews, observation) counts. A customer's interview about cancelling a subscription may reveal more than sales figures.

  • Evidence must be relevant. Arguing for a new Hyderabad office with employee satisfaction scores from the Bangalore office: accurate numbers, but not relevant. Evidence is information that helps answer the question under discussion.

Reasoning (warrant, the logical bridge)

Reasoning explains why the evidence supports the claim. It is the most overlooked part. "Customer satisfaction increased 15%" is evidence, but alone it could support many claims (service improved, a redesign worked, lower prices improved service). Without reasoning, evidence is facts sitting beside a conclusion, and the audience may draw a different connection from the one intended.

Worked example: AI customer support at an online grocer

Part Content
Claim Use AI to handle first-level customer inquiries
Evidence Three-month pilot: average response time fell from 12 to 1 minute; support costs fell 35%; customer satisfaction unchanged
Reasoning If routine inquiries can be handled faster and cheaper without reducing satisfaction, then automating first-level support improves organisational efficiency while maintaining service quality

The reasoning sentence adds no new evidence; it only explains why the evidence leads to the conclusion.

Evidence that does not match the claim

"Our customers love our new app" supported by "one million downloads". Downloads show people were willing to try it, not that they enjoyed it (they might have deleted it after a day, left poor reviews, or downloaded for a discount). The evidence supports a different claim.

AI and the building blocks

AI reports present evidence and conclusions confidently, but confidence is not reasoning. Asked "Should we expand into Vietnam?", it may cite GDP growth, rising internet penetration, a growing middle class and government investment in digital infrastructure, then conclude "enter next year". Missing: why do these facts justify your entry? Your products may not fit local preferences, regulations may create barriers, competitors may dominate. Don't stop at "are the facts correct?"; ask how exactly the facts support the conclusion.

Key takeaways

  • Argument = claim + evidence + reasoning; missing any weakens it.
  • Claims can be implied, but clarity demands making them explicit.
  • Evidence can be qualitative and must be relevant.
  • Reasoning explains why evidence supports the claim; it adds no new evidence.
  • Mismatch between evidence and claim (downloads vs. love) is a common flaw; AI often supplies facts without the bridge.

Types of Reasoning: Deductive and Inductive

Reasoning can run in two directions, and each type is evaluated differently: deduction can offer certainty; induction only probability.

Deductive Inductive
Direction General rule/principle → specific case Specific observations → broader pattern/conclusion
Conclusion Follows necessarily if premises are true and the rule is applied correctly Only probable; never guaranteed
Key question Does it necessarily follow? How likely/strong; could other explanations or new evidence change it?
Business uses Compliance, financial reporting, tax regulations, employment laws, internal policies; rule-based AI such as fraud detection Market research, surveys, forecasts, trends; most strategic decisions

Types of Reasoning: Deductive and Inductive

Deduction

  • All employees must complete cybersecurity training to access company systems; Priya has not completed it; therefore Priya cannot access the systems.
  • Every expense above ₹50,000 requires board approval; the proposal requests ₹75,000; therefore it requires board approval. (Assuming the policy is accurate, the conclusion is not debatable; it is built into the rule.)
  • A fraud-detection algorithm with rules such as "transactions above a certain value from an unfamiliar location require additional verification" follows the rule once conditions are met; human or AI, the logic is the same.

Induction

Electric scooter sales have risen every year for five years, so demand is likely to keep growing. The key word is likely: a technological breakthrough, policy change or recession could change the trend. That does not mean the reasoning was poor; inductive arguments stay open to revision when new evidence appears.

Most business decisions are inductive. Managers cannot know the future; they look for patterns and predict what is likely.

  • Retailer opening a store: study similar locations, population growth, income levels, traffic patterns, purchasing behaviour. These make success more or less probable, never certain.
  • Supermarket plant-based foods: six quarters of rising sales, surveys showing interest in sustainable diets, social media awareness of environmental issues, so allot more shelf space. Not guaranteed: it could be a temporary fashion, inflation could change buying habits, preferences could shift. The conclusion is reasonable, not certain.

Reasoning in AI

  • ChatGPT predicting the next word does not apply rigid lexical or logical rules; it makes probabilistic predictions from patterns in vast text. It reasons inductively, identifying statistical regularities.
  • Strength: fluent, fast responses. Weakness: because it predicts probabilities, it can produce convincing statements that are incorrect or hallucinated. It is not lying; it is extending a pattern beyond what the evidence supports, which is a risk of induction itself.

Case: UPI and digital payments (a 2021 retail CEO)

Digital payments and UPI transactions rising for a decade, smartphone ownership expanding, more merchants accepting digital payments, consumers comfortable with QR codes. Therefore future stores should be designed around digital payments rather than cash. This is inductive (pattern → prediction about the future), and the prediction was largely correct, yet strategy always involves uncertainty.

Why the distinction matters

People expect the wrong level of certainty. "The data proves customers will buy this" should make you pause: business data rarely proves; it suggests, supports, indicates, makes likely (inductive words). Using deductive language ("our research proves", "clearly demonstrates", "guarantees") for inductive conclusions overstates the case; the confidence comes from language, not logic. Ask: is this certain or merely probable?

  • Deduction: ask whether the logic is valid.
  • Induction: ask how strong the evidence is and whether other explanations are possible.

Recognising uncertainty does not weaken business thinking; it makes it more honest.

Key takeaways

  • Deduction: rule → case, conclusion necessary. Induction: observations → pattern, conclusion probable.
  • Most business and strategic decisions, and generative AI, work inductively.
  • Induction is always open to revision; that is not poor reasoning.
  • Overstating certainty with "proves/guarantees" is a red flag.
  • Evaluate deduction by validity, induction by strength of evidence.

The Toulmin Model

Real arguments are messy: they assume shared background, skip the link between evidence and recommendation, sound confident, and rely on unstated, unexamined assumptions. Stephen Toulmin studied how people actually argue (boardrooms, courtrooms, newspapers, home, scientific papers). His model identifies every piece, stated or not. It shifts you from "Do I agree?" to "Which part is weak?" Most business arguments leave out one or two of the six parts unstated, but they are still there.

Part Role 4-day workweek example
Claim Conclusion the speaker wants accepted The company should introduce a 4-day workweek
Grounds Facts supporting the claim (Toulmin's word for evidence); observations, not proof by themselves Productivity up 12% in a 6-month trial; sick leave down; retention improved
Warrant Reasoning: why grounds support the claim; the bridge, the most important and most forgotten part If productivity rises while well-being also improves, adopting the schedule benefits both organisation and staff
Backing Supports the warrant itself when people challenge your reasoning ("how do we know the gain wasn't temporary?"); e.g. peer-reviewed studies, industry reports, longitudinal evidence from other companies. Goes beyond claim-evidence-reasoning Research linking employee well-being, motivation and performance
Qualifier Indicates the strength of the claim; makes arguments more honest, not weaker "Probably", "likely", "under current conditions", "in most cases", "assuming current market trends continue"
Rebuttal Conditions under which the argument would no longer hold: "when would I be wrong?" See below

The Toulmin Model

Microsoft Japan trial

Microsoft tested a 4-day workweek in Japan with full salaries. Reportedly productivity rose significantly, meetings got shorter, electricity costs fell and employee satisfaction improved.

  • Claim: Microsoft should continue exploring more flexible work arrangements.
  • Grounds: higher productivity, lower operational costs, greater employee satisfaction.
  • Warrant: if an organisation improves productivity while cutting costs and increasing employee well-being, changing the schedule benefits the business.
  • Backing: organisational psychology research linking well-being, motivation and performance.

Making the backing explicit explains how the evidence supports the recommendation instead of just "the trial worked".

Qualifier and rebuttal

  • Managers say "this strategy will succeed"; reality is rarely that certain. A doctor claiming "this treatment will definitely work" (100%) invites suspicion; a trustworthy one cites a very high success rate based on available evidence. Confidence is valuable, overconfidence is not.
  • A rebuttal requires being critical of your own thinking; most people defend ideas and few imagine them failing, but experienced leaders do. For AI-powered customer service: "assumes routine inquiries remain simple; if customers ask more complex questions or regulations demand more human oversight, expected savings may not materialise." The recommendation became more credible, showing intellectual honesty by stating limitations before someone else does.

Arguments across professions

The skeleton of an argument is similar; the "cladding" changes because each profession has its own standards for convincing evidence, accepted reasoning and tolerable uncertainty.

Context Features
Law Advocates build arguments to persuade judges and juries. Lawyers rarely dispute the facts, but what they mean (company acted negligently vs. reasonably under the circumstances). Heavy weight on warrants, grounded in established legal principles and precedent
Science Almost never claims absolute certainty; cautious phrasing ("the evidence suggests", "results indicate", "findings are consistent with"). This is intellectual honesty and fundamentally inductive

Key takeaways

  • Toulmin's six parts: claim, grounds, warrant, backing, qualifier, rebuttal.
  • Use it to find which part of an argument is weak.
  • Backing defends the warrant; qualifiers calibrate confidence; rebuttals state when you'd be wrong.
  • Rebuttals and qualifiers raise credibility.
  • Professions differ in standards (law: interpretation and precedent; science: cautious, inductive).

Evaluating Arguments: Validity, Soundness, Strength, Cogency

Example: "All fish live in water. A dolphin lives in water. Therefore, a dolphin is a fish." It feels like it works, yet the conclusion is false (dolphins are mammals). We must separate whether an argument is put together correctly from whether it is true.

Evaluating Arguments: Validity, Soundness, Strength, Cogency

Validity vs. soundness (deductive arguments)

  • Valid: the structure is solid; if the premises are true, the conclusion must be true. Validity is about form, not content, and says nothing about whether premises are true in the real world.
  • Test: is there any possible way for the premises to be true while the conclusion is false? If no, valid; if yes, invalid, however true the premises are.
  • Sound: valid and all premises actually true. The gold standard: airtight structure plus accurate material.

The dolphin argument is invalid: even with both premises true, the conclusion does not follow, because many non-fish live in water (the lecture mentions otters, and loosely submarines). Failing validity, it cannot be sound.

Every sound argument is valid, but not every valid argument is sound. Validity is the minimum bar; soundness is the full package (correct reasoning plus correct facts).

Valid-but-unsound arguments are used constantly in marketing, casual arguments and political rhetoric: a structure that feels logical built on false or unproven premises. Checking only "does the conclusion follow?" and never "are the premises true?" is half the job.

Strength vs. cogency (inductive arguments)

Validity and soundness suit deductive arguments; much everyday reasoning is inductive. Example: it has rained every day this week in Bangalore during monsoon, so it will probably rain tomorrow; a probability claim.

  • Strong: if the premises are true, the conclusion is probably true. Strength comes in degrees.
    • "Every crow I have ever seen is black, therefore all crows are black": moderately strong (reasonable inference from a large pattern).
    • "I saw one black crow yesterday, therefore all crows are black": very weak; same conclusion, much thinner evidence base.
  • Strength, like validity, concerns the relationship between premises and conclusion, so a strong argument can have false premises. Example: "All lottery winners publicly announce within 24 hours; nobody announced winning last week's lottery; therefore probably nobody has claimed the prize." Structurally reasonable even if the first premise is false.
  • Cogent: strong and its premises are actually true (or well supported and reasonable to accept). The inductive gold standard.
Deductive Inductive
Claims Certainty Probability
Structure test Validity Strength (degrees)
Structure + true premises Soundness Cogency

Evaluating Evidence

Unlike validity, which can often be checked cleanly, evaluating evidence is messy and judgment-based, but five questions apply to any piece of evidence.

# Question Notes
1 Relevant? Must bear on the claim. "This phone has a great camera" is not evidence that it is a good investment (unless camera quality is all you want). Weak arguments are often not wrong but beside the point: true statements that don't support the claim
2 Sufficient? One anecdote rarely supports a general claim (one black crow is one data point used for a population claim). Size and diversity of the evidence base matter. "Most people say..." is not a valid sample; it is a cherry-picked case
3 Credible source? Check expertise, track record, and bias or conflict of interest. A health-benefit study funded entirely by the company selling the product is not automatically wrong, but deserves more scrutiny than independent research: it may have expertise but also has something to gain
4 Current? Evidence has a shelf life: a 15-year-old internet-usage statistic or a medical study later revised may now be actively misleading. Ask where and when it is from and what has changed
5 Survives counter-evidence? Strong evidence withstands contrary evidence. An unaddressed body of contrary evidence is a huge red flag. Cherry picking = using only data that fits the conclusion and ignoring data that doesn't

Evaluating Evidence

Putting it all together: an evaluation workflow

The point is not cynicism or trusting nothing; it is telling an argument that earns agreement from one that merely sounds polished, i.e. the difference between feeling persuaded and being justifiably persuaded.

Key takeaways

  • Valid = true premises would force the conclusion (form); sound = valid + true premises. Every sound argument is valid, not vice versa.
  • Strong = true premises make conclusion probable (degrees); cogent = strong + true premises.
  • Evidence checklist: relevant, sufficient, credible (no conflict of interest), current, survives counter-evidence.
  • Cherry picking and anecdotes are classic failures of sufficiency and counter-evidence.
  • First decide whether an argument claims certainty or probability, then pick the matching tests.
Revise before your examAll 5 exam tips from this module in one list
  1. Being persuaded is not the same as being convinced. You can trust the speaker and care about the issue and still be handed a weak argument; hence the need to examine argument structure.
  2. Reasoning (warrant) is the most commonly missing piece. If evidence and conclusion merely sit side by side, name the missing bridge.
  3. Inductive conclusions are "likely", not "proven". Spot deductive language ("proves", "guarantees") attached to inductive evidence.
  4. Claim-evidence-reasoning versus Toulmin: Toulmin adds backing (supports the warrant), qualifier (strength) and rebuttal (limits). Grounds = evidence; warrant = reasoning.
  5. Soundness : validity :: cogency : strength. Validity is all-or-nothing; strength comes in degrees.

Test yourself on this module

Questions from these notes, one at a time, with feedback. Your weak topics are saved.

Set it up free Takes about a minute. Free plans work.