Seeing Is Not Always Believing
We tend to treat what we see, read and hear (a graph, a dashboard, a news piece, a video) as a direct picture of reality. It is not. Data never speaks for itself: people interpret it, ask questions of it, make assumptions about it and construct explanations from it. Because people are involved, errors in reasoning are inevitable.
Seeing Is Not Always Believing
The dashboard trap
A business analyst at an online retailer sees website traffic up 30% over the weekend. Tempting conclusion: the marketing campaign worked. But the increase could be caused by bots, a technical error in the analytics platform, or a viral social post that generated curiosity but no purchases. The numbers and the graph are real; the conclusion may still be entirely wrong.
Why we reason badly
Humans are remarkably poor at evaluating information objectively:
- We notice evidence that supports our opinions and ignore evidence that challenges them.
- We jump to conclusions on very little information.
- We rely on heuristics (mental shortcuts), which are often useful but sometimes misleading.
- The smarter we are, the better we often become at defending existing beliefs rather than questioning them.
So critical thinking is not simply about intelligence; it is about the intellectual discipline to question our own existing beliefs.
Case: Target's pregnancy-prediction model (2012)
Statistician Andrew Pole and his team at Target built a predictive model to identify pregnant customers so that promotions could arrive exactly when new buying habits were forming.
- Signal: subtle purchase shifts, e.g. from regular scented lotion and makeup to unscented lotion, calcium and magnesium supplements.
- Output: a pregnancy prediction score, indicating likely first/second trimester, with coupons timed to each stage.
- Failure: a man outside Minneapolis complained that his teenage daughter was receiving baby-clothes, crib and maternity mailers (the story became a New York Times investigation). The manager apologised, not knowing what had triggered the mailers. Days later the father admitted she was pregnant, due in August. The model knew before he did.
The model was technically correct, the data real, the analysis rigorous, the predictions accurate. What failed was the thinking around the analytics:
| Question asked | Questions neglected |
|---|---|
| Technical: "Can we predict this with great precision?" | Human: "Should we?" What are the second-order effects of being this intrusive? What does it do to trust and to the company-customer relationship? |
Target's response was to insert irrelevant ads (tools, beer, lawnmowers) into the mailers so targeting looked less surgical. The model and its precision stayed; they just got better at hiding it. Analytical thinking was sophisticated; critical thinking around it was absent.
What critical thinking is (and is not)
- It is not criticising everything, nor being sceptical of every claim.
- It is a careful process of evaluating information before reaching a conclusion, asking:
- What exactly is the claim?
- What evidence supports it?
- Is there another explanation?
- What assumptions am I making?
- Could I be wrong?
- Managers, judges, consultants, scientists and journalists all ask these questions.
Key takeaways
- Real numbers and real charts can still support a wrong conclusion; data is always interpreted.
- Smart people are often better at defending beliefs, so critical thinking is discipline, not IQ.
- Target: accurate analytics, absent critical thinking ("Can we?" asked, "Should we?" not).
- Critical thinking = evaluating information before concluding, by asking the five questions above.
Logic and Critical Thinking in Reasoning
Reasoning
Scriven and Paul (two of the most-cited names in the field) describe critical thinking as "self-guided, self-disciplined thinking which attempts to reason at the highest level of quality in a fair-minded way."
Reasoning is the process of moving from what you already know or believe to a new conclusion: start with information, do something with it, end up somewhere new. Two people can start from identical information and reach opposite conclusions; the process, not the information, makes reasoning good or bad.
Example: signups doubled after a price cut.
- Founder 1: lower prices drove signups, so cut prices further.
- Founder 2: signups doubled, but was it the price cut or a marketing push in the same week?
- Same fact, different reasoning; only the second checks whether the conclusion is earned.
Logic vs critical thinking
| Logic | Critical thinking | |
|---|---|---|
| Checks | The structure of the move from premise to conclusion | The material going in: the premises |
| Asks | Does the conclusion follow from what you started with? | Are the starting points true, complete, relevant, and the right question? |
| Failure it catches | Conclusion does not follow, even if premises are true | Premises are incomplete, biased or false; the argument "stands on nothing" |
- 1Premises / evidenceCritical thinking: are they true, complete, relevant?
- 2Reasoning stepLogic: does the conclusion follow?
- 3Conclusion
- Reasoning runs constantly and fast, unnoticed: reading a metric, sizing up a competitor, deciding whether to trust a claim. A conclusion simply seems to "arrive" and feels normal.
- The skill is to notice that moment and run both checks before accepting the conclusion.
- Logic and critical thinking are two expressions of one discipline, not separate skills.
- Critical thinking is not a checklist or an inborn trait; it is a skill that can be learned.
When reasoning goes wrong: the Challenger disaster (28 Jan 1986)
- Conditions: unusually cold morning at Kennedy Space Center, colder than any previous shuttle launch; ice on the launch tower.
- Morton-Thiokol (maker of the solid rocket boosters) held an emergency call with NASA the evening before. Concern: the O-ring, a rubber seal about the thickness of a shoelace between booster segments that keeps hot gases from escaping the joint.
- Engineers had data: on a previous cool launch at 53°F, O-rings showed erosion and blackening (hot gas partially escaped). Forecast for launch: 29°F.
- Engineer Roger Boisjoly argued for delay: cold had been linked to O-ring damage; tomorrow would be colder than any previous case; so risk was higher than ever.
- NASA management pushed back, asking engineers to prove with data that cold caused failure. The engineers charted only flights where O-ring damage occurred, against temperature. This omitted every flight with no damage, including several warm launches. Without that comparison the pattern looked weak; with it, the pattern is obvious (colder launch, more likely damage).
- The launch proceeded; 73 seconds after liftoff an O-ring failed and Challenger broke apart, killing all 7 crew.
Lessons:
- Given the evidence on the chart, the conclusion followed reasonably. The logic was not the problem; the evidence was incomplete and nobody caught it in time.
- Logic says whether a conclusion flows from premises, nothing about whether the premises are the right or complete ones. A valid argument can yield a wrong answer if its starting material is flawed.
- "Management ignored the engineers" is true but not useful. The sharper question: what evidence and what kind of questions were left out of the thinking process?
Logic and critical thinking in practice
Logic = the structure of an argument, the formal rules determining whether a conclusion flows from its premises. Illustration: "All recorded O-ring failures happened below 40 degrees. Tomorrow will be 29 degrees. Therefore there is a meaningful risk of O-ring failure": valid structure, conclusion follows. Logic alone cannot say whether the premises are true, complete or the right ones; critical thinking examines where premises come from, what is missing, and whether you are asking the right question.
| Combination | Result |
|---|---|
| Logic without critical thinking | Argument holds together internally but rests on a shaky foundation: faulty assumptions, selective evidence, vested interests |
| Critical thinking without logic | A vague sense "something is off" that you cannot explain, so you cannot convince others and can be talked out of a correct instinct by someone who simply sounds more confident |
| Both | Good reasoning: construct an argument that holds together and keep testing whether it should; follow the answer where it leads, even to an undesired conclusion (Scriven and Paul's "fair-minded") |
Same failure elsewhere
| Case | What was defensible | What was missing |
|---|---|---|
| IL&FS (2018), a large Indian infrastructure financing group | Rated at the highest grade by 3 major rating agencies shortly before defaulting on roughly 91,000 crore of loans | Debt had grown 44% in the previous 3 years; numbers were in public filings (not hidden). No one with authority asked why a line of credit meant to cover overdue payments had not been used for that purpose |
| BYJU'S (2022-23) | Valued at $22 billion (India's most valuable startup then); growth was genuine (millions of students joined in the pandemic); aggressive acquisitions to compound it followed reasonably | Written down 86% to under $3 billion by late 2023. Missing: rigorous ongoing check on whether pandemic-era growth, and the losses incurred chasing it, could continue once conditions shifted |
| Challenger | Engineers' conclusion followed from the chart | Complete evidence (non-damage flights) |
Common thread (space agency, credit rating, startup): reasoning was defensible on its own terms; what was absent was someone stepping back to test whether the starting assumptions deserved the confidence placed in them.
Practical questions for the "decision room" (founder pressured to launch early, investor deck resting on one untested optimistic assumption, team debating shipping a feature):
- The data shows X, but which data? Was some left out?
- Not only "does this argument make sense?" but "what would I need to see or hear to know whether it is right?" (logic alone cannot answer this).
- "Data-driven" does not mean carefully reasoned; data can form a perfectly logical case while missing the one piece that would change the conclusion.
Capable, well-informed people still decide badly because the mind has default settings that work against it: confirmation bias, motivated reasoning, anchoring.
Key takeaways
- Logic = structure (does it follow?); critical thinking = premises (do they deserve trust?).
- Challenger, IL&FS and BYJU'S show sound-looking reasoning on unexamined premises.
- Logic without critical thinking is brittle; critical thinking without logic is unpersuasive.
- "Data-driven" is not a guarantee: ask which data, and what was left out.
Barriers to Sound Reasoning
Failures of the premise check come not from carelessness or missing tools but from default mental settings. Three main obstacles:
| Obstacle | Core idea | Working parts |
|---|---|---|
| Confirmation bias | Favouring information that supports what you already believe | Selective search, selective interpretation, selective memory |
| Motivated reasoning | Reasoning shaped by what you want to be true rather than what the evidence supports | Asymmetric scrutiny, rationalization |
| Anchoring | The first number or claim heard shapes all later judgments, even if baseless | The anchor, insufficient adjustment |
Barriers to Sound Reasoning
Confirmation bias
- Selective searchlooking only where you are likely to find what you already think.
- Selective interpretationreading ambiguous evidence in line with your belief.
- Selective memoryrecalling supporting instances more easily than others.
Example: a founder believes the product's advantage is speed. They read only reviews mentioning load times (search); file "It's fast but a bit expensive" as a compliment, not a pricing complaint (interpretation); a month later recall "fast" immediately but not "expensive" (memory).
Remedy: before reviewing evidence, write down what would change your mind, then check whether you looked for that, not just more support.
Motivated reasoning
- Asymmetric scrutinyevidence for the preferred conclusion is accepted quickly; evidence against is picked apart and held to a higher standard before it counts.
- Rationalizationafter reaching a conclusion this lopsided way, a story is built making the process look neutral, even to the person who did it.
Example: a team spent 8 months building a feature. Strong week-1 engagement is accepted as proof it works; the sharp week-3 drop is challenged ("tracking may be off", "seasonal dip", "needs another month of data"). Same dashboard, only one number cross-examined.
Remedy: ask who benefits if the conclusion is true (is it you?), and whether you would apply the same scrutiny to evidence that does not align with yours.
Anchoring
- The anchoran initial figure entering the conversation, often with little justification (e.g. a pen priced Rs 9.99).
- Insufficient adjustmentlater estimates move from the anchor, but not far enough; the pull persists even after better information arrives (the actual price is Rs 10, yet 9.99 still pulls).
Evidence: people spun a wheel landing on a random number, then estimated the percentage of UN member countries that are African; those who spun higher numbers gave higher estimates, though the wheel was irrelevant. Business example: an investor opens a negotiation at a specific valuation; every later discussion adjusts up or down from it instead of being rebuilt from the company's actual numbers, though the number may have been picked with little rigour.
Remedy: notice the first number that entered the conversation and ask whether it was ever justified or just the first thing anyone said.
How they combine
- They rarely act alone; they reinforce each other.
- This does not mean every conclusion is suspect. The conclusions deserving the closest scrutiny are those that come very fast, feel obvious and look convenient.
- Usually one of the three is doing quiet work underneath.
Critical thinking in professional life
It is risk management for decisions involving other people's money, time and careers.
- AI hallucination: large language models embedded in research, drafting and financial-analysis tools produce confident, fluent but fabricated output. In 2023 a New York lawyer filed court citations to cases that did not exist, having used an AI tool for precedents without checking; the result was sanctions and loss of licence. Convincing and true are not the same. A confident tone and a correct claim often travel together, which is why people stop checking, but a confident tone may not support a correct claim.
- Hiring: panels form a strong impression in the first 5 minutes (confident delivery, shared background), then unconsciously seek confirming evidence and explain away the rest: anchoring plus confirmation bias. The cost is a bad hire with lasting effects.
- Startup metrics: improving metrics get prominence, deteriorating ones a brief mention or none. Not outright fraud, but motivated reasoning at organisational scale. Investors/board members asking "What number are we not looking at right now?" are running the critical-thinking check, which will not run on its own under incentives.
Critical thinking in civic life
- Anchoring at civic scale: a context-free video clip or made-up/out-of-context/AI-generated image can reach millions before verification, during disasters, breaking news and elections. Whichever version arrives first sets the frame; corrections arrive later and rarely spread as far or fast.
- COVID-19: claims about treatments, vaccines and risk spread at extraordinary speed; willingness to believe tracked whether they matched prior beliefs about institutions, governments and vaccines. People on opposing sides looked at overlapping studies and coverage but noticed, remembered and trusted different subsets: confirmation bias with life-and-death stakes.
- Elections: voters get statistics, projections and claims from sources with competing incentives. Same questions as a boardroom: where did the number come from, what might be missing, am I accepting it because it is well supported or because it is what I wanted to hear? Same tools; only the setting changes.
Why it matters more now: the volume of claims encountered daily (news, social media, messaging apps, AI content) is vastly larger than a decade ago, and generating a false but convincing claim is far cheaper (a fake image or video once took skill and time; now an AI can make one in a second). Checking habits have not kept pace. Hence critical thinking is a civic responsibility: a citizenry that anchors on the first claim and reads evidence only to confirm itself can be moved by whoever is fastest and most emotionally compelling, regardless of truth.
Consolidation
- Reasoning moves from what you know to a new conclusion. Logic tests structure; critical thinking tests whether premises deserve trust. Each catches failures the other cannot see.
- Confirmation bias shapes what you search for and notice; motivated reasoning shapes which conclusions you want and how sceptically you examine evidence; anchoring lets the first number or claim frame everything.
- Together they yield conclusions that feel fluent and confident but rest on a non-neutral process. The same discipline protects boardroom decisions and democratic judgment; it is "civic infrastructure."
Key takeaways
- Three obstacles: confirmation bias, motivated reasoning, anchoring; each has two parts and a specific remedy.
- They reinforce one another; be most wary of conclusions that are fast, obvious and convenient.
- Professional cost: fabricated citations, biased hiring, selective metrics. Civic cost: viral misinformation, polarised health beliefs, election claims.
- The scale of unchecked claims makes critical thinking a civic duty.
Evidence, Interpretation and Framing
Premises arrive as evidence (a number, report, claim or chart). Evidence is interpreted, and is already framed before it reaches you, shaping the conclusion it seems to support. Obstacles like confirmation bias operate on evidence. Framing example: "headcount reduction" vs "gutted workforce" (cognitive vs emotive meaning, after Patrick Hurley).
Professionally you will read (and produce) investor memos, market research and competitor analysis; civically you will meet evidence selected and framed by those wanting you to reach a particular conclusion. The response is not cynicism but asking: what if this were presented differently, and would that change what I believe?
Evidence, Interpretation and Framing
Evidence versus assertion
| Assertion | Evidence | |
|---|---|---|
| Definition | A statement offered with no support attached | Something offered in support of a claim |
| Basis for acceptance | Speaker's authority, confidence or repetition | Can be checked, traced and verified by someone other than the claimer |
| Example | "Our product is the fastest on the market." | "Independent benchmarking by a third-party testing firm found our product loads pages 40% faster than our 3 closest competitors." |
- The difference is not tone (both sound confident) but what stands behind the sentence. The evidence example gives a named source (third-party firm), a method (independent benchmarking) and a specific comparison (40% faster than 3 closest competitors).
- Confident unsupported claims can be as persuasive as, or more than, real evidence, since real studies come with caveats and margins of error.
- Test: what would I have to do to check this myself? A specific source, method or data set to consult means evidence; none means assertion.
- Numbers are not evidence by default: they need a traceable source and method. "9 out of 10 professionals recommend this approach" is an assertion unless you know who the professionals were, how many were surveyed and who ran the survey. Specificity does the persuasive work that verification should do.
- Professional example: a market-sizing slide saying "$50 billion opportunity" is an assertion unless it cites its source and what it measures.
- Civic example: "many people say" with no attribution, circulating alongside claims citing named sources, looks identical on social media; delivery confidence says nothing.
- A citation is not enough: a cited claim can still fail if the citation does not say what is claimed, the chart's axes mislead, or a screenshot is altered or out of context. The step people skip: has someone equally credible confirmed the source says what it is being used to say?
- Not every assertion should be distrusted: a doctor's decades of experience can yield a well-informed clinical judgment without a CT scan, MRI or blood tests.
Facts, inferences and interpretations
Most claims blend three components that people wrongly treat as equal in weight.
| Component | Definition | Answers | Examples |
|---|---|---|---|
| Fact | A verifiable state of affairs, true regardless of who looks; checkable by anyone | What happened? | Quarterly revenue was $2.3 million; unemployment was 4.1% last month; the video was recorded on March 3rd |
| Inference | A conclusion drawn from facts by reasoning, a step beyond what was observed; follows a visible path that can be assessed for validity or strength | What does it likely lead to? | Revenue rose in each of the last 3 quarters, so it will likely rise again next quarter |
| Interpretation | Meaning or significance assigned to facts, shaped by the person's perspective, values or framework; two people can reach different, equally reasonable interpretations | What does it mean? | "Revenue growth shows our new sales strategy works" vs "it reflects a market-wide seasonal upswing that would have happened anyway" |
- Choosing between competing interpretations needs additional evidence, judgment and context, not repetition of the fact.
- Civic example (crime statistics): crime incidence rose 15% in a city last year.
- Fact: the reported 15% rise (if the reporting methodology is sound).
- Inference: it reflects a real increase in underlying crime, rather than changed reporting behaviour or policing focus; reasonable only if those alternatives are ruled out.
- Interpretation: a particular policy failed or the public-safety approach should be abandoned; shaped by the interpreter's values.
- The three are often delivered in the same tone of certainty, so listeners treat value-laden interpretations as if they had the footing of facts. "Crime rose" does not by itself tell you what to do.
- Interpretation is necessary (facts rarely say what to do; refusing to interpret just lets someone else's interpretation fill the space). The aim is to know which one you are doing and stop fact-certainty leaking into inference and interpretation.
- Habit: split a claim into layers. Fact: checkable, in dispute? Inference: evaluate like any argument. Interpretation: a judgment about significance, open to reasonable disagreement, not settled truth.
Key takeaways
- Evidence is checkable (source, method, specific comparison); assertion rests on confidence or authority.
- Numbers and citations need verification that they say what is claimed.
- Fact = what happened; inference = what it likely leads to; interpretation = what it means (value-shaped).
- Separating the three is the base for studying framing, omission and the "objective record."
Framing and the Limits of Evidence
How framing influences interpretation
Framing is how presentation shapes reception. It does not change what is true but changes what the audience takes away as true.
Demonstration (research): doctors and patients received identical data on a medical procedure, described as "90% survival rate" or "10% mortality rate". The survival framing group was consistently more willing to choose the procedure though risk was identical.
| Mechanism | How it works | Example |
|---|---|---|
| Word choice | Cognitive vs emotive meaning (Hurley); same fact, different emotional charge, so the conclusion is reached before facts are evaluated | "reduced headcount by 15%" vs "gutted its workforce" |
| Comparison point | What the number is measured against | "Sales grew 5%" (modest) vs "...best result in company history" (triumph) vs "...well below the industry's 12% average" (problem) |
| Visual framing | Charts look objective but scale persuades | Bar chart with y-axis starting at 90 not 0 makes a 2-point shift dramatic; 3-month line shows a spike, the same data over 3 years a flat trend |
| Order and emphasis | What gets the headline vs a footnote | Report leads "revenue grew 8%", mentions in paragraph 6 that costs grew 14%; technically complete, practically hidden |
| Anchoring | The nearby presence of a number biases judgment | See below |
Anchoring (Kahneman)
Two distinct operations:
- Deliberate but insufficient adjustment: you consciously move from the starting number toward a better estimate but stop short, so the answer stays biased toward the start.
- Suggestion: a number shifts judgment even when never treated as a serious estimate and even when known to be arbitrary.
Studies described:
- Real estate agents shown the same house with different listing prices valued it biased toward the listing price they saw, while insisting it played no role.
- Quick estimate of gave a noticeably lower guess than the reversed ; the first numbers seen set the anchor.
Takeaway: the anchor is often just the first number a presenter chooses; later numbers are judged relative to it rather than fresh.
Framing in practice
- Professional: a pitch deck can describe the same unit economics as "capital efficient compared to our category" or "margins below what most investors expect at this stage"; same numbers, opposite framing, driven by comparison set and headline paragraph.
- Civic: "Unemployment falls to lowest level in a decade" vs "Millions still without work despite reported job gains": both accurate, neither fabricated, each steering the reader before the article begins. Tax policy as "tax relief for working families" vs "a tax cut that mostly benefits high earners": both defensible readings of the same bill.
- There is no neutral, frame-free way to present information (choosing words, order and comparison is unavoidable). Not every frame is manipulative. The goal: notice the frame and ask what a different, equally accurate frame would look like before letting it set your conclusion.
Habits
- Substitution testreframe the statistic (failure rate instead of success rate; smaller time window; harsher comparison). Would you draw the same conclusion?
- Anchoring checkidentify the first number; has your estimate moved far from it? If not, the conclusion came from the frame, not the fact.
Selection bias and the objective record
Selection bias is the sharpest form of framing: not how evidence is presented, but what never reaches you. You can notice a loaded word or a misleading axis, but not something never shown.
WYSIATI
Kahneman's WYSIATI: "What You See Is All There Is." The mind builds a coherent story from whatever information is available without asking whether important information is missing. It is near-automatic and affects everyone. A coherent conclusion built on a small, selective set of facts feels as solid as one built on the full picture. Hence "what would change my mind?" must be preceded by "What wasn't I shown that might have changed it?"
Survivorship bias: Wald's bombers (WWII)
- Military researchers mapped damage on returning bombers to decide where to add armour; damage clustered on wings and tail, so the natural conclusion was to armour those.
- Statistician Abraham Wald saw the flaw: the data included only planes that made it back, showing where a plane can be hit and survive. Planes hit in undamaged-looking areas (engines, cockpit) never returned.
- Armour belonged where the data showed no damage; the absence of data was the key signal, invisible until someone asked what was missing.
- Business parallel: an accelerator showcasing successful alumni is accurate but by construction excludes every failed company from the same programme, so a prospective founder's sample misrepresents real odds.
Availability and risk perception (civic)
- Per Kahneman, people overestimate dramatic, vivid causes of death (plane crashes, shark attacks, terrorist incidents) relative to undramatic but dangerous ones (heart disease, diabetes).
- Not from misjudging accurate statistics: vivid causes get disproportionate media coverage, so they are more available in memory, and availability is mistaken for frequency.
- Consequence: public concern and even policy are shaped by what is dramatic and reportable rather than common and serious. Nobody needs to suppress heart-disease data; one category being more visible is enough, and visibility is mistaken for the whole picture.
The objective record
The belief that raw evidence (video, dataset, transcript) is neutral and free of framing and selection is wrong. Even raw records reflect choices made at collection, before any framing or selective presentation.
- Video ("the tape doesn't lie"): a camera has a field of view; everything outside the frame is absent from the record however relevant. The recording is real but not automatically complete.
- Datasets feel even more objective, but exist only because someone decided what counted as a data point worth recording; that decision is a choice, not a neutral fact of the world.
- Linked to Kahneman's coherent-narrative tendency: a clean, legible dataset or clearly shot video feels authoritative, but clarity does not guarantee nothing relevant lies outside the frame.
- A record can be entirely genuine yet partial. Recognising this separates careful readers of evidence from those who trust whatever looks like proof.
- 1Collection: what is captured / counted
- 2Selection: what reaches you
- 3Framing: words, comparison, scale, order, anchor
- 4Your interpretation and conclusion
Key takeaways
- Identical data framed as 90% survival vs 10% mortality changes decisions; five mechanisms: word choice, comparison point, visual scale, order/emphasis, anchoring.
- No frame-free presentation exists; notice the frame and test alternatives (substitution test).
- WYSIATI: missing information is invisible; ask what you were not shown. Wald's bombers and accelerator alumni illustrate survivorship bias; vivid risks illustrate availability.
- Even "raw" records (video, data) embed collection choices and are partial; genuine does not mean complete.
Revise before your examAll 4 exam tips from this module in one list
- Reasoning fails in exactly two places, structurally (logic) or at the source (premises). Name which one a case illustrates.
- Match each remedy to its bias: write down what would change your mind (confirmation); ask who benefits and apply equal scrutiny (motivated); identify the first number and whether it was justified (anchoring).
- First question for any claim: "Is this evidence, or an assertion wearing evidence's clothing?"
- Substitution test is for framing ("is this framed favourably?"); the selection-bias question is distinct: "What population, dataset or cases were excluded before I saw the data, and would including them change my conclusion?"
Test yourself on this module
Questions from these notes, one at a time, with feedback. Your weak topics are saved.