Level 5 — Advanced · Technical & Conceptual Reading
Lesson 76: Advanced Reading: Bias, Framing, Selection & Missing Evidence
The fifth part of this course's argument-analysis arc — reading a text that is entirely factual and still slanted, through what it selects, what it omits, how it frames the same numbers positively or negatively, and whose perspective and interests shaped it.
Learning Objectives
By the end of this lesson, you should be able to:
- Distinguish bias from falsehood, recognising that a text can be entirely factual while still selecting, omitting, and framing information to create a slanted overall picture
- Identify loaded language, euphemism, dysphemism, and labelling as tools that carry evaluation inside seemingly descriptive words, and distinguish a material omission (one that would change interpretation) from an unavoidable one
- Read percentage and framing claims for base rate neglect, denominator tracking, and relative-versus-absolute framing, connecting quantitative claims back to their actual scale
- Ask the counterfactual question — 'what would we expect to see if the author's explanation were wrong?' — and consider perspective bias and conflict of interest as sources shaping a text's overall picture
- Apply the 'missing column' technique to separate what an argument shows from what it doesn't, and judge whether that missing information would materially change the conclusion
Introduction
Lesson 75 asked how much to trust a piece of evidence based on its source and type. Today's lesson — the fifth part of this course's argument-analysis arc — asks a different, equally important question: what evidence is an author showing you, and what are they not showing you? A text can be entirely factual and still present a slanted picture, through what it selects, what it leaves out, and how it frames the numbers it does include.
Bias Does Not Mean Falsehood
The single most important distinction this lesson makes
An author's stated information can be completely true, while still producing a biased overall picture — by giving some evidence more weight, leaving other evidence out, choosing particular words, foregrounding one perspective, and backgrounding another. A text can contain true information and still present a biased picture. Bias ≠ falsehood — a distinction worth holding onto through everything that follows.
Selection: What Information Was Chosen?
True statements can still be an incomplete picture
The new automated billing system reduced processing time by 30%. This can be entirely true. But if the report never mentions that maintenance costs increased, training was expensive, smaller terminals saw no improvement, or error rates stayed unchanged, the reader's overall impression can end up considerably more positive than the full picture warrants. Ask, always: what information was selected?
Omission: What Wasn't Said?
Missing information can change an entire interpretation
The company introduced automation and productivity increased by 20%. If the text never mentions that the company also hired 30 additional employees during the same period, the interpretation is incomplete — additional staffing is a genuine alternative explanation for the productivity increase, and it has simply been left out.
Not Every Omission Is Bias
No text can include everything — the real test is materiality
A 500-page book cannot include every possible piece of relevant evidence. Omission ≠ automatically bias. The real question: was the omitted information important enough to change the reader's interpretation? If yes, the omission is material. If not, it's simply the ordinary, unavoidable compression every piece of writing requires.
What Is Framing?
The same underlying facts, presented to steer attention or interpretation
Frame A: the new system reduced processing time by 30%. Frame B: the new system failed to reduce overall operating costs despite a 30% reduction in processing time. Both frames can share the identical underlying fact — a 30% processing-time reduction — while directing the reader's attention in opposite directions.
Positive Versus Negative Framing
Identical numbers, opposite psychological effect
95% of invoices were processed successfully. Positive framing: a 95% success rate. Negative framing: a 5% failure rate. Numerically identical information — but the two framings land very differently. The reading skill: separate the underlying fact from the framing applied to it.
Loaded Language
Some words carry evaluation, not just description
Compare: the company changed its procedures / the company overhauled its procedures / the company disrupted its procedures. Same underlying event, three different tones. More pairs worth recognising on sight: reform vs. interference, investment vs. spending, streamlined vs. cut, cautious vs. hesitant, ambitious vs. unrealistic, efficient vs. aggressive, traditional vs. outdated, innovative vs. experimental. A word's connotation often reveals an author's attitude even when the underlying fact stays constant.
Euphemism and Dysphemism
Softening a negative, or sharpening a neutral, through word choice
Euphemism — describing something negative in softer language: workforce reduction instead of layoffs. Dysphemism — describing something neutral or ordinary in deliberately negative language: bureaucratic interference instead of regulatory oversight. Both are framing tools, working in opposite directions from the same underlying event.
Labeling
The name given to a person or group shapes how they're perceived
Compare: government regulators vs. government bureaucrats; industry experts vs. industry lobbyists. Each label carries its own attitude toward the same role. Ask, whenever a group is named: what label is being used, and what attitude does it carry?
Evidence Selection Bias
True evidence, incompletely represented
Four studies found that automation improves productivity. This can be technically true. But if six other studies found no meaningful improvement, citing only the four positive ones leaves an incomplete picture. Ask: how were these particular studies selected?
Cherry Picking
Choosing the favorable evidence, and the same caution as with omission
Cherry picking means selecting only the evidence that supports a preferred conclusion while ignoring inconvenient evidence: the company highlights three successful automation projects while ignoring seven projects that failed to meet their targets. But — exactly as with omission — mentioning only three studies does not, by itself, prove cherry picking. The question remains: why were those three selected?
Base Rate Neglect
A percentage can sound excellent while hiding a very different underlying reality
A new security system detected 99% of suspicious transactions. Sounds excellent — until you factor in how rare suspicious transactions actually are. If only 100 of 100,000 transactions are genuinely suspicious, the number of false positives could still be substantial. Whenever a percentage appears, ask: what is the base rate it's actually operating against?
Denominator Matters
Connecting back to Lesson 56's percentage-tracking work
Complaints increased by 50%. Sounds like a large increase. It could mean 2 complaints became 3. It could mean 10,000 became 15,000 — the same 50% increase, on a completely different absolute scale. Always identify the denominator and baseline before reacting to a percentage change.
Relative Versus Absolute Framing
Connecting back to Lesson 56 — the same distinction, applied to bias detection
Error rates fell by 50% — if the change was 2% → 1%, the absolute reduction is one percentage point, while the relative reduction is 50%. Conflating the two can make a genuinely modest change sound dramatic, or the reverse — always check which framing a stated percentage is actually using.
Positive Evidence Versus Missing Negative Evidence
One good metric does not make the whole picture positive
Customer satisfaction increased by 15%. How many customers? Which customers? What measurement method? What was the previous baseline? What about dissatisfied customers, complaints, or churn? A single positive metric reported on its own says nothing about whatever negative metrics weren't mentioned.
Survivorship Bias, Revisited
Connecting back to Lesson 75
The world's most successful startups used strategy X. Therefore, strategy X causes startup success. Where's the data on unsuccessful startups that used the exact same strategy? Ask, always: who is missing from this dataset?
Counterfactual Missing Evidence
A genuinely powerful question for testing any causal claim
Ask: what would we expect to see if the author's explanation were wrong? Author's claim: automation caused productivity growth. Counterfactual question: if automation were not the cause, what else might we observe? Possible answers: productivity also rose in non-automated teams; staffing changed; workload changed; management policy changed. Each of these, if actually observed, would constitute a genuine alternative explanation — and actively imagining them is how a critical reader tests a causal claim rather than simply accepting it.
Perspective Bias
The same event looks different depending on whose eyes it's seen through
Take automation deployment: management sees cost reduction; employees see workload and job security; the IT team sees maintenance complexity; customers see speed and service quality; finance sees total cost of ownership. A report drawing only on management's perspective isn't necessarily false — it's perspective-limited, presenting one genuine stakeholder's view as though it were the whole picture.
Whose Interests Matter?
Incentive analysis, not conspiracy thinking
Ask: who benefits if the reader accepts this conclusion? A software vendor claiming "our platform dramatically improves operational efficiency" naturally invites the follow-up questions: is there independent evaluation? Is this only the vendor's own data? Is there a comparison group? Are negative findings reported? Is cost data included? A source's incentives can shape how evidence gets interpreted and presented, without that source ever stating a single false fact.
Conflict of Interest
When a source's conclusion and its financial interest point the same direction
A company-funded study finds that its own product is highly effective. This is not automatically invalid — but it deserves additional scrutiny: was the methodology transparent? Were independent researchers involved? Were negative results reported alongside positive ones?
Author Bias Versus Text Bias
A careful, defensible distinction — don't overclaim intent
Author bias is a person's actual preference or viewpoint — usually unknowable with certainty from the text alone. Text bias is the effect actually produced by the final text's selection, framing, omission, language, and evidence balance — visible directly on the page. Prefer "the text appears to frame the issue positively" over "the author deliberately manipulated the reader" — the second is a far stronger claim about intent than a critical reader can usually support from the text alone.
Tone and Evidence Together
Check whether the language's strength actually matches the evidence's strength
Although the pilot produced modest improvements, management's ambitious automation strategy represents a major step toward modernization. Evidence: modest improvements. Evaluative phrase: major step. Whenever evaluative language sounds considerably stronger than the evidence actually reported, that mismatch itself is worth flagging explicitly — don't just notice the tone or the evidence in isolation; check them against each other.
A Strong Critical Reading Example
Positive evidence, positive framing, and a list of what's missing, weighed together
The automated billing system increased invoice-processing speed by 30% during its first year. Management therefore considers the project a major success. Employee surveys also showed high satisfaction with the new workflow. However, the report does not provide data on implementation costs, long-term maintenance, or performance in smaller terminals.
Positive evidence: processing speed up 30%; employee satisfaction high. Positive framing: "major success." Missing evidence: implementation cost, maintenance cost, smaller-terminal performance. Evaluation problem: success according to which criteria? Final judgment: the evidence supports operational improvements, but it is insufficient to establish overall project success.
A Bias Detection Framework
Ten questions for any serious text
The "Missing Column" Technique
A concrete tool for turning a vague suspicion of bias into a specific list
Build a two-column mental table: what the author shows, and what the author does not show.
| Author shows | Author does not show |
|---|---|
| Processing time ↓ | Long-term cost |
| Error reports ↓ | Actual error rate |
| Satisfaction ↑ | Non-users' opinions |
| Four terminals improved | Smaller terminals |
| Six-month results | Five-year reliability |
Then ask: does the missing column materially change the conclusion? If it does, the conclusion should be qualified accordingly.
Bias Spectrum
Avoid binary thinking — most texts sit somewhere in between
Picture a spectrum: balanced → selective → strongly selective → misleading framing. No serious text is purely 100% objective or 100% biased; the useful question isn't "is this text biased, yes or no?" but "where does it sit on this spectrum, and in which direction?"
Lessons 72–76, Combined: A Master Framework
The full argument-and-bias reading architecture built across this five-lesson arc
Vocabulary in Context
framingnounformal
presenting the same underlying facts in a way that steers a reader's attention or interpretation in a particular direction (উপস্থাপনার ছাঁচ)
Often with media framing, framing of, deliberate framing
“The framing of the story as a 'crackdown' rather than 'enforcement' shaped how readers reacted.”
loaded languagenoun phraseneutral
words that carry built-in evaluation or attitude rather than neutral description, even when describing the same event as a plainer word would (আবেগপূর্ণ ভাষা)
Often with use loaded language, loaded language such as, avoid loaded language
“The press release used loaded language, calling rival firms 'reckless' instead of naming any fault.”
cherry pickingnoun phraseneutral
selecting only the evidence that supports a preferred conclusion while leaving out evidence that doesn't (পছন্দসই তথ্য বাছাই)
Often with accused of cherry picking, cherry picking of data, avoid cherry picking
“The vendor's report showed only its best quarter, a clear case of cherry picking.”
base rate neglectnoun phrasetechnical
ignoring how common or rare an underlying event actually is when interpreting a percentage or detection rate about it (মূল হার উপেক্ষা)
Often with commit base rate neglect, avoid base rate neglect, example of base rate neglect
“Reporting that 'ninety percent of the delayed ships were foreign' shows base rate neglect if most ships are foreign.”
counterfactual questionnoun phrasetechnical
asking what evidence would be expected to appear if a claimed explanation were actually false, as a test of that explanation (বৈপরীত্যমূলক প্রশ্ন)
Often with ask a counterfactual question, useful counterfactual question, counterfactual question about
“A useful counterfactual question is what the data would show if the new system had no effect.”
conflict of interestnoun phraseformal
a situation in which a source's financial or personal interest points in the same direction as its stated conclusion, warranting extra scrutiny (স্বার্থের দ্বন্দ্ব)
Often with declare a conflict of interest, potential conflict of interest, avoid a conflict of interest
“The consultant declared a conflict of interest because her firm had bid for the contract.”
perspective biasnoun phrasecourse term
a text's tendency to present one stakeholder's view of an event as though it were the complete picture, while other stakeholders' views go unmentioned (দৃষ্টিভঙ্গিগত পক্ষপাত)
Often with perspective bias, show perspective bias, risk of perspective bias
“The article shows perspective bias: it quotes only management and never the drivers.”
material omissionnoun phraseformal
missing information important enough that including it would change how the reader interprets the text (তাৎপর্যপূর্ণ বাদ পড়া তথ্য)
Often with a material omission, disclose a material omission, material omission from
“The prospectus contained a material omission: it did not mention the pending lawsuit.”
Guided Reading Practice
Read this passage once, straight through. Then apply the missing- column technique, name any loaded language or framing choice, and ask the counterfactual question, before checking your reading against the notes that follow.
Since adopting the new scheduling software, the clinic has seen appointment no-shows drop by an impressive 40%. Staff describe the transition as a resounding success, crediting the software's automated reminder feature. The clinic has not published data on total patient volume during this period, nor on whether reminder calls placed by staff — a practice that also began around the same time — might have contributed to the improvement.
Loaded language and framing: "impressive," "resounding success" — strongly positive evaluative language attached to a single metric. Missing column — shown: no-show rate down 40%, staff endorsement. Not shown: total patient volume (a lower volume alone could reduce no-shows without the software helping at all), and the concurrent introduction of staff reminder calls — a genuine alternative explanation the passage names but doesn't rule out. Counterfactual question: if the software weren't the actual cause, what would we expect to see? Possibly the same no-show reduction, explained instead by the reminder calls, or by a coincidental drop in patient volume. The passage's own framing ("resounding success," crediting the software specifically) outruns what its own reported evidence actually establishes.
Golden Rule
Golden Rule
Do not ask only "is this information true?" — ask what picture the selection, framing, language, and omission of information creates, and whether that picture is actually justified by the full evidence, always asking what is missing that could change your interpretation.
Lesson Summary
This lesson added a layer the previous four lessons in this arc hadn't directly addressed: what an author chooses to show, and what they leave out. You learned that bias does not mean falsehood — a text can be entirely factual while still producing a slanted picture through selection, omission, framing, and word choice. You worked through loaded language, euphemism, dysphemism, and labelling as tools that smuggle evaluation into apparently neutral description, and learned to judge an omission by materiality rather than assuming every gap is bias. You revisited percentage framing through base rate neglect, denominator tracking, and relative-versus-absolute framing, and practised the counterfactual question as a genuine test of a causal claim. You considered perspective bias and conflict of interest as forces shaping a text's overall shape, carefully distinguished author bias from text bias, and closed with the "missing column" technique — a concrete way to ask, of any argument, what is missing, and whether that absence would actually change the conclusion.
Five lessons, one connected skill
Lessons 72 through 76 have built one continuous reading capability: taking apart an argument's anatomy, naming its missing bridges, judging its evidence sufficiency, tracing that evidence to its actual source and quality, and now, checking what the whole picture leaves out. Held together, these five lessons replace "the author makes a good case" or "the author seems biased" with something far more useful — a precise account of exactly what was shown, what wasn't, and whether the conclusion has actually earned the confidence it's stated with.
Practice: Test What You've Learned
Work through every question yourself before checking anything.
Before you start
For every passage below, apply the missing-column technique explicitly — name at least one thing shown and one thing not shown — before judging whether the conclusion is justified.
Part A — Selection, Omission, and Materiality
- A company report states: 'Revenue grew 12% this year,' without mentioning that a major competitor exited the market during the same period. Explain why this omission could be material.
Show answer
A competitor's exit is a genuine alternative explanation for the revenue growth — customers may simply have shifted to this company by default, rather than the company's own strategy or product quality driving the gain. Leaving this out lets the reader credit the 12% entirely to the company's own performance, which could change the interpretation considerably. - A 400-page history book does not mention every minor treaty signed during the period it covers. Explain why this omission is not automatically evidence of bias.
Show answer
No text can include every possible fact, and minor treaties are unlikely to change a reader's overall interpretation of the period's major events. Since the omission isn't material — it wouldn't shift the reader's understanding if included — it's simply ordinary compression, not evidence of a slanted picture. - A product review lists five positive features and does not mention price. Explain what additional information would help you judge whether this omission is material.
Show answer
Knowing whether the product is priced dramatically higher or lower than comparable alternatives would help — if the price is unusually high, omitting it would materially change a reader's sense of whether the product is actually worth recommending, whereas an average price would make the omission far less consequential.
Part B — Framing and Loaded Language
- Compare: 'The policy cut funding by 10%' versus 'The policy streamlined funding by 10%.' Identify the loaded word and the attitude it carries.
Show answer
"Streamlined" is the loaded word, replacing the more neutral or negative "cut." It carries a positive, approving attitude, framing the same 10% funding reduction as an efficient improvement rather than a loss, even though the underlying fact is identical. - A report states: '80% of employees did not oppose the merger.' Reframe this statement in a way that presents the identical fact more negatively.
Show answer
"20% of employees actively opposed the merger" presents the identical underlying fact but frames it negatively, drawing attention to a meaningful minority of resistance rather than the majority who simply did not object. - Identify whether 'whistleblower' and 'informant' are likely to be euphemisms, dysphemisms, or neither, depending on context, and explain your reasoning.
Show answer
"Whistleblower" tends to function as a positive label, casting the person as a principled truth-teller, which can act as a euphemism when describing someone who broke confidentiality rules. "Informant" tends to carry a more negative, suspicious connotation, closer to a dysphemism, even when describing the identical act of reporting wrongdoing to authorities. Which label applies often reveals the author's own attitude toward the person described.
Part C — Base Rates, Denominators, and Relative Framing
- A fraud-detection tool flags 95% of fraudulent transactions. Explain what additional information about the base rate you would need before judging how many false alarms this might generate.
Show answer
You would need to know how rare fraudulent transactions actually are relative to the total number of transactions processed. If fraud is very rare, even a small false-positive rate on the much larger pool of legitimate transactions could generate a substantial number of false alarms despite the tool's impressive 95% detection figure. - 'Customer complaints rose by 200%.' Explain why this statement alone does not tell you whether this is a serious problem.
Show answer
Without the denominator and baseline, a 200% increase could mean 1 complaint became 3, which is trivial, or 5,000 became 15,000, which is serious. The relative percentage alone hides the absolute scale, so the statement can't be judged as a serious problem until the actual starting and ending numbers are known. - 'Deaths from the disease fell by 50%,' where the rate went from 4 in 100,000 to 2 in 100,000. Identify the absolute change and the relative change separately.
Show answer
The absolute change is a reduction of 2 deaths per 100,000 people (from 4 to 2). The relative change is 50%, since 2 is half of the original 4. Both are accurate, but the relative framing sounds far more dramatic than the modest absolute change actually is.
Part D — Perspective, Incentives, and the Missing Column
- A press release from a software company states its own product 'sets the new industry standard.' List two questions a critical reader should ask before accepting this claim.
Show answer
Two useful questions: is there any independent evaluation or third-party benchmark supporting this claim, rather than only the company's own data? And what criteria is "industry standard" actually being judged against — performance, adoption, price, or something else entirely? - A report on a new workplace policy quotes only senior managers about its effects. Identify the perspective bias, and name one other stakeholder group worth hearing from.
Show answer
The perspective bias is management-only framing — the report presents senior managers' view of the policy's effects as though it were the complete picture. Frontline employees, who directly experience the policy's day-to-day impact, are a stakeholder group worth hearing from as well. - Explain, using this lesson's terms, why 'the text appears to frame the issue positively' is a more defensible statement than 'the author is lying to readers.'
Show answer
"The text appears to frame the issue positively" describes text bias — an effect visible directly in the words on the page, which a reader can point to and support. "The author is lying" is a claim about author bias and deliberate intent, something usually unknowable with certainty from the text alone, so it is a far stronger claim than a critical reader can typically defend.
Part E — Full Passage Analysis
Return to this lesson's guided reading passage about the clinic's scheduling software.
- List what the passage shows and what it does not show, in two columns.
Show answer
Shows: a 40% drop in appointment no-shows, and staff describing the transition as a resounding success. Does not show: total patient volume during the period, and whether the concurrently introduced staff reminder calls contributed to the improvement instead of, or alongside, the software. - Identify the alternative explanation the passage names but does not rule out.
Show answer
The passage names staff reminder calls, a practice that began around the same time as the software, as a possible alternative cause of the no-show reduction — but it never rules this out or separates its effect from the software's. - State the counterfactual question this passage invites, and what evidence would help answer it.
Show answer
The counterfactual question: if the software weren't the actual cause of the drop, what would we expect to see? Evidence that would help answer it includes data isolating the no-show rate before and after the reminder calls started separately from the software rollout, and patient volume data over the same period. - Explain why "resounding success" is stronger language than the passage's own reported evidence actually supports.
Show answer
The only reported evidence is a 40% drop in one metric, no-shows, with at least one unexamined alternative explanation still on the table. "Resounding success" asserts a sweeping, confident judgment about the software's overall impact that the single, potentially confounded metric has not yet earned.
Next: Lesson 77 — Advanced Reading: Fact, Interpretation, Opinion, Evaluation & Inference: the sixth part of this course's argument- analysis arc, separating a text's six reasoning layers from raw observation to full recommendation.