about SCIE1001
0 · Exam map course data + lecture 13
Final Exam 35% · Course Quiz 2% · In-class Test 20% · Scientific Report 15% · Weekly Short Response 10% · Participation 18%
“Secure settings - no phones, smart watches, notes permitted” (SCIE1001 2026 Lecture 13 'Exam Details & Conditions' slide (PDF_Lecture_13_2026_pdf.txt))
| Part | What happens |
|---|---|
| 1 · 25% | 6 multiple-choice; one from each week 7–12 |
| 2 · 25% | choose week 7 or week 10 short answer |
| 3 · 25% | choose week 8 or week 9 short answer |
| 4 · 25% | choose week 11 or week 12 short answer |
Operating facts: 1.5 hours, on campus, final covers weeks 7–12 only; short answers are roughly half a page. The final is compulsory: you must sit it to pass. No minimum exam mark is stated.
1 · Recitable core C2–C7
Falsification exposes a claim to observations that could show it false. Inductive risk is possible harm from accepting or rejecting a claim under uncertainty. Epistemic values concern knowledge quality; non-epistemic values concern ethical, social or political priorities shaping questions, methods and uses.
Check → A scientific answer says what observation would count against it; a policy answer also says whose harms and benefits set the decision threshold.
1b · Four jobs, in order C1 · C12
| Job | Recite this check |
|---|---|
| orient | What system, population, scale or condition is being analysed? |
| connect | What verb-led mechanism changes the starting condition? |
| discriminate | What pattern favours this explanation over a plausible alternative? |
| bound | What uncertainty or missing link stops the conclusion? |
A complete response makes the task visible before detail: define the relevant terms, connect evidence to mechanism, use a comparison to test it, and answer the question actually asked. Do not report a label or group difference as a mechanism.
2 · Claim → test → revise C2 · C3 · C4
testable-claim frameDefine variable + population + comparison → predict an observable difference → name a possible disconfirming result → revise if contrary evidence arrives.
Demarcation problem: distinguish scientific inquiry and claims from non-science or pseudoscience without assuming one simple rule fits every discipline. One demarcation rule cannot settle every disputed field.
| Audit point | Question to ask |
|---|---|
| Validity | Does the variable measure the intended thing? |
| Selection | Who or what entered the data, and what was omitted? |
| Visual | Are scale, denominator and comparison preserved? |
| Inference | Association, tested mechanism, or causal claim? |
p-value and p-hacking
A p-value is the probability, assuming a specified null model, of data at least as extreme as observed. It is not the probability the hypothesis is true. p-hacking selectively analyses or reports results to obtain significance.
Trap → More rows cannot repair biased measurement or an invalid comparison.
2b · Graph-to-claim audit C3 worked method
Start at the denominator. Define who or what each mark represents, the population from which observations were selected, and whether the axes or display preserve the comparison. A familiar term is not enough: state the precondition under which the measure is relevant.
| Finding | Permitted wording |
|---|---|
| pattern in selected data | “is associated with” in that sample |
| controlled comparison | difference under specified conditions |
3 · Uncertainty language C4
| Type | Effect on conclusion |
|---|---|
| Random uncertainty | Variation across observations; report interval/range and avoid false precision. |
| Systematic uncertainty | Bias or untested assumption; more observations alone do not remove it. |
| Unknown mechanism | Association may remain useful, but causal language must narrow. |
calibration ladderestablishes → supports → is consistent with → suggests → cannot distinguish
Uncertainty is not ignorance and does not make every claim equally credible. State whether the problem is random variation, systematic bias or an untested assumption because each changes the conclusion differently.
4 · Repeat the right thing C5
| Term | What repeats? |
|---|---|
| Reproducibility | same data + same analysis → same result |
| Replicability | new data test the finding |
| Robustness | reasonable analytical changes preserve it |
One failed replication is evidence to investigate, not an automatic verdict of fraud. Check power, method and boundary conditions; publication incentives and questionable practices can affect what becomes visible.
4b · Fair comparison + report C6
Controls and comparison change the intended factor while holding plausible alternatives stable. A confounder offers another route to the outcome. Report architecture keeps observation separate from interpretation.
| Design check | Report check |
|---|---|
| comparison addresses the question | state what was observed first |
| plausible confounders controlled | interpret through evidence, not rhetoric |
| a result could count against claim | report uncertainty and variation |
5 · Values inside science C7 · week 7
Values can enter question selection, method and policy response. Make each entry point explicit: what is prioritised, whose interests or harms matter, and how the choice changes the evidence sought or the action threshold.
| Stage | Evidence job | Value job |
|---|---|---|
| Question | define answerable comparison | choose which problem matters |
| Method | measure consequences reliably | set ethical constraints |
| Decision | estimate outcomes + uncertainty | weigh harms, benefits, distribution |
Value influence does not mean evidence can be replaced by preference. A decision threshold may depend on consequences of error; the empirical claim still needs evidential support.
6 · FLICC denial audit glossary
| Fake experts | credentials or consensus misrepresented |
| Logical fallacies | conclusion does not follow |
| Impossible expectations | certainty demanded before action |
| Cherry-picking | selective evidence hides the whole pattern |
| Conspiracy theories | counter-evidence becomes proof of a cover-up |
MCQ → Name the technique from the operation, not from whether you agree with the speaker.
6b · Decision-threshold logic C7
Under uncertainty, accepting and rejecting a claim can each produce harm. The threshold is therefore not merely “is the evidence perfect?” but “what level of support is appropriate given the consequences of each possible error?”
exam sentenceBecause a false [accept/reject] would harm [group/consequence], the decision threshold should [rise/fall]; however, the empirical claim remains limited by [evidence/uncertainty].
| Bad move | Repair |
|---|---|
| hide the value judgement | name the prioritised harm/benefit |
7 · Ecosystem evidence → action C8 · week 8/9 choice
Begin with the biodiversity measure at the relevant spatial and temporal scale. Connect a driver to the pressure it exerts, identify the expected ecological indicator, then ask whether intervention evidence supports the management decision.
mechanism chaindriver → pressure → ecological change → measured indicator → intervention → monitored response
| Link | Failure question |
|---|---|
| measure → state | Does the indicator capture the claimed biodiversity feature? |
| driver → pressure | Could another driver create the same pattern? |
| pressure → change | Is the mechanism shown or only associated? |
| intervention → response | Is comparison evidence local and scale-appropriate? |
A global trend cannot be transferred to a local site without scale-appropriate evidence. A global decline motivates inquiry; it does not estimate a site's loss or identify its cause.
Fast comparison
Indicator describes a pattern. Driver/pressure proposes why it changes. Intervention evidence tests whether acting on that mechanism changes the indicator. Do not collapse them into one claim.
7b · Scale + intervention ledger C8
| Write down | Why it earns space |
|---|---|
| indicator + unit | prevents an undefined “biodiversity” claim |
| site + time window | fixes the scale of inference |
| driver + pressure verb | makes the causal route inspectable |
| alternative driver | shows what the same pattern cannot distinguish |
| intervention comparator | turns action into a test |
| response + uncertainty | states what would maintain, narrow or reverse action |
Cause–effect wording: “The indicator shows [pattern] at [scale]. Driver X could produce it through pressure Y, but alternative Z remains plausible. Compare [intervention] with [alternative/control] and monitor [response] before claiming local effectiveness.”
8 · Pollution causal chain C9 · week 8/9 choice
Build the causal chain before recommending monitoring or control. A named pollutant is only a hazard; realised risk requires a viable pathway to a receptor plus relevant exposure or dose.
source–pathway–receptorsource → transport/contact pathway → receptor → exposure/dose → possible effect → risk communication
| If the chain breaks at… | Do this |
|---|---|
| source | verify presence, form and release |
| pathway | test transport/contact route |
| receptor | define who or what can be exposed |
| dose | measure magnitude, duration and relevant variation |
| effect | separate association from mechanism |
Hazard presence and realised risk are different claims. Communicate what is known, what remains uncertain and which measurement could change the recommendation.
9 · Knowledge systems C10–C11
| Concept | Distinct job |
|---|---|
| disciplinary diversity | brings different methods and questions |
| standpoint + expertise | social position can reveal hidden relations |
| collaborative evidence | combines contributions without extraction |
| Country + relational knowledge | changes system boundary and relevant relations |
| observation across generations | extends place-based pattern recognition |
| ethical engagement | preserves holder, context and permission |
Do not homogenise distinct peoples, countries or knowledge traditions.