Unit 9 activity: Grounding and citation checking

Time: about 30 minutes. Read page.md first, at least as far as the six codes.

This is the module's second worked case, and it has the same shape as the first: build the reference, compare item by item, stop at the first divergence, name the fault, and finish with a statement you could defend.

Part 1 uses a specimen supplied below, so you can do the whole comparison without leaving this page. Part 2 puts it on a topic of your own, at a scale you can finish honestly.

Part 1 — the specimen (15 minutes)

About these sources

[FICTIONAL TEACHING SOURCES] The four sources below are invented for this exercise. Brackenford does not exist, none of these documents exists, and nothing in them should be quoted, cited or repeated anywhere else. They are here so you can carry out a full claim-by-claim comparison in fifteen minutes rather than an afternoon in a library, with every edition fixed and every passage available. Everything needed to adjudicate every claim is in the extracts.

The request

Write a short referenced summary of the Brackenford kerbside recycling scheme.
Eight claims, each with a citation to a specific location in a source.

Constructed specimen summary

This response was written for this exercise to exhibit the failure modes. It is not an observed model output, and nothing here says how often any system produces any of them.

1. Brackenford's kerbside recycling scheme began in 2011
   (Brackenford Borough Council 2019, section 1.1).

2. Participation increased by 19.6 percentage points between 2012 and 2018
   (Hallam 2021, p. 94).

3. Hallam (2021, p. 96) concluded that fortnightly collection causes higher
   contamination of recyclate.

4. Kerbside glass collection was introduced in 2015
   (Brackenford Borough Council 2019, section 4.2).

5. The council's current target is 60 per cent household recycling by 2028
   (Brackenford Borough Council 2019, section 2.3).

6. Okonjo and Lind (2020, p. 112) report that, where authorities printed sorting
   guidance on bin lids, sorting errors fell in the following year.

7. Okonjo and Lind (2020, p. 112) surveyed 1,200 households across three
   northern towns.

8. The 2023 strategy commits to weekly separate food-waste collection from 2026
   (Brackenford Borough Council 2023, section 5.2).

The sources

[FICTIONAL] SOURCE A -- Brackenford Borough Council (2019).
Kerbside Collection Review 2019. A 26-page document. Contents: section 1
Background, section 2 Targets, section 3 Materials, plus appendices. There is
no section 4.

  1.1  "Brackenford's kerbside recycling scheme began in 2011 with the
        collection of paper, card and mixed plastics."
  2.3  "The Council's target is 60 per cent household recycling by 2028."
  3.4  "Glass was added to the kerbside collection in 2015."

[FICTIONAL] SOURCE B -- Hallam, R. (2021). "Kerbside recycling
schemes." Journal of Municipal Waste Studies 14(2), 88-101.

  p. 94  "In Brackenford, participation increased by approximately 20
          percentage points between 2012 and 2018."
  p. 96  "Contamination was higher in the fortnightly-collection authorities.
          We are not able to separate the effect of collection frequency from
          that of housing type, which differs systematically between the two
          groups."
  p. 97  "Where authorities printed sorting guidance on the bin lid itself,
          sorting errors fell in the following year."

[FICTIONAL] SOURCE C -- Brackenford Borough Council (2023).
Waste Strategy 2023-2028.

  5.1  "This strategy supersedes the targets set in the 2019 Kerbside
        Collection Review."
  5.2  "The Council will introduce weekly separate food-waste collection from
        2026."
  5.4  "The Council commits to 65 per cent household recycling by 2028."

[FICTIONAL] SOURCE D -- Okonjo, T. and Lind, S. (2020). Household Sorting
Behaviour. Riverbank Press. 240 pages.

  p. 112  "We surveyed 840 households across three northern towns. The towns
           are not named, at the request of the participating authorities."

  (The book does not discuss labelling or bin design anywhere.)

What to do

Build the reference first, then compare. Do not skip to the comparison — the order is the exercise, exactly as it was in Unit 6.

1. Build the reference. One row per locator the summary cites, plus a row for any passage that changes how another source may be used. Use the three columns from the reading: source and locator, what the passage supports, and the limitation, scope or version note.
2. Compare, claim by claim, in order. For each, ask: does the cited passage support this, as stated, at this strength?
3. Record the first divergence. Which claim number, and which code.
4. Then keep going and classify the rest. Stopping at the first divergence is where the diagnosis stops, not the work.

Use this table:

Work the decision order from the reading rather than picking the label that feels right. Several claims could be described in more than one way; the order is what makes one of them the primary diagnosis.

Part 2 — two claims of your own (12 minutes)

Now on a real topic, at a scale you can actually finish.

1. Ask a system for a short referenced summary of something in your own subject. Six to eight claims with citations, exactly as above. Keep the prompt and the output untouched — Unit 7's discipline.
2. Pick two claims only. Not all of them. Two claims checked properly is worth more than eight skimmed, and it is the honest amount of work for twelve minutes.
3. Retrieve the sources for those two and build the reference: what does the passage at the cited locator actually say?
4. Classify each. If a claim is supported, say so — that is a result too.

Record what you could not retrieve. A paywall, a dead link, an edition you cannot get: those are findings, and they change what you can claim at the end.

Part 3 — the scoped statement (3 minutes)

Write two or three sentences about your own two claims, in the form the reading gives: what you checked, what you found, what you could not do, and what follows.

Then apply one test to what you wrote:

Could a reader who does not trust me, and cannot see my screen, tell what evidence exists?

Keep all three parts. Unit 17 asks for an evidence pack, and this comparison table is what goes into it if you took the source route rather than the code route. They are the same deliverable.

Now take the checkpoints

cp09-diagnose-source gives you six fresh observations and asks which code each one is. cp09-scope-the-claim-source is about Part 3.

python3 selfcheck.py run cp09-diagnose-source
python3 selfcheck.py run cp09-scope-the-claim-source  # or in the browser

Run the Python route from course/lab/, or use the browser self-check. Nothing in either checkpoint requires you to have run any code.

Optional extensions

Optional. These sit outside the core study time, are not required for the checkpoints, and nothing later in the module depends on them.

Ask for the quotation (about 6 minutes). Go back to one of your Part 2 claims and ask the system to quote the exact sentence its citation rests on. Then compare that quotation with the source. There are three outcomes — the sentence is there and says this, it is there and says something adjacent, or it is not there — and all three are more informative than the original claim was.

Check a claim that has no source at all (about 8 minutes). Take an unsourced claim from the same summary and try to find out whether it is true. Time it. The gap between that and the thirty seconds a locator costs you is the whole argument for asking for citations, and it lands harder when you have paid it once.