Week 3 Tutorial: Data Acquisition: Photography as a Quantitative Asset

Treat profile photos as traceable information assets. Test lighting, identity ambiguity and detection confidence using fictional image records.

11 March 2027

Inputs#

Bring the Platform audit and the Week 3 photo kit, or three equivalent fictional asset descriptions/drawings. Use the audit’s source/claim/unknown fields to start the manifest; keep unknowns instead of filling them from appearance. No identifiable real people are needed.

Finish with The photo desk needs receipts: the interactive quiz. Six new cases include a crop diagram, lighting readings and a reviewer table. Apply the manifest rules and calculations, then submit all six responses to unlock the score and worked explanations. The quiz is practice and carries no course marks.

Follow the 60-minute lecture selection and independent-work plan. Complete the preparation and open the workbook before Thursday. The four blocks below are the entire 90-minute tutorial; use existing lecture or study working for review and refinement. Remaining pack material belongs to the week’s independent study allocation.

90-minute tutorial#

1. 20 minutes — define fields (slides 4–9 and 36–39). Record asset ID, provenance, exercise permission, intended subject, source size, crop rectangle, revision, ambiguity and alt text. A has a supplied Alex label; B’s identity is unresolved; C’s exercise permission is unknown. Keep mock detection confidence separate from identity evidence.
2. 20 minutes — calculate (slides 10–21 and 28–35). Reproduce 200/100 = 2:1 and the synthetic 6/10, 2/10 and 3/10 failure rates. Verify that A’s 300×400 crop retains half a 600×400 source. Compare the two group-reviewer columns: six failures each, but only 8/10 label agreement. Name every denominator.
3. 25 minutes — challenge the metric (slides 22–26 and 30). Inspect B’s source and partial-person crop. Explain why the mock 0.98 detection score does not resolve subject identity. Flag one ambiguous crop among three records without inventing a relationship. State why 1/3 crop ambiguity does not imply that the other two assets pass every check.
4. 25 minutes — select, freeze and take the quiz (slides 40–47). Use fifteen minutes to implement pass/block with reasons; test a blank source and an out-of-bounds rectangle. Select A-portrait-v1 in the supplied fixture, preserve B and C with their reasons, and record file, crop, revision, alt text and byte fingerprint. Repair the proposed bio comparison that changes both photo and text. Use the final ten minutes for the six-case quiz: about eight minutes to respond and two to review the explanations after all cases are answered.

Deliverable: Photo asset manifest#

Save three records with sources and derivatives, the selected revision, calculations and a reason for each rejection. Use worked-examples.mjs from the extracted kit to review the reference calculations, or show equivalent working in another language. Solo work and review are accepted. A text-only placeholder with informative alt text remains sufficient; artistic skill is not assessed.

Week 4 must hold file bytes, crop, alt text and display treatment constant while comparing bios. That dependency prevents a photo change from masquerading as a bio effect. The manifest also prepares the data report’s explanation of controlled non-bio inputs.

Interactive quiz · 6 cases · about 8 minutes + 2 to review

The photo desk needs receipts

The crop looks good. The metadata would like a word.

Make a decision in every case, then select Reveal answers. You can change any response before submitting. Scores and explanations stay hidden until the whole quiz is answered.

All images, records, readings and reviewer labels are authored teaching fixtures. These separate examples add no facts about Alex. Practice only: no course marks. Responses stay on this page; reloading starts a fresh attempt.

The cases are readable without JavaScript. Enable JavaScript to navigate, submit your responses and unlock the explanations.

Case 01 / 06The suspiciously tidy record

Preserve provenance and uncertainty

A fictional portrait has a recorded source, a supported subject label, an in-bounds crop and useful alt text. Its deliberately incomplete exercise record says permission: unknown. A teammate wants to change that field to yes because the image looks professional.

What should your manifest do with this record?

AApprove it: a professional-looking image supplies the missing permission.BRetain the record and block selection until the required permission evidence is supplied.CDelete the rejected row so the manifest contains only clean-looking records.DMark permission as yes, but lower the detection score to compensate.

Case 02 / 06The five-to-one sales pitch

Use the declared lighting convention

In the same setup and arbitrary unit, the isolated key-source reading is 300 and the isolated fill-source reading is 75. A teammate adds them and announces a 5:1 ratio. This exercise defines key-to-fill as isolated key divided by isolated fill.

Which correction follows this convention?

A5:1: the key reading must include the fill reading.B4:1, which proves that this setup will get more matches.C4:1. Doubling both readings preserves the ratio, without proving a better photo.D225:1: subtract the fill from the key before reporting the ratio.

Case 03 / 06The crop invoice

Calculate retained area

An 800 × 600 source uses a top-left origin. The crop starts at (200, 100), has width 400 and height 300, and lies inside the source. No resizing happens in this calculation.

Model supplied
Retained area (%) = 100 × (crop width × crop height) / (source width × source height).

What percentage of the source area is retained? Enter the percentage as a number, without the % sign.

Retained source area (%)

Enter one number. A calculator and your lecture notes are welcome.

Case 04 / 06The extremely confident stranger

Separate detection from identity

A course-drawn group image shows two unlabelled fictional adults. Its crop keeps one adult and part of the other at the edge. A mock face-presence score is 0.99; there is no supplied subject label or relationship history.

Which record is supported by those facts?

ASubject: Alex. A score of 0.99 identifies the complete figure.BAlt text: Alex has cropped out an ex. The partial figure supplies the history.CPermission: yes. High face-presence confidence establishes consent.DSubject unresolved; crop ambiguous. Describe the two unlabelled adults and retain the source.

Case 05 / 06Same totals, different witnesses

Compare annotations row by row

Two fictional reviewers label the same eight identification tasks: 1 means the supplied cues fail to identify the intended subject uniquely; 0 means pass. Each reports four failures. A teammate claims their annotations therefore agree completely.

Invented trial labels · a separate fixture from the portrait records

Trial	Reviewer 1	Reviewer 2
T1	1	1
T2	1	0
T3	1	0
T4	1	1
T5	0	1
T6	0	1
T7	0	0
T8	0	0

What should the audit report about agreement?

AThey agree on 4/8 labels, or 50%. Agreement alone does not establish correctness.BThey agree on 8/8 labels because each reported four failures.CThey agree on 2/8 labels; matching passes should not count.DReviewer 1 is more accurate because their failures occur earlier in the table.

Case 06 / 06One test, two makeovers

Freeze non-bio inputs for Week 4

A proposed bio comparison gives bio A portrait revision v1 and bio B a brighter, tighter crop exported as v2. Their filenames share the same prefix, but their bytes, crop and alt text differ. The planned question concerns the bio wording.

Which repair makes the image treatment constant?

ARename both exports portrait.svg; matching names guarantee identical image treatment.BUse whichever export has the higher detection score in each arm.CUse the same frozen bytes, crop, alt text and display treatment in both arms; change the bio text.DKeep both photo changes and report any response difference as the bio effect.

Reveal answers

Answer all 6 cases to unlock your score and the worked explanations.