# Extraction report: wundfolkspuristspragmatic

## What I processed

- Target PDF: `inbox/Wu_2023.pdf`
- Ingested + processed to: `papers/wundfolkspuristspragmatic/`

## Evidence sources used (pipeline outputs first)

- `papers/wundfolkspuristspragmatic/out/tables/*` for counts/percentages (used for all extracted effects).
- `papers/wundfolkspuristspragmatic/out/fulltext.md` and `papers/wundfolkspuristspragmatic/out/tei.xml` for study descriptions and quotes.
- Rendered page images (pipeline output) used to confirm bibliographic details on the first page (e.g., journal year/volume/pages): `papers/wundfolkspuristspragmatic/out/images/pages/p001.png`.

## Computations

- Computed `d` and `v` from exact 2×2 count tables (stakes × knowledge attribution) using `esc::esc_2x2` (OR -> d) in:
  - `papers/wundfolkspuristspragmatic/analysis/effect_sizes.qmd`
  - Rendered output: `papers/wundfolkspuristspragmatic/analysis/effect_sizes.html`

## Recoding decision — paradigm classification

- Updated on 2026-04-21: for analytic purposes, all three Wu experiments are now coded as `Agreement with knowledge claim` rather than `Other`.
- This includes:
  - Experiment 1 and Experiment 2, where the raw response options jointly combine knowledge attribution and action/rechecking content;
  - Experiment 3, where the binary choice is phrased as `knows without checking` vs `does not know without checking`.
- The rationale for the recode is analytic rather than taxonomic: in all three experiments, the extracted effect size is ultimately a knowledge-attribution contrast, and the project decision was to treat these studies as direct knowledge-attribution variants in the meta-analysis.
- This coding decision should be disclosed in the manuscript/reporting text if paradigm definitions are discussed.

## Issues / uncertainties for human review

### 1) Airport vignette: evidence source missing in extracted text (moderators)

The extracted outputs summarize the Airport cases’ stakes but do not specify what evidence the protagonist has for the key proposition (e.g., ticket/itinerary/testimony/memory/observation). This blocks moderator coding for:

- `effects[].moderators.evidence` (First Person vs External)
- potentially `effects[].moderators.evidence_reliability`

Impacted effects:
- Study 1: `s1_e2`
- Study 2: `s2_e2`
- Study 3: `s3_e2`

**Answer here (what is the evidence source in the Airport vignette, and is any reliability cue present?):**

```text
ANSWER:
```

### 2) Spelling/Typos vignette: evidence source missing in extracted text (moderators)

Similarly, the extracted outputs summarize stakes for the Spelling cases but do not specify the protagonist’s evidence source for the proposition (or any explicit reliability cue).

Impacted effects:
- Study 1: `s1_e3`
- Study 2: `s2_e3`
- Study 3: `s3_e3`

**Answer here (what is the evidence source in the Spelling/Typos vignette, and is any reliability cue present?):**

```text
ANSWER:
```

### 3) Skeptical-pressure coding for the Bank vignette

I coded `skeptical_pressure: "Yes"` for the Bank vignette because the vignette text includes an explicit alternative possibility (“Many banks are closed on Saturdays…”), but this is not the canonical “another character raises doubt” skeptical-pressure manipulation used in some other bank-case papers.

Impacted effects:
- Study 1: `s1_e1`
- Study 2: `s2_e1`
- Study 3: `s3_e1`

**Answer here (should “Many banks are closed on Saturdays” be coded as skeptical pressure under this project’s rules?):**

```text
ANSWER:
```

### 4) Publication year choice (2025 vs online-first year)

I set `paper.year: 2025` based on the first page header showing Episteme (2025), volume 22, pages 304–332. The DOI includes “2023”, and the paper may have been available online earlier.

**Answer here (preferred year coding rule for this project: issue year vs online-first year?):**

```text
ANSWER:
```