---
name: brycewang-stanford/aos-methods
source: https://app.decimal.ai/s/brycewang-stanford-aos-methods@1/SKILL.md
source_sha256: 4300e742c8de
---

# Research Design & Methods (aos-methods)

## When to trigger

- Choosing between field study, experiment, survey, or historical design for a settled question
- Negotiating field access, case selection, or interview/observation protocols
- Designing manipulations and participant pools for a behavioral experiment
- Ethics approval, informed consent, or site-anonymity commitments are still open
- A reviewer says the design cannot support the paper's interpretive or causal claims

## Methodological pluralism, single standard

AOS accepts any method suited to studying accounting's social operation, and holds each to the rigor bar of its own tradition. The design must let the *theoretical lens* meet the *empirical material* on fair terms — enough depth for interpretation, enough control for causal process claims, enough sources for historical reconstruction.

## Qualitative field study (the modal AOS design)

- **Case logic.** Justify the site theoretically: extreme, revelatory, or transparently typical for the phenomenon. "We had access" is a fact, not a rationale — convert access into an argument about what this site lets you see.
- **Data plan.** Triangulate interviews, observation, and internal documents; log counts, roles, durations, and periods as you go (reviewers will ask for a data-inventory table). Plan for prolonged or repeated engagement — one-shot interview rounds read as thin at AOS.
- **Theory–data discipline.** Follow the methodology the journal itself canonized: Ahrens & Chapman (2006) — the field problem, the theory, and the data must be *positioned* against each other continuously, not assembled at the end.
- **Reflexivity and access terms.** Record your own role at the site, what informants were promised (anonymity, review of quotes), and any sponsorship that could shape what you were shown.

## Behavioral experiment

- Theorize the process first (see `aos-theory-development`); the manipulation must move the *theorized construct* and nothing else that plausibly moves the outcome.
- Choose the pool the inference needs — professional participants (auditors, managers, analysts) when expertise is the point; online panels only where the process is expertise-independent, with justification.
- Build in manipulation checks, attention screens, randomization, and a planned process measure (mediator) — AOS behavioral papers live or die on process evidence.
- Pre-specify exclusion rules and, where feasible, pre-register.

## Survey and historical designs

- **Survey:** validated or carefully developed instruments; a defined frame and response-rate accounting; common-method-bias mitigation by design (separation, sources) before statistics.
- **Historical/genealogical:** name the archives, their custodians, and coverage gaps; treat documents as traces produced for purposes — source criticism is the identification strategy of historical work.

## Ethics and anonymity (settle before collection)

- Human-participant work (interviews, experiments, surveys) needs institutional IRB/REB approval and documented informed consent; organize this **before** entering the field or the lab, and keep the paperwork — editors can ask.
- Field sites usually require pseudonyms and disguised identifying details; agree the terms with the organization in writing, and check that disguise does not destroy the theoretically relevant detail.
- Declare in the manuscript how consent and anonymity were handled; AOS review is double-anonymized, so scrub author-identifying access stories from the submitted files.

## Checklist

- [ ] Design chosen because the question needs it; the inference each claim requires is stated
- [ ] (Field) case logic argued; data inventory planned; engagement duration adequate; reflexivity noted
- [ ] (Experiment) manipulation isolates the construct; pool justified; process measure and checks designed
- [ ] (Survey) instrument validity and method-bias mitigation designed in
- [ ] (Historical) archives named; source criticism planned
- [ ] IRB/REB approval and consent documented; site-anonymity terms agreed in writing

## Anti-patterns

- **Convenience ethnography**: a handful of interviews at an accessible firm presented as a field study.
- **Confounded manipulations** that move attention or effort along with the theorized construct.
- **Anonymity that erases the evidence** — disguising the site so heavily the practices become uninterpretable.
- **Ethics as afterthought**: collecting first, seeking approval later.
- **Method mimicry**: dressing an interpretive study in hypothesis-testing scaffolding to look "rigorous."

## Output format

```
【Design】qualitative-field / experiment / survey / historical / archival+theory
【Case or pool logic】why this site / these participants / these archives ...
【Data plan】sources, counts, periods, triangulation ...
【Process controls (if experimental)】manipulation, checks, mediator ...
【Ethics & anonymity】IRB/REB status, consent, site terms ...
【Next step】aos-data-analysis
```