Agent Skills
› brycewang-stanford/Awesome-Journal-Skills
› amr-theory-development
amr-theory-development
GitHub用于构建AMR论文理论，将谜题转化为明确构念、关系、命题及边界条件。按顺序定义构念、假设、关系、机制、命题和边界，确保逻辑严谨，避免仅做断言。
触发场景
已设定谜题和对话，需构建理论
有命题但未定义底层构念
关系仅为断言而非推导得出
未说明边界条件或假设
安装
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill amr-theory-development -g -y
SKILL.md
Frontmatter
{
"name": "amr-theory-development",
"description": "Use when building the actual theory for an Academy of Management Review (AMR) manuscript — turning a positioned puzzle into defined constructs, explicit relationships, propositions, and boundary conditions. Constructs the theory; it does NOT stress-test the argument's logic (that is amr-data-analysis) or design any data collection (AMR has none)."
}
Theory Development: Constructs, Relationships, Propositions (amr-theory-development)
When to trigger
- The puzzle and conversation are set; now you must build the theory
- You have propositions but no defined constructs underneath them
- Your relationships are asserted, not derived
- You have not stated boundary conditions or assumptions
The build order
AMR theory is assembled in a disciplined sequence. Skipping a step is the most common reason a draft reads as "assertions, not theory."
- Define the constructs. For each new or re-specified construct: a precise conceptual definition, its domain (what it includes and, crucially, excludes), scope conditions, and how it differs from neighboring constructs. This is the construct-clarity standard from Suddaby's AMR editorial (2010, DOI 10.5465/amr.2010.0419): definition, scope conditions, semantic relationships, coherence. A construct that overlaps an existing one without distinction will be challenged as a relabel.
- State the assumptions. What must be true about actors, the setting, and the world for the theory to operate? Make them explicit; hidden assumptions are where reviewers attack.
- Specify the relationships. Name the form of each link: causal, recursive, moderating, mediating, constitutive. "X relates to Y" is not a relationship — specify direction, shape, and why.
- Articulate the mechanism. The why. The mechanism is the engine that makes the
relationship hold. A proposition without a mechanism is a guess. Whetten's AMR editorial
(1989, DOI 10.5465/amr.1989.4308371) is explicit that the Why is what makes theory
theory; the What and How without the Why is description. (Develop the mechanism logic with
amr-data-analysis.) - Write the propositions. Each proposition is a falsifiable theoretical claim that follows from the constructs + mechanism. Number them (P1, P2...). Each must be preceded by the argument that earns it.
- Set boundary conditions. State when the theory holds and when it does not — the contextual, temporal, and level limits. Boundary conditions are a contribution, not a disclaimer.
Construct definition template
For each construct, fill:
- Label (one term, used consistently throughout)
- Conceptual definition (one or two precise sentences)
- Domain — included instances / excluded instances
- Distinction — how it differs from the nearest existing construct
- Level of theory (individual / dyad / group / organization / field)
Proposition discipline
- A proposition states a relationship between constructs, not an empirical magnitude.
- Every proposition is preceded by its logical argument; the proposition is the summary of an argument already made.
- Propositions build on each other; later ones may depend on earlier constructs.
- Keep them theoretical (P: "the stronger A, the more likely B, because mechanism M"), never operational ("A measured by survey predicts B at p<.05").
- Never write "H1" or "supported / not supported" — that support/rejection vocabulary is AMJ's, not AMR's. Each proposition should be testable in principle (AMR's scope: "testable knowledge-based claims") but you do no testing.
Exemplar
Oliver's "Strategic Responses to Institutional Processes" (AMR 1991, DOI 10.5465/amr.1991.4279002) builds a typology of five responses (acquiescence, compromise, avoidance, defiance, manipulation) and derives propositions linking antecedent conditions (cause, constituents, content, control, context) to them — defined constructs, explicit mechanisms, numbered propositions, stated boundaries, and not a single datum.
Checklist
- Every construct has a definition, a domain, and a distinction from neighbors
- All assumptions are stated explicitly
- Each relationship's form is specified (causal / recursive / moderating / mediating / constitutive)
- Each proposition has an explicit mechanism behind it
- Propositions are numbered and each is earned by a preceding argument
- Boundary conditions specify where the theory holds and fails
- Level(s) of theory are consistent (or level shifts are theorized, not accidental)
Anti-patterns
- Propositions stated as bullet points with no preceding logical argument
- A new construct that is an old construct with a new name
- "X affects Y" with no specified form and no mechanism
- Assumptions left implicit, then exposed by reviewers
- A "process model" figure standing in for the theoretical argument it should summarize
- Boundary conditions written as apologies ("limited to...") rather than as theory
- Smuggling in empirical claims ("data show...") — AMR has no data
Output format
【Constructs】[name: definition / domain / distinction / level] for each
【Assumptions】[explicit list]
【Relationships】[construct → construct : form + mechanism]
【Propositions】P1...Pn (each with one-line mechanism)
【Boundary conditions】where the theory holds / fails
【Next step】amr-methods (refine construction) → amr-data-analysis (logic check)
版本历史
- 1839142 当前 2026-07-05 12:14
