Research State Machine¶

This document specifies the orchestration model used by using-humanities-superpowers.

Purpose¶

The state machine prevents fluent output from being mistaken for completed research. States describe which scholarly objects currently exist and which operation is valid next. They do not certify truth or quality.

States¶

State	Minimum observable condition	Common blocking condition
`IDEA`	Topic, intuition, phenomenon, or problem is recorded	No arguable research question
`QUESTION`	A provisional research question exists	Scope, feasibility, or hidden conclusion unresolved
`SCOPE`	Inclusion, exclusion, corpus, and claim scale are explicit	Necessary materials unavailable
`SOURCES`	Source records and verification states exist	Central evidence absent or unverified
`INTERPRETATION`	Bounded readings connect observable features to interpretive moves	Primary object absent or alternative reading ignored
`ARGUMENT`	Claims, evidence, warrants, qualifications, and objections are mapped	Central warrant or evidence path missing
`STRUCTURE`	Researcher material is organized by argumentative function	Prose conceals unresolved argument gaps
`REVIEW`	Manuscript or argument is under systematic challenge	Review lacks manuscript, scope, or evidence context
`REVISION`	Problems or reviewer comments have revision actions	Planned actions are falsely reported as completed
`SUBMISSION`	Final manuscript and package are being verified	Citation, privacy, package, or venue requirement fails
`BLOCKED`	A required item or decision prevents valid progression	Source, permission, ethics review, or author decision absent
`UNKNOWN`	Evidence is insufficient to classify the state	No discriminating artifact supplied

Transition rule¶

A transition is valid only when:

the current state's required object exists;
the selected skill's prerequisites are satisfied;
the controlling gate is not FAIL;
carried conditions remain explicitly represented;
no researcher-controlled decision has been silently automated.

Normal transitions¶

IDEA → QUESTION → SCOPE → SOURCES
SOURCES → INTERPRETATION and/or ARGUMENT
INTERPRETATION → ARGUMENT
ARGUMENT → STRUCTURE → REVIEW → REVISION → SUBMISSION

The path is intentionally nonlinear. A source discovery may return a project from ARGUMENT to SCOPE; a reviewer may return it from REVISION to CONCEPT LINEAGE; a submission audit may stay within SUBMISSION while repairing file metadata.

Gate-driven actions¶

Gate status	Default orchestration action
`PASS`	Advance if the next skill's prerequisites are satisfied
`CONDITIONAL PASS`	Advance only with named conditions and no invalid conversion
`FAIL`	Roll back, pause, request researcher decision, or stop

Rollback principle¶

Rollback targets the earliest causal defect, not the latest visible symptom.

Unsupported generalization → scope.
Unsupported central claim → literature dialogue, close reading, or argument planning.
Concept drift → concept lineage and terminology audit.
Unverified citation → citation audit; if central support collapses, argument planning.
Reviewer request requiring a new corpus → researcher decision and scope.
Hidden author identity in blinded files → submission package repair only.

Completion semantics¶

SUBMISSION is a workflow state. PASS is a gate result. They are not synonyms.

A project may end with:

SUBMISSION + PASS: the named checks passed for the inspected package;
SUBMISSION + CONDITIONAL PASS: non-blocking conditions remain and are disclosed;
SUBMISSION + FAIL: submission should not proceed until blocking defects are repaired.

No state authorizes claims of truth, originality, or publication success.