Scientific Framework for structured, traceable and collaborative scientific work with AI.
The repository contains the General Scientific Layer (GSL) and the depending Genealogy Model (GM). The Genealogy Model specializes the general scientific architecture for genealogical research and provides concrete GM Executions for defined scientific tasks.
The GSL is intended to provide a domain-independent scientific foundation; GM is its first domain-specific specialization. The framework is currently validated primarily through genealogical research. Future specialization of the GSL into other domain models would provide an important test of whether its architecture is sufficiently general and domain-independent.
The framework is designed around explicit scientific responsibilities, traceable artifacts and relationships, controlled human--AI collaboration, and graph-based research structures.
General Scientific Layer (GSL)
↓
Genealogy Model Core (GM Core)
↓
GM Executions
The GSL defines the general scientific principles and architecture used by the framework. It provides the normative foundation for scientific working context, evidence and knowledge handling, traceability, uncertainty, human--AI collaboration, and controlled scientific evolution.
The GSL is intended to support future domain models that do not inherit from GM Core. Such non-genealogical specialization is a future validation path, not a currently demonstrated field-test result.
Current baseline:
GM Core specializes the GSL for genealogical research. It defines the common genealogical scientific architecture, including artifacts, relationships, project context, processing boundaries, and the interface between GM Core and GM Executions.
Current baseline:
GM Executions define controlled processing for specific genealogical scientific tasks. They inherit from GSL and GM Core and specialize only the execution-specific scientific responsibilities required for their task.
Published Executions currently include:
GM-Exec_Person_Identity_Research_v2.3_Baseline.md--- BaselineGM-Exec_Source_Registration-v1.2_Candidate.md--- CandidateGM-Exec_Transcription-v1.2_Baseline.md--- Baseline
The GM-Exec Specification Development Template supports the controlled development of additional GM Executions.
It is a development aid and is not an additional normative layer. A GM Execution developed with the template inherits normatively from GSL and GM Core.
Recommended repository location:
The framework is intended to be usable directly with a capable Large Language Model (LLM), such as ChatGPT:
- Load the applicable framework files into the LLM --- GSL, GM Core, and the GM Execution required for the research task.
- Provide the research material and define the research objective and applicable project context.
- Begin the research collaboratively within the scientific and execution boundaries defined by the loaded specifications.
- Preserve validated results in the project's persistent research structure so that later work can continue from an explicit scientific state.
The specifications are designed to provide the LLM with a structured scientific working context rather than relying only on conversational instructions.
The framework is designed for scientific work with explicit provenance, traceability, uncertainty, validation, and controlled human--AI collaboration. Scientific responsibilities are assigned to defined architectural owners instead of being left implicit in the conversation.
Source-near observation is kept distinct from subsequent interpretation, correlation, hypothesis formation, identity research, and other research activities. This helps preserve what was actually observed in a source while allowing later conclusions to remain reviewable and revisable.
Research knowledge is represented through scientific artifacts and explicit relationships between them. This graph-oriented structure allows persons, events, information, sources, observations, and other research elements to remain individually addressable while their scientific context is preserved through relationships.
Validated scientific results can be preserved as persistent project state rather than existing only in the LLM conversation. Persistent artifacts, relationships, provenance, and authoritative register states provide a basis for continuing research across sessions, tools, and processing steps.
The framework separates general scientific responsibilities from genealogy-specific semantics and execution-specific procedures.
A central architectural rule is Single Owner: each normative scientific responsibility has exactly one normative owner. Other architectural elements reference or inherit that responsibility instead of redefining it.
The framework also follows the development principle "Normiere -- entwickle nicht neu." Established scientific or technical concepts are reused where they adequately satisfy the scientific requirement; the framework specifies the conditions for their scientific use instead of independently redefining them.
This publication is intended as a public field test of the framework.
The current architecture has been developed and repeatedly tested through practical genealogical research. Publication is intended to extend that evaluation to other researchers, source types, research questions, workflows, and tool environments.
Feedback is especially valuable when it identifies a concrete scientific or architectural problem rather than only proposing a textual change.
Where possible, feedback should therefore be submitted as a qualified issue describing:
- the concrete research use case or source context;
- the observed problem or limitation;
- the affected framework layer or specification, if identifiable;
- the current behavior or applicable rule;
- the expected scientific requirement;
- evidence or a reproducible example;
- whether the issue appears to require clarification, specialization, correction, or a specification change.
For proposed normative changes, contributors are encouraged to formulate the issue in terms compatible with the Scientific Change Proposal (SCP) concepts defined by the GSL. This makes the scientific motivation, affected responsibility, evidence, and proposed change assessable before implementation.
A Pull Request may then implement an already understood and assessable change. A Pull Request alone is not a substitute for describing the scientific problem that motivates a normative change.
This distinction is intentional:
Research experience
↓
Qualified issue / evidence
↓
Scientific assessment
↓
SCP where a normative change is required
↓
Controlled specification evolution
↓
Implementation / Pull Request where appropriate
Reports showing that the framework works without modification in a new use case are also valuable field-test evidence.
The framework supports both a central contribution path and independent experimentation.
Central contribution path
Field test
↓
Qualified issue / evidence
↓
Scientific assessment
↓
SCP where required
↓
Controlled integration
Contributors can report findings to the central repository for assessment against the current GSL, GM Core, and applicable GM Execution. This provides a common place for evaluating whether an observed problem requires clarification, specialization, correction, or normative evolution.
Independent development path
Repository
↓
Fork
↓
Adaptation / independent field test
↓
Findings
↓
optional contribution back to the central project
A fork may explore another application, adaptation, or development path independently. Findings from such work can later be contributed back as qualified issues, evidence, SCP input, or implementation proposals where appropriate.
The central repository therefore provides a controlled reference lineage without preventing independent scientific experimentation.
This repository is under active development. Documents explicitly marked Baseline represent released reference states. Documents marked Candidate are publication candidates still subject to their controlled validation and promotion process.
This project is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) license.
See LICENSE for the full license terms.