The Moral Seismograph: A Causal Machine Learning Framework for Testing Ethical Theories
Fuente:
Zenodo
Salvato in:
| Autore principale: | |
|---|---|
| Natura: | Recurso digital |
| Pubblicazione: |
Zenodo
2025
|
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866901535307333632 |
|---|---|
| author | YANG, JIHOON |
| author_facet | YANG, JIHOON |
| contents | <p>This paper presents a validated, reproducible pipeline for translating philosophical theories into empirically testable predictions about structural transitions in complex systems. While designed with ethics in mind, the framework has been proven viable through successful applications in linguistics (N=40, AUC≈0.944) and genomics (N=520-1,150, AUC 0.675-0.948).</p> <p><strong>Three Methodological Principles:</strong></p> <ol> <li> <p><strong>Autonomous Event Detection</strong> - Eliminates human labeling bias by detecting structural transitions through statistical threshold-crossing (Z-score > 1.5) in intrinsic dynamics, removing documentation bias and enabling universal applicability.</p> </li> <li> <p><strong>No-Leak Filtering</strong> - Systematically removes temporal contamination (39% of features in linguistics study) to preserve causal validity, ensuring future information cannot contaminate past predictions.</p> </li> <li> <p><strong>Time-Aware Cross-Validation</strong> - Enforces strict temporal ordering (train on t ≤ threshold, test on t > threshold) across multiple percentiles to test genuine predictive capacity rather than spurious correlations.</p> </li> </ol> <p><strong>Theoretical Foundation:</strong> The framework operationalizes two philosophical constructs—Recovery-Transition Dynamics and Conditional Universality—as measurable computational features (recovery_signal, transition_signal, roughness, memory). These concepts, previously validated across music, constitutions, brain activity, and chromosomes, now demonstrate cross-domain applicability in symbolic (linguistic) and biological (genomic) systems.</p> <p><strong>Empirical Validation:</strong></p> <ul> <li> <p><strong>Linguistics Study (Yang, 2025):</strong> Historical explanation AUC≈0.944 with N=40; empirically determined N>100 required for future prediction (Exp2 with N=25 showed AUC range 0.33-1.0, honest boundary-finding).</p> </li> <li> <p><strong>Genomics Study (Yang, 2025):</strong> Demonstrated conditional universality through gene-specific minimal sufficient contexts: PIK3CA requires lineage only (AUC 0.745), KRAS requires lineage+expression synergy (AUC 0.675), MYC achieves high prediction across multiple contexts (AUC >0.92).</p> </li> </ul> <p><strong>What This Paper IS:</strong></p> <ul> <li>A theory-agnostic methodology for testing ANY structural transition theory</li> <li>A complete implementation guide (Section 5) with code specifications</li> <li>Proof-of-concept validation in two disparate domains</li> <li>A reproducible framework ready for ethics application</li> </ul> <p><strong>What This Paper is NOT:</strong></p> <ul> <li>Direct validation in ethical systems (explicitly stated in Section 6.2)</li> <li>Proof of a specific ethical theory's truth</li> <li>A completed ethics study (that is Paper 2, forthcoming)</li> </ul> <p><strong>Practical Applications:</strong> Researchers can now test their own theories of ethical norm transitions, political regime changes, financial crises, or any system exhibiting structural transitions. Complete pipeline specifications provided: data requirements (N≈40 for explanation, N≈100+ for prediction), feature engineering protocols, validation procedures, and expected challenges with mitigation strategies.</p> <p><strong>Scientific Honesty:</strong> This paper exemplifies rigorous boundary-finding by honestly reporting both successes and failures, establishing when the methodology works (historical explanation) and when it requires more data (future prediction), following the principle that "honest failure is as valuable as success."</p> <p><strong>Related Work:</strong> Part of a systematic research program applying recovery-transition dynamics across domains: music (χ²=12.34, p=0.0063), constitutions, neuroscience, genomics, and linguistics. This paper synthesizes methodological insights from these validations into a unified, generalizable framework.</p> <p><strong>Next Steps:</strong> Application to ethical norm transitions in social media (Yang, forthcoming) will provide the first direct test in the target domain.</p> <p><strong>Keywords:</strong> causal machine learning, computational ethics, structural transitions, conditional universality, recovery-transition dynamics, autonomous detection, no-leak validation, cross-domain methodology, empirical philosophy, time-series prediction</p> <p><strong>Reproducibility:</strong> Complete methodology, code specifications, and validation protocols provided. All prior validation studies openly available on Zenodo with DOIs.</p> <p> </p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_17254676 |
| institution | Zenodo |
| language | |
| publishDate | 2025 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | The Moral Seismograph: A Causal Machine Learning Framework for Testing Ethical Theories YANG, JIHOON <p>This paper presents a validated, reproducible pipeline for translating philosophical theories into empirically testable predictions about structural transitions in complex systems. While designed with ethics in mind, the framework has been proven viable through successful applications in linguistics (N=40, AUC≈0.944) and genomics (N=520-1,150, AUC 0.675-0.948).</p> <p><strong>Three Methodological Principles:</strong></p> <ol> <li> <p><strong>Autonomous Event Detection</strong> - Eliminates human labeling bias by detecting structural transitions through statistical threshold-crossing (Z-score > 1.5) in intrinsic dynamics, removing documentation bias and enabling universal applicability.</p> </li> <li> <p><strong>No-Leak Filtering</strong> - Systematically removes temporal contamination (39% of features in linguistics study) to preserve causal validity, ensuring future information cannot contaminate past predictions.</p> </li> <li> <p><strong>Time-Aware Cross-Validation</strong> - Enforces strict temporal ordering (train on t ≤ threshold, test on t > threshold) across multiple percentiles to test genuine predictive capacity rather than spurious correlations.</p> </li> </ol> <p><strong>Theoretical Foundation:</strong> The framework operationalizes two philosophical constructs—Recovery-Transition Dynamics and Conditional Universality—as measurable computational features (recovery_signal, transition_signal, roughness, memory). These concepts, previously validated across music, constitutions, brain activity, and chromosomes, now demonstrate cross-domain applicability in symbolic (linguistic) and biological (genomic) systems.</p> <p><strong>Empirical Validation:</strong></p> <ul> <li> <p><strong>Linguistics Study (Yang, 2025):</strong> Historical explanation AUC≈0.944 with N=40; empirically determined N>100 required for future prediction (Exp2 with N=25 showed AUC range 0.33-1.0, honest boundary-finding).</p> </li> <li> <p><strong>Genomics Study (Yang, 2025):</strong> Demonstrated conditional universality through gene-specific minimal sufficient contexts: PIK3CA requires lineage only (AUC 0.745), KRAS requires lineage+expression synergy (AUC 0.675), MYC achieves high prediction across multiple contexts (AUC >0.92).</p> </li> </ul> <p><strong>What This Paper IS:</strong></p> <ul> <li>A theory-agnostic methodology for testing ANY structural transition theory</li> <li>A complete implementation guide (Section 5) with code specifications</li> <li>Proof-of-concept validation in two disparate domains</li> <li>A reproducible framework ready for ethics application</li> </ul> <p><strong>What This Paper is NOT:</strong></p> <ul> <li>Direct validation in ethical systems (explicitly stated in Section 6.2)</li> <li>Proof of a specific ethical theory's truth</li> <li>A completed ethics study (that is Paper 2, forthcoming)</li> </ul> <p><strong>Practical Applications:</strong> Researchers can now test their own theories of ethical norm transitions, political regime changes, financial crises, or any system exhibiting structural transitions. Complete pipeline specifications provided: data requirements (N≈40 for explanation, N≈100+ for prediction), feature engineering protocols, validation procedures, and expected challenges with mitigation strategies.</p> <p><strong>Scientific Honesty:</strong> This paper exemplifies rigorous boundary-finding by honestly reporting both successes and failures, establishing when the methodology works (historical explanation) and when it requires more data (future prediction), following the principle that "honest failure is as valuable as success."</p> <p><strong>Related Work:</strong> Part of a systematic research program applying recovery-transition dynamics across domains: music (χ²=12.34, p=0.0063), constitutions, neuroscience, genomics, and linguistics. This paper synthesizes methodological insights from these validations into a unified, generalizable framework.</p> <p><strong>Next Steps:</strong> Application to ethical norm transitions in social media (Yang, forthcoming) will provide the first direct test in the target domain.</p> <p><strong>Keywords:</strong> causal machine learning, computational ethics, structural transitions, conditional universality, recovery-transition dynamics, autonomous detection, no-leak validation, cross-domain methodology, empirical philosophy, time-series prediction</p> <p><strong>Reproducibility:</strong> Complete methodology, code specifications, and validation protocols provided. All prior validation studies openly available on Zenodo with DOIs.</p> <p> </p> |
| title | The Moral Seismograph: A Causal Machine Learning Framework for Testing Ethical Theories |
| url | https://doi.org/10.5281/zenodo.17254676 |