走向社会科学中的适定问题：Hadamard准则作为认识论护栏

中文摘要 AI 辅助

我们开发了一个跨学科框架，通过Hadamard的适定问题准则（存在性、唯一性和稳定性）来评估社会科学探究的认识论稳健性。我们将这些准则重新解释为方法论护栏，而非对确定性必然性的要求，并展示它们如何诊断定量和定性范式中反复出现的识别与推断失败。在定量研究（如计量经济学建模）中，当实质性结论对模型设定或数据过滤敏感依赖时，不稳定性便显现出来。在解释主义定性研究中，当理论框架具有足够弹性以容纳相互矛盾的观察结果而无需预设的拒绝标准时，非唯一性和不可证伪性便会出现。我们将这些失败框定为逆问题，即从经验数据到实质性主张的映射未能满足存在性、唯一性或稳定性。最后，我们提出跨范式保障措施：事前证伪标准、经验边界条件、多重宇宙敏感性审计和跨观察者验证，以确保社会科学主张保持适当约束、可证伪，并对证据和解释的扰动具有稳健性。

英文摘要

We develop an interdisciplinary framework for evaluating the epistemic robustness of social-scientific inquiry through Hadamard's criteria for well-posed problems: existence, uniqueness, and stability. Reinterpreting these criteria not as demands for deterministic certainty but as methodological guardrails, we show how they diagnose recurrent failures of identification and inference across quantitative and qualitative paradigms. In quantitative research (e.g., econometric modeling), instability manifests when substantive conclusions depend sensitively on model specifications or data filtering. In interpretivist qualitative research, non-uniqueness and non-falsifiability arise when theoretical frameworks are elastic enough to accommodate contradictory observations without pre-specified rejection criteria. We frame these failures as inverse problems where the mapping from empirical data to substantive claims fails to satisfy existence, uniqueness, or stability. Finally, we propose cross-paradigmatic safeguards: ex-ante falsification criteria, empirical boundary conditions, multiverse sensitivity auditing, and cross-observer validation, to ensure social-scientific claims remain appropriately constrained, falsifiable, and robust to perturbations in evidence and interpretation.

发表机构

• Portland State University（波特兰州立大学）

机构由 AI 辅助整理，请以论文原文为准。