Overview
This work proposes a framework for process transparency within AI-assisted ethics research. It addresses the philosophical specification of AI disclosure mandates, arguing that existing requirements for reporting AI assistance lack sufficient explanation regarding their purpose within the context of ethical inquiry. The proposed framework grounds the duty of transparency in agent-integrity, defined as the legibility of an author's identity-constituting commitments before a community of inquiry.
Research Context
The core issue identified is that existing mandates for disclosing AI assistance in scholarship are philosophically underspecified. While they establish a duty to report AI use, they do not elaborate on the philosophical basis or purpose of this duty. The paper posits that ethical inquiry is characterized by deep disagreements, being "essentially contested at two independent levels"—concerning its nature and its demands on the inquirer. This inherent contestability, it is argued, renders output-only evaluation and a welfare-economic dismissal of transparency insufficient. Reproducibility framings, often imported from empirical sciences, are also considered inadequate due to this contested nature of ethical study.
Approach
The researchers developed a documentation-adequacy framework designed to operationalize Meaningful Human Control in AI-assisted philosophical research. This framework comprises five distinct transparency elements:
- Declaration: Explicit notification of AI assistance.
- Navigation: Guidance on how to access and understand the transparency documentation.
- Documentation Account: A narrative description of how AI was used.
- Process Documentation: Detailed records of the AI interaction process.
- Development Records: Information pertaining to the creation or adaptation of AI tools used.
The paper itself serves as a demonstration of this framework, with a full documentation record archived at a persistent identifier. The framework is presented as a first iteration, subject to revision, rather than a definitive standard.
Findings
The research suggests that the transparency duty in AI-assisted ethical inquiry is not primarily for evaluation against communally settled criteria, as such criteria are not yet established for this type of work. Instead, the achievable goal for transparency is "tracking." This 'tracking' involves accumulating an evidentiary record that enables various traditions within the community of inquiry to assess the work based on their own specific terms. Furthermore, this evidentiary record is crucial for making future normative judgments possible. The transparency framework directly addresses the problem of output-only evaluation by focusing on the process of inquiry rather than solely on its conclusions.
Why This Matters
This framework is significant because it provides a structured approach to transparency in AI-assisted philosophical research, addressing the unique challenges posed by the contested nature of ethical inquiry. By grounding transparency in agent-integrity, it aims to foster legibility of authorial commitments, which is essential for communal assessment and the evolution of normative judgments in a field where evaluative standards are not yet settled.