A Follow-up Work to AISA–T: Elucidating the Human Evaluation Rubric for AGI Self-Awareness
Year: 2025
Type: article
Abstract: The original AISA–T paper established a critical framework for assessing AI self-awarenessthrough a three-way comparison of self-ratings, intrinsic quality, and independent human-assigned scores. However, the specific rubric used for the human evaluation component wasnot detailed. This follow-up work proposes a comprehensive rubric designed to standardize thehuman evaluation process for AGI-level responses. Our proposed rubric evaluates responseson a five-point scale across three core dimensions: Depth, Coherence, and ArchitecturalAccuracy. By formalizing this methodology, we aim to enhance the reproducibility andtransparency of future AGI self-awareness studies, providing a clearer path for validating thenascent capabilities of highly advanced AI models.
Author: de ceuster, peter
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W6911963573
FWCI:
Cites:
Cited by:
Related to: