Paper Number
ECIS2026-1565
Paper Type
CRP
Abstract
Generative AI tools promise to transform literature review practices, yet concerns about transparency and methodological rigor persist. This paper develops the AI-Augmented Hermeneutic Review Framework (AI-HRF), integrating mechanistic interpretability with Gadamerian dialogical hermeneutics and Neohumanist values. Through grounded theory analysis of an Anthropic interpretability transcript, we demonstrate how researchers can navigate tensions between AI efficiency and interpretive depth. The framework offers three accessibility tiers enabling rigorous AI-augmented reviews across varying resource contexts(1) essential practices (prompt logging, output verification), (2) enhanced techniques (hermeneutic validation, meaning checking in outputs with awareness of which features or representations the AI is activating), and (3) specialized methods (mechanistic interpretability). Our contribution lies in bridging technical AI transparency with interpretive research traditions, providing IS researchers with practical guidance for responsible integration of Generative AI while maintaining scholarly rigor.
Recommended Citation
Mashele, Billy BH, "From Opaque Black Boxes To Dialogues: Integrating Mechanistic Interpretability With Human Interpretation In Hermeneutic Literature Reviews" (2026). ECIS 2026 Proceedings. 3.
https://aisel.aisnet.org/ecis2026/litrev/litrev/3
From Opaque Black Boxes To Dialogues: Integrating Mechanistic Interpretability With Human Interpretation In Hermeneutic Literature Reviews
Generative AI tools promise to transform literature review practices, yet concerns about transparency and methodological rigor persist. This paper develops the AI-Augmented Hermeneutic Review Framework (AI-HRF), integrating mechanistic interpretability with Gadamerian dialogical hermeneutics and Neohumanist values. Through grounded theory analysis of an Anthropic interpretability transcript, we demonstrate how researchers can navigate tensions between AI efficiency and interpretive depth. The framework offers three accessibility tiers enabling rigorous AI-augmented reviews across varying resource contexts(1) essential practices (prompt logging, output verification), (2) enhanced techniques (hermeneutic validation, meaning checking in outputs with awareness of which features or representations the AI is activating), and (3) specialized methods (mechanistic interpretability). Our contribution lies in bridging technical AI transparency with interpretive research traditions, providing IS researchers with practical guidance for responsible integration of Generative AI while maintaining scholarly rigor.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.