From GPT to Human Critique: A Look at Our Model
July 20, 2025
Our InterXAI pipeline combines three layers of interpretation:
- Model Explanation (e.g., SHAP, LIME, attention weights)
- Human Commentary (scholarly or critical notes)
- Intertextual Overlay (matching or clashing sources)
In this post, we provide an overview of our architecture—from GPT-based content generation to side-by-side comparison with curated human annotations.
We believe this hybrid approach opens up a new form of critical XAI that’s especially useful in education, ethics, and digital humanities research.
🧩🧑🏫💡