Application of integrated gradients explainability to sociopsychological semantic markers
Ali Aghababaei, Jan Nikadon, Magdalena Formanowicz · 2025
Read Application of integrated gradients explainability to sociopsychological semantic markers by Ali Aghababaei, Jan Nikadon, Magdalena Formanowicz (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-04989.txt.
Description
Ali Aghababaei and six co-authors open the black box of a classifier that scores psychological traits in text, word by word. Sentence-level classifiers for markers like agency are popular, but users cannot see which words drive a score. Integrated gradients, an attribution method, can assign each word its contribution.
The team applies the technique to agency, one of few markers with a verified deep classifier, BERTAgent, available. They tune how the attribution model is trained, test alternatives, and check usefulness in a real application scenario. A second scenario uses the method with only a small labeled dataset, deliberately encouraging overfitting to sharpen class-distinctive words. The goal is explainable sociopsychological text analysis that researchers can audit.
Catalog record
| CREATOR | Ali Aghababaei, Jan Nikadon, Magdalena Formanowicz |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 53 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-04989