Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Epistemological Bias As a Means for the Automated Detection of Injustices in Text

Kenya Andrews, Lamogha Chiazor

TL;DR

Epistemic biases in text often conceal subtle injustices that are hard to detect manually. The paper presents a modular, explainable pipeline that fuses epistemology-inspired tagging with transformer-based bias detection and stereotype reasoning (Tagger, CO-STAR, SBF), augmented by lexicon lookup and an editor-facing UI. Validation includes a human survey and a scalable comparative test across Meghan Markle and Kate Middleton headlines, demonstrating reasonable alignment with human judgments and revealing systematic framing patterns. The work offers a practical tool to reduce time burdens in identifying implicit injustices in media content while preserving explainability and editorial accountability.

Abstract

Injustices in text are often subtle since implicit biases or stereotypes frequently operate unconsciously due to the pervasive nature of prejudice in society. This makes automated detection of injustices more challenging which leads to them being often overlooked. We introduce a novel framework that combines knowledge from epistemology to enhance the detection of implicit injustices in text using NLP models to address these complexities and offer explainability. Our empirical study shows how our framework can be applied to effectively detect these injustices. We validate our framework using a human baseline study which mostly agrees with the choice of implicit bias, stereotype, and sentiment. The main feedback from the study was the extended time required to analyze, digest, and decide on each component of our framework. This highlights the importance of our automated framework pipeline that assists users in detecting implicit injustices while offering explainability and reducing time burdens on humans.

Epistemological Bias As a Means for the Automated Detection of Injustices in Text

TL;DR

Epistemic biases in text often conceal subtle injustices that are hard to detect manually. The paper presents a modular, explainable pipeline that fuses epistemology-inspired tagging with transformer-based bias detection and stereotype reasoning (Tagger, CO-STAR, SBF), augmented by lexicon lookup and an editor-facing UI. Validation includes a human survey and a scalable comparative test across Meghan Markle and Kate Middleton headlines, demonstrating reasonable alignment with human judgments and revealing systematic framing patterns. The work offers a practical tool to reduce time burdens in identifying implicit injustices in media content while preserving explainability and editorial accountability.

Abstract

Injustices in text are often subtle since implicit biases or stereotypes frequently operate unconsciously due to the pervasive nature of prejudice in society. This makes automated detection of injustices more challenging which leads to them being often overlooked. We introduce a novel framework that combines knowledge from epistemology to enhance the detection of implicit injustices in text using NLP models to address these complexities and offer explainability. Our empirical study shows how our framework can be applied to effectively detect these injustices. We validate our framework using a human baseline study which mostly agrees with the choice of implicit bias, stereotype, and sentiment. The main feedback from the study was the extended time required to analyze, digest, and decide on each component of our framework. This highlights the importance of our automated framework pipeline that assists users in detecting implicit injustices while offering explainability and reducing time burdens on humans.
Paper Structure (16 sections, 2 equations, 9 figures, 8 tables)

This paper contains 16 sections, 2 equations, 9 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Methodology
  4. Tagger model
  5. CO-STAR Model and Social Bias Frames
  6. Lexicon Lookup
  7. Interactive Interface for Learning about Bias
  8. Datasets
  9. Validation Survey
  10. Results and Discussions
  11. Tagger Model Ablation Study
  12. Comparative Test
  13. Scaling the comparative test
  14. Human Survey Discussions
  15. Conclusions
  16. ...and 1 more sections

Figures (9)

  • Figure 1: Testimonial Injustice Technical Framework
  • Figure 2: Process of Getting Data to the Framework
  • Figure 3: The interactive interface showing the highlighted tagged word and associated information for a sentence.
  • Figure 4: UI showing the tagged words, their associated probability for bias, the potential stereotype and stereotype concept and some explainable data created from the lexicon search.
  • Figure 5: Training and Validation Results for the Tagger Model using 10 epochs. no features = the expert features as mentioned in Section \ref{['ssec:methodtagger']} was not added, with expert features = combine the expert features as mentioned in Section \ref{['ssec:methodtagger']} and with expert + tmi features = combine the expert features whilst using the data samples containing TMI class no tmi or tmi
  • ...and 4 more figures