Gender Bias in Coreference Resolution
This addresses fairness issues in NLP for users affected by biased AI systems, though it is incremental as it builds on existing bias detection methods.
The study tackled gender bias in coreference resolution by introducing Winogender schemas, a novel set of minimal pair sentences, and found systematic bias in three systems, correlating it with real-world and textual gender statistics.
We present an empirical study of gender bias in coreference resolution systems. We first introduce a novel, Winograd schema-style set of minimal pair sentences that differ only by pronoun gender. With these "Winogender schemas," we evaluate and confirm systematic gender bias in three publicly-available coreference resolution systems, and correlate this bias with real-world and textual gender statistics.