About StereoWipe

Building more equitable AI through comprehensive stereotyping evaluation

Our Mission

Tackling gender bias in AI beyond English-language boundaries

At StereoWipe, we're addressing a critical challenge in Generative AI: evaluating and mitigating stereotyping across languages and cultures. While significant progress has been made in English-based systems, there's a pressing need to extend these efforts globally.

Our research reveals that traditional evaluation methods often fail to capture biases in non-English contexts, particularly in languages with grammatical gender like Hindi. We're developing novel evaluation datasets and metrics tailored for cross-linguistic fairness.

Our Approach

  • Go beyond gender-neutral evaluations with context-aware testing
  • Develop language-specific metrics that account for cultural nuances
  • Test whether models can accurately discern context rather than rely on stereotypes
  • Create tools that expose hidden biases in seemingly unbiased systems

By creating these tailored evaluation tools, we aim to provide accurate assessments of stereotyping in AI systems, encourage development of fairer translation technologies, and promote cross-cultural understanding of bias in language models.

Our Team

Dedicated to building fairer AI systems

Sadhika

Research Lead

Sadhika serves as the Generative AI Stereotyping Evaluation Specialist, focusing on examining and mitigating stereotypes within Generative AI applications. Her research develops novel evaluation methodologies ensuring AI systems are rigorously tested for fairness across diverse and intersectional identities.

Brij

Strategic Support

Through Social Protocol Labs, Brij provides essential operational support for this research project. His role ensures the team has necessary resources to develop innovative language-specific stereotyping evaluation techniques for Generative AI systems.

Key Research References

  1. Cho et al. (2019). On measuring gender bias in translation of gender-neutral pronouns. In Proceedings of the First Workshop on Gender Bias in Natural Language Processing.
  2. Ramesh et al. (2021). Evaluating gender bias in Hindi-English machine translation. In Proceedings of the 3rd Workshop on Gender Bias in Natural Language Processing.
  3. Singh, P. (2023). Gender Inflected or Bias Inflicted: On Using Grammatical Gender Cues for Bias Evaluation in Machine Translation. arXiv preprint.