# Systemic Biases in Sign Language AI Research: A Deaf-Led Call to Reevaluate Research Agendas

**Authors:** Aashaka Desai, Maartje De Meulder, Julie A. Hochgesang, Annemarie Kocab, Alex X. Lu  
**arXiv:** [2403.02563](https://arxiv.org/abs/2403.02563)  
**Date:** 2024-03-05  
**Summary page:** https://www.summarizepaper.com/en/paper/2403.02563v1/

## Key points

- Sign language AI research is biased and overlooks linguistic and cultural aspects of sign languages
- Assumption of direct connection between sign language use and hearing ability neglects diverse experiences within the Deaf community
- Focus on continuous sign language datasets with little consideration for isolated sign language or fingerspelling data
- Lack of linguistic grounding in annotations and development methods leads to flawed models
- Importance of intentional inclusion of Deaf stakeholders in shaping research direction emphasized
- Need for greater diversity and representation in sign language AI research highlighted
- Call for more collaboration with Deaf researchers to lead the conversation and prioritize ethical development practices

## Summary

A systematic literature review of 101 recent papers in sign language AI revealed significant biases in the current state of research. The majority of papers were motivated by addressing perceived communication barriers for Deaf individuals, but overlooked the rich linguistic and cultural aspects of sign languages. Many assumed a direct connection between sign language use and hearing ability, failing to acknowledge the diverse experiences within the Deaf community. Additionally, most papers focused on continuous sign language datasets with little consideration for isolated sign language or fingerspelling data. While some utilized publicly available datasets, others collected their own private ones. However, there was a lack of linguistic grounding in annotations and development methods that built upon flawed models. Inspired by critical literature reviews in other fields, researchers emphasized the importance of intentional inclusion of Deaf stakeholders in shaping the direction of sign language AI research. Without meaningful input from those directly impacted by these technologies, the field risks being driven solely by convenience or what is deemed important by hearing researchers. This study highlights the need for greater diversity and representation in sign language AI research and calls for more collaboration with Deaf researchers to lead the conversation and prioritize ethical development practices.