A new position paper proposes a unified agenda for Speech AI, highlighting the disconnect between technical NLP and sociotechnical HCI communities. The paper argues that current Speech AI systems have an incomplete understanding of communication and identity, and their metrics fail to capture the right constructs. It suggests solutions focused on designing for human variability, integrating both quantitative and qualitative research methods, and reforming venue structures to foster cross-community collaboration. AI
IMPACT Could lead to more inclusive and effective speech AI systems by bridging technical and social science approaches.
RANK_REASON The item is a position paper published on arXiv proposing a research agenda. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →