analiza aiqbench
This research identifies how large language models reinforce standard language ideology, systematically granting legitimacy to certain dialects while marginalizing others through mechanisms like stereotyping and erasure. The authors propose a sociotechnical taxonomy to map these biases and advocate for accountability frameworks that prioritize community agency and linguistic diversity. For a tech audience, the key takeaway is the urgent need to address power relations in model training to prevent the automation of linguistic inequality.
Standard Language Ideology in AI-Generated Language
arXiv:2406.08726v3 Announce Type: replace Abstract: Large language models (LLMs) generate text that reinforces standard language ideology: a bias towards certain language varieties that are granted more prestige, authority, and legitimacy than others. This paper contributes a sociotechnically grounded faceted taxonomy that illustrates how generative AI systems reproduce standard language ideology and its societal implications. We introduce the concept of standard AI-generated language ideology to explain how AI systems confer legitimacy on certain language varieties while marginalizing others, structuring patterns of performance disparity, stereotyping, appropriation, and erasure. We then discuss ongoing tensions around what constitutes desirable system behavior, as well as advantages and drawbacks of generative AI tools attempting or refusing to imitate different language varieties. To address the power relations shaping generative AI and the mechanisms identified in our taxonomy--legitimation, stereotyping, appropriation, and erasure--we offer recommendations that emphasize accountability, community agency, control, and ownership. These recommendations recognize linguistic diversity as a resource to be protected, valued, and sustained as part of a just AI future.
Abstract:Large language models (LLMs) generate text that reinforces standard language ideology: a bias towards certain language varieties that are granted more prestige, authority, and legitimacy than others. This paper contributes a sociotechnically grounded faceted taxonomy that illustrates how generative AI systems reproduce standard language ideology and its societal implications. We introduce the concept of standard AI-generated language ideology to explain how AI systems confer legitimacy on certain language varieties while marginalizing others, structuring patterns of performance disparity, stereotyping, appropriation, and erasure. We then discuss ongoing tensions around what constitutes desirable system behavior, as well as advantages and drawbacks of generative AI tools attempting or refusing to imitate different language varieties. To address the power relations shaping generative AI and the mechanisms identified in our taxonomy--legitimation, stereotyping, appropriation, and erasure--we offer recommendations that emphasize accountability, community agency, control, and ownership. These recommendations recognize linguistic diversity as a resource to be protected, valued, and sustained as part of a just AI future.
Computation and Language (cs.CL)
arXiv:2406.08726 [cs.CL]
(or arXiv:2406.08726v3 [cs.CL] for this version)
Text integral sub licența deschisă a publicației · 172 cuvinte