Ongoing developments in natural language processing (NLP) and natural language generation (NLG) raise critical questions about linguistic bias and inequalities in datasets. Problematizing the concept of digital “inclusion” for linguistic vitality, this article addresses an adjacent research gap concerning dilemmas that are emerging from the increasing operability of globally marginalized languages with NLP and NLG technologies. We draw on our previous work on how Google Search “Autocomplete” algorithms interact with three languages indigenous to East Africa – Amharic, Kiswahili and Somali – each with their own historical, political and orthographic features that condition this automated interaction in complex and unpredictable ways. Highlighting different forms of harm that might result from the operation of predictive text in each language, we argue for the necessity of situating the algorithmic experiences of marginalized languages within specific historical, cultural and political contexts. From here, we consider the ways that the predictive logics of autocomplete have been expanding through the rapid intrusion of generative AI into the results of mainstream search engines. We grapple with the global implications of this rapidly changing AI information landscape and consider the potential impacts of such developments on linguistic contexts in East Africa and the dynamics of multi-scalar language inequality within and beyond the region.
Chonka et al. (Thu,) studied this question.