Euphemism in AI Discourse as Colonial Harm & Effect: Learnings from Subaltern Puerto Rico

Through three integrated studies, this dissertation explores complex human-AI interactions by comparing the perceptions of two groups of U.S. citizens: subaltern Puerto Ricans and stateside Americans’ interactions with ChatGPT. This corpus-assisted Critical Discourse Analysis or CLCDA, triangulated with 1:1 semi-structured interviews, validates scholars’ concerns about the biases inherent in generative AI. Contributing to those concerns, this dissertation reveals that ChatGPT deploys discourse strategies that, when executed, facilitate digital colonialism. Euphemistic language is prominent among its discourse strategies, replacing controversial words, phrases, and topics with more palatable ones. When prompted on events of historical significance, I find that ChatGPT’s discourse obscures colonialism and its effects in its narratives – causing further harm to already disempowered subaltern Puerto Ricans. This work presents euphemistic discourse as a ChatGPT phenomenon and discusses its broad implications for diverse groups worldwide.