Bad apples just for friends: A large language model approach to research on the strategic framing in government’s human rights advocacy narratives

Abstract

Existing scholarship shows that transnational human rights advocacy depends on successful framing and audience mobilization. While most of this literature has focused on human rights framing by NGOs and advocates, governments can also frame human rights interests to their advantage. Focusing on the topic of police violence in the United States government’s transnational human rights advocacy practices, this article argues that a government may frame advocacy narratives for its benefit. For the government, strategic framing could be used to mobilize pressure against its geopolitical rivals, ease condemnation against its friends, and actively define advocated issues in a way favorable to its regime. With a novel network approach for text representation based on pre-trained large language models (LLMs), this article proposes an effective method to measure strategic framing from text data. Using the US State Department’s human rights reports, the results show that police violence accusations in the US government’s human rights advocacy narratives are strategically framed with reporting in favor of countries closer to the US. This research contributes to human rights scholarship by highlighting how governments’ national interests considerations could be incorporated into transnational human rights advocacy activities through strategic framing. The proposed LLM-based text data representation method also shows promising potential for broader text analysis tasks like topic modeling.