Exploring the Llama 2 Large Language Model Through Critical AI Ethics

• In Event: Social Justice and Liberation for Marginalized Communities

Abstract

This paper aims to contest the adoption of large language models in education. Specifically, I employ a critical discourse analysis of Meta’s Llama 2 large language model through a critical AI ethics (Waelen, 2022) perspective to analyze the Llama 2 paper (Touvron et al., 2023) in relation to two ethical principles that underpin the public school ideal outlined by Winton (2022). The ethical principles used are fairness and non-discrimination, and sustainability (UNESCO, 2022). The analysis demonstrates an incommensurability between how the Llama 2 model is trained and the public school ideal. Furthermore, the paper demonstrates to educational stakeholders considering the adoption of alternative large language models how a critical AI ethics approach can inform their thinking.

Author

• Kevin Oliver, University of Manitoba