Linguistic Bias and Cultural Diversity in the Evaluation of Language Models

This editorial reflects on how the biases of language models can particularly affect those of us who use AI in Spanish. Drawing on the findings of SESGO and the concept of Digital Linguistic Bias, it examines how the unequal representation of Spanish varieties and the lack of cultural context can lead these systems to reproduce, and even reinforce, existing inequalities. In light of this, it highlights the need to develop an AI that not only speaks Spanish, but is also able to represent the diversity of those who speak it.