[New Course in SP27] LING 350: Are Algorithms Biased?

Flyer of a new course

LING 350 Are Algorithms Biased?

Instructor: Dr. Gyu-Ho Shin

This course investigates the intersection of linguistics and Artificial Intelligence (AI) to uncover how AI systems inherit, amplify, and encode human biases. Through readings and hands-on Python activities, students will critically evaluate real-world AI models across four core modules: measuring the multilingual tax, diagnosing structural failures in morphosyntactically diverse languages, probing representations and model outputs for semantic and pragmatic harms, and quantifying dialectal perplexity gaps across (marginalised) speech varieties. By the end of this course, you will be able to empirically audit language technologies and design more inclusive, linguistically grounded AI systems.

  • Prerequisite: LING 150 (or consent of the instructor).
  • Recommended background: LING 370, LING 310, and/or credit in at least one 200-level LING course; some Python knowledge and experience (if you are unsure about this, check with me first).

[Course schedule (subject to change)]

  • The Linguistic Architecture of AI Bias
  • Typology & Cross-Lingual Disparities
  • Morphosyntactic Breakdown
  • Semantic & Pragmatic Harms
  • Sociolinguistics Gaps