Ethics in NLP: Multilingual and Multicultural Aspects

Module Description

Participants

All advanced CL Bachelor students and all CL master students. Students from MSc Data and Computer Science or MSc Scientific Computing with Field of Application Computational Linguistics are welcome after getting permission from the lecturer. Solid linguistic background is necessary for some of the topics (such as multilingual topics).
MSc Scientific Computing students can only take the course as HS for 8 LP. If the seminar should be oversubscribed, CL students will have priority.

If you took another ethics seminar already, please discuss with me whether another participation is possible. If you took my "Linguistics in Modern NLP" seminar in SS25 participation is possible as overlap is small (2-3 papers) but you need to present a paper that is not the same as the one you presented in the previous seminar.

Prerequisites for Participation

The types of papers are very wide ranging from basic to advanced therefore a wide range of students can take part. The course is particularly suitable for students with a good linguistic knowledge or a secondary subject in the humanities.

• Module Prerequisites
• For Bachelor Students: ECL, Introduction to Linguistics, Prog I
• For MA Students: none
• For 25% Students: Programming knowledge corresponding to programming I as well as ECL and solid linguistic knowledge is required. FF-FM (Grundlagen der Mathematik) is not required.

Assessment

• Presentation
• Discussion contributions: literature critiques, coassessment for presentations
• Project or oral exam dependent on the number of participants. The Camilla and Georg Jellinek Center for Ethics at the university is planning to conduct a hackathon in the winter semester, for which I will likely be a topic giver. Dependent on time and topic, participation in this hackathon can either give extra credit, ÜK points for the Bachelor students or even replace the project for some students. However, the hackathon is in its early planning stages so I cannot yet fully guarantee that this is feasible.

Content

Given the large-scale development and deployment of NLP methods, and in particular but not only LLMs, a wide variety of ethical questions for NLP have arisen, including privacy and copyright issues, the representation of and impact on minorities, environmental impact of training and using LLMs, working conditions and exploitation for human annotators for LLms, and the power and political structure that can be solidified or upended.

As this can cover several seminars, we will concentrate mostly on multilingual and multicultural issues in this seminar. This includes:

• Representation bias against minorities and groups with protected characteristics in embeddings, LLMs and other NLP systems (translation, summarization etc). Methods to test for and mitigate such bias. An example paper is https://www.nature.com/articles/s41586-024-07856-5
• Disadvantaging minority languages, low-resource languages and dialects. This can happen in several different ways, including (i) higher expenses for minority language LLM users, often due to tokenization problems (ii) worse performance of NLP models on lower-resource languages and (iii) negative associations and biases against some dialects, such as African American Vernacular or German dialect speakers. An example paper of the latter is https://www.nature.com/articles/s41586-024-07856-5 or https://aclanthology.org/2025.emnlp-main.415.pdf
• NLP and culture, in particular whether NLP should be multicultural and culture-aware and what this means for NLP algorithm development and NLP alignment methods. We are looking at impact of culture on human preference and human value collection as well as benchmark performance and benchmark construction . Example papers are https://arxiv.org/pdf/2404.16019 or https://aclanthology.org/2024.acl-long.345/ or https://aclanthology.org/2025.acl-long.919.pdf

Literature

Will be given out at beginning of class.

» More Materials