Karina Halevy

CMU LTI | Harvard CS + History

headshot_2022.jpg

Image Description: Karina (white skin, brown hair, black top) smiling at the camera outside. Picture credits: Oriana Li Halevy, 2022.

Hello! My name is Karina, and I am a PhD student at the Carnegie Mellon University Language Technologies Institute, supported by a National Science Foundation Graduate Research Fellowship and co-advised by Mona Diab and Maarten Sap! Previously, I did my undergrad in Computer Science and History (jointly) with a minor in Statistics at Harvard College, where I worked with Professors Stuart Shieber, Elena Glassman, Cemal Kafadar, and Damián Blasi. I have also spent summers working with Shaghayegh Agah (Comcast), Carolina Brum (Apple), Antoine Bosselut, Syrielle Montariol (EPFL), and Jordan Hashemi (BBN Technologies). My main academic interest is in natural language processing/computational linguistics.

I am specifically interested in:

  • carefully doing NLP for social science/social good (e.g. information extraction or visualization for legal/medical/historical material, hate speech/misinformation detection, public health advocacy)
  • NLP for low-resource/non-Indo-European languages
  • algorithmic fairness/ethics (especially for disabled people, ethnically minoritized people, & subpopulations too small to be included in most research due to lack of statistical significance)
  • constructing evaluation metrics and benchmark datasets to make language models more understandable
  • intersections of NLP with human-computer interaction and data visualization

I also dance.

Selected Publications

  1. methods.png
    Who’s the (Multi-)Fairest of Them All: Rethinking Interpolation-Based Data Augmentation Through the Lens of Multicalibration
    Karina Halevy, Karly Hou, and Charumathi Badrinath
    Proceedings of the AAAI Conference on Artificial Intelligence, Apr 2025
  2. Karina_v1-1.png
    “Flex Tape Can’t Fix That”: Bias and Misinformation in Edited Language Models
    Karina H Halevy, Anna Sotnikova, Badr AlKhamissi, and 2 more authors
    In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, Nov 2024
  3. preview_p2.png
    Evaluating Word Embeddings with Categorical Modularity
    Sı́lvia Casacuberta, Karina Halevy, and Damián Blasi
    In Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, Aug 2021