PUBLICATIONS

In Preparation:

Constructing Normative Evaluators: Subjection Techniques of Representational Harms

Nasanbayar Ulzii-Orshikh, Mark Ackerman, Justine Zhang

For a Submission to FAccT, 2026

Generative Cultural Erasures in GenAI-Mediated Articulation

Nasanbayar Ulzii-Orshikh, Mark Ackerman, Justine Zhang
For a Submission to FAccT, 2026

In Publication:

The Subject(s) of Representational Harm

Nasanbayar Ulzii-Orshikh, Mark Ackerman, Justine Zhang
In Synthetic Data: Representation and/vs Representativeness workshop, Aarhus Conference 2025

Quality at a Glance: An Audit of Web-Crawled Multilingual Datasets
Isaac Caswell, Julia Kreutzer, Lisa Wang, Ahsan Wahab, Daan van Esch, Nasanbayar Ulzii-Orshikh, et al.
In Transactions of the Association for Computational Linguistics (TACL), 2022
Previously, in AfricaNLP Workshop, EACL, 2021

Clustering via Information Access in a Network
Hannah Beilinson*, Nasanbayar Ulzii-Orshikh*, Ashkan Bashardoust, Sorelle Friedler, Carlos Scheidegger and Suresh Venkatasubramanian
In arXiv:2010.12611, 2022
[Github]

Iteration with Intention: Project-Based Learning of Computational Thinking
Nasanbayar Ulzii-Orshikh and John Dougherty
In Proceedings of the 51st ACM Technical Symposium on Computer Science Education (SIGCSE), 2020
[In Proceedings]     [Poster]

Maladaptive Coping Predicts Fixed Mindset in Asian Students
Jenny Yu, Nasanbayar Ulzii-Orshikh, Meredith A. Henry, Louise K. Charkoudian, Benjamin Le, Lisa A. Corwin, and Jennifer M. Heemstra
In Abstracts of the Southeastern Teaching of Psychology Conference (SETOP), 2020

Multispectral imaging reveals the design of iridescent visual signals in peacocks & related pheasants
Kane, S.A., Xia, S., Fang, R., Lu, Y., Ulzii-Orshikh, N., Wu, J., Dakin, R.
In Abstracts of the Society for Integrative and Comparative Biology Conference (Oxford University Press), 2020

SELECTED PROJECTS

The 114th Congressional Co-sponsorship Network: Principal Component Analysis (PCA)
Explored three representations of the 114th Congressional co-sponsorship network — information access signatures, adjacency matrix, and representation with node2vec — using Principal Component Analysis (PCA) to find and experimentally validate a computationally less expensive pipeline of information access clustering (Beilinson et al. 2020).

Algorithmic Guarantee of Equity: Solving the Registrar Problem
Designed a new algorithm that guarantees a structurally equitable allocation of resources based on a feasibility of circulation problem, solved using the Max-flow min-cut theorem, and provided its qualitative analysis against the baseline approach (p. 10-13).

Faster K-Nearest Neighbors with Locality-Preserving Hashing
Designed a new Locality-Preserving Hashing algorithm for KNN, achieving (analyzed a posteriori) a minimum accuracy of 87% within 1/10th of the time by the classic algorithm (Hal Daumé III (2017). A Course in Machine Learning, 33.).

Iris Form
Implemented from scratch a Convolutional Neural Network on a personally collected human iris dataset and experimentally confirmed the consistency in the model's knowledge system.

SELECTED WRITING

Understanding Cognitive Affordances
Tracing the historical progression of how logicians have conceptualized the way that the mind comes to grips with the world: Aristotle's terms, Kant's intuition and concept, Wittgenstein's object name and pictorial form, and Frege's sense (Sinn) and designation (Bedeutung).

On Kant’s Abstracting of Rules
Exploring how the Kantian logical framework, in contrast to the Aristotelian one, explains judgments through analogies.

So What Exactly is a ‘Tree’ Again?
Attempting to understand how concepts mediate the world and the mind with the help of John McDowell, Charles Travis, and Danielle Macbeth.

Recollecting Errors
A meditation on Plato's argument on the innateness of knowledge.

Knowledge: Its Origin & Consequences
A comparative analysis of "knowledge" across Aristotle, Descartes, Hume, and Kant.

SELECTED INDUSTRY EXPERIENCE

Product Manager Intern, Tomujin Digital
Researched the commercial potential of YODA, a $2,000,000 in valuation digital counseling platform for low-income college applicants, and launched its early release, achieving a 93% conversion rate for students and 79% for counselors.

Chief Technology Officer, One Minute
Co-founded an ed-tech social enterprise, valued at $50,000 by Shark Tank, to change the way rural Mongolian students plan their education. Currently at 25,000 followers on Facebook.