About

Jordan Kodner is an assistant professor in the Department of Linguistics at Stony Brook University, with affiliations to the Institute for Advanced Computational Science, the Department of Computer Science, the Institute for AI-Driven Discovery and Innovation, and the Natural Language Processing (NLP) group. His primary research focuses on computational approaches to child language acquisition and their broader implications. Specifically, he investigates algorithmic models of grammar acquisition, especially morphology, how these processes drive language variation and change, the insights they provide for low-resource NLP, and what they reveal about the intersection of NLP and cognitive science. Kodner is currently authoring a book titled "Child Language Acquisition in the Past: A Mechanistic View of Language Change" for Edinburgh University Press, which explores the role of child language acquisition in language change by integrating algorithmic modeling, corpus methods, historical linguistics, cognitive science, and variationist sociolinguistics. Kodner's academic background includes a PhD in Linguistics from the University of Pennsylvania, completed in 2020 under the supervision of Charles Yang and Mitch Marcus, and a master's degree from the University of Pennsylvania Department of Computer and Information Science obtained in 2018. Prior to his academic appointments, he worked as an Associate Scientist at Raytheon BBN Technologies from 2013 to 2015 on defense and medical-related projects and interned with Amazon Alexa AI-Natural Language Understanding in 2020. His broader academic interests encompass a wide range of topics including Chinese language varieties, computing history, evolutionary theory, formal language theory, general NLP, human geography, Indo-European historical linguistics, Latin language, paleontology and cladistics, programming languages and software engineering, Roman history and culture, Semitic languages, Shona language, Singlish and Singaporean English, and writing systems.

Research topics

• Artificial Intelligence
• Computer Science
• Natural Language Processing
• Linguistics
• Psychology
• Mathematics
• Humanities
• Cognitive psychology
• Statistics
• Philosophy
• Epistemology
• Geography
• History
• Art
• Physics

Selected publications

• Artificial Knowledge of Language

  Vernon Press eBooks · 2026-01-01

  book

  The central question that this volume seeks to answer is: What are the similarities and differences between how human beings know language and how artificial intelligence knows language? The recent development and popularization of artificial intelligence systems called Large Language Models (such as ChatGPT) have led to a proliferation of opinions regarding the relevance of these systems beyond the practical purposes for which they were designed. It is not uncommon to find statements in social networks and popular magazines, as well as in academic publications, to the effect that these language models have solved the problems that sciences such as linguistics aim to solve, that their success in generating text can be seen as a refutation of some particularly influential theories of language, or that Language Models are actually scientific theories of language. These statements seem to be based on the premise that the linguistic knowledge acquired by these systems is comparable to that developed by humans. This book aims to evaluate whether this assumption is warranted. To this end, the opinions of renowned linguists and other cognitive scientists have been gathered to answer questions such as what kind of language knowledge these artificial systems have, to what extent they are faithful models of natural language knowledge, and what we can learn about the human language faculty by examining their inner workings. Anyone interested in the nature of human language and mind and in artificial intelligence can follow the eight chapters of the book without being an expert in linguistics or computer science. This is the first comprehensive work to present the views of experts in linguistic theory on the relevant questions mentioned above, and to provide an accessible presentation of current research on the nature of artificial knowledge of language.

  PublisherDOI

• Syllable Structures Across Arabic Varieties

  2026-01-01

  articleOpen access

  This study compares the syllable structures of nine Arabic varieties from Wiktionary, using a computational syllabifier.It further investigates methods for learning syllable boundaries in unsyllabified words transcribed in the International Phonetic Alphabet (IPA).The syllabification algorithm is evaluated under three conditions: (i) Default, employing fixed rules; (ii) Joint, learning onsets and codas across all varieties collectively; and (iii) Per-variety, learning onsets and codas specific to each variety.Results indicate that the default configuration yields the highest accuracy, ranging from 97.05% to 100%.The per-variety approach achieves 90.64% to 100% accuracy, while the joint approach ranges from 84.63% to 94.74%.A cross-variety analysis using Jensen-Shannon divergence reveals three principal groupings: Egyptian, Hejazi, and Modern Standard Arabic are closely related; Levantine and Gulf varieties constitute a second cluster; and Juba Arabic, Maltese, and Moroccan emerge as outliers.A cleaned dataset encompassing all nine varieties is also provided.

  PublisherOA PDFDOI

• Analogical Extension from Irregular Paradigms in Iranian Armenian: A Case of "Elsewhere Reversal"

  OSF Preprints (OSF Preprints) · 2026-03-03

  otherSenior author

  Supplementary materials for Dolatian & Kodner 2026. The materials include a readme that explains all the raw data and calculations. The calculations are based on a frequency dictionary of Eastern Armenian, and on the Universal Dependencies corpus of Eastern Armenian.

  Publisher

• Exploring Limitations of LLM Capabilities with Multi-Problem Evaluation

  2025-01-01 · 3 citations

  articleOpen access

  We propose using prompts made up of multiple problems to evaluate LLM capabilities, an approach we call multi-problem evaluation.We examine 7 LLMs on 4 related task types constructed from 6 existing classification benchmarks.We find that while LLMs can generally perform multiple homogeneous classifications at once (Batch Classification) as well as when they do so separately, they perform significantly worse on two selection tasks that are conceptually equivalent to Batch Classification and involve selecting indices of text falling into each class label, either independently or altogether.We show that such a significant performance drop is due to LLMs' inability to adequately combine index selection with text classification.Such a drop is surprisingly observed across all LLMs attested, under zero-shot, few-shot, and CoT settings, and even with a novel synthetic dataset, potentially reflecting an inherent capability limitation with modern LLMs.

  PublisherOA PDFDOI

• “Generative Perspectives of Language Change and Acquisition” in Diachrony and Language Evolution

  Elsevier eBooks · 2025-01-01

  book-chapter1st authorCorresponding
  PublisherDOI

• LLMs can Perform Multi-Dimensional Analytic Writing Assessments: A Case Study of L2 Graduate-Level Academic English Writing

  2025-01-01 · 1 citations

  articleOpen access

  The paper explores the performance of LLMs in the context of multi-dimensional analytic writing assessments, i.e. their ability to provide both scores and comments based on multiple assessment criteria.Using a corpus of literature reviews written by L2 graduate students and assessed by human experts against 9 analytic criteria, we prompt several popular LLMs to perform the same task under various conditions.To evaluate the quality of feedback comments, we apply a novel feedback comment quality evaluation framework.This framework is interpretable, cost-efficient, scalable, and reproducible, compared to existing methods that rely on manual judgments.We find that LLMs can generate reasonably good and generally reliable multi-dimensional analytic assessments.We release our corpus and code 1 for reproducibility.

  PublisherOA PDFDOI

• Evaluating learning trajectories of neural morphology acquisition models

  Linguistics Vanguard · 2025-09-08

  article1st authorCorresponding

  Abstract Computational models of morphology acquisition have played a central role in debates over the nature of morphological representations since the origin of the “past tense debate” in the 1980s. The apparent success of recent artificial neural network architectures for morphological inflection in natural language processing has revitalized this debate. However, despite their often good performance, the actual suitability of these advanced neural networks as models of human morphology acquisition remains uncertain. We argue that much of this confusion stems from inconsistent methods of training and evaluation. In this work, we demonstrate that more careful dataset creation and an evaluation combining quantitative analysis and comparison with human development puts the evaluation of neural models on firmer ground.

  PublisherDOI

• Lemmas Matter, But Not Like That: Predictors of Lemma-Based Generalization in Morphological Inflection

  2025-01-01

  articleOpen accessSenior author

  Recent work has shown that overlap -whether a given lemma or feature set is attested independently in train -drives model performance on morphological inflection tasks.The impact of lemma overlap, however, is debated, with accuracy drops from 0% to 30% reported between seen and unseen test lemmas.In this paper, we introduce a novel splitting algorithm designed to investigate predictors of accuracy on seen and unseen lemmas.We find only an 11% average drop from seen to unseen test lemmas but show that the number of lemmas in train has a much stronger effect on accuracy on unseen than seen lemmas.We also show that the previously reported 30% drop is inflated due to the introduction of a near-30% drop in the number of training lemmas from the original splits to the novel splits.These results help us better understand the factors affecting morphological generalization by neural models.

  PublisherOA PDFDOI

• LLMs can Perform Multi-Dimensional Analytic Writing Assessments: A Case Study of L2 Graduate-Level Academic English Writing

  ArXiv.org · 2025-02-17

  preprintOpen access

  The paper explores the performance of LLMs in the context of multi-dimensional analytic writing assessments, i.e. their ability to provide both scores and comments based on multiple assessment criteria. Using a corpus of literature reviews written by L2 graduate students and assessed by human experts against 9 analytic criteria, we prompt several popular LLMs to perform the same task under various conditions. To evaluate the quality of feedback comments, we apply a novel feedback comment quality evaluation framework. This framework is interpretable, cost-efficient, scalable, and reproducible, compared to existing methods that rely on manual judgments. We find that LLMs can generate reasonably good and generally reliable multi-dimensional analytic assessments. We release our corpus and code for reproducibility.

  PublisherOA PDFDOI

• Some Innate Characteristics of Neural Models of Morphological Inflection

  Underline Science Inc. · 2025-06-18

  otherOpen access

  Neural Network Models of Morphological Inflection (NNMIs) have deep relevance to cognitive science stemming from the central role that they played in the Past Tense Debate of the 1980s and 1990s. Critics of the connectionist approach to the mind frequently pointed to NNMIs’ shortcomings in the area of developmental realism: they argued that regardless of their ultimate accuracy, they failed to capture patterns of child language acquisition including developmental regressions and a propensity for over-regularization rather than irregularization. However, NNMIs have seen impressive improvement in the deep learning era of the 2010s and 2020s. Have modern NNMIs solved the old problems of developmental realism? We find that they have not. The persistence of these shortcomings suggests that they reflect “innate” characteristics of NNMIs as a class of learner, and that even substantial advancement in neural architectures and subsequent performance increases do not necessarily entail increased cognitive plausibility.

  PublisherDOI

Frequent coauthors

• Sarah R. Payne

  9 shared

• Salam Khalifa

  8 shared

• Spencer Caplan

  The University of Texas at Austin

  8 shared

• Charles Yang

  8 shared

• Chun Yang

  University of Science and Technology of China

  5 shared

• Christopher Cerezo Falco

  California University of Pennsylvania

  5 shared

• Natalia Krizhanovsky

  University of Massachusetts Amherst

  4 shared

• Nona Atanalov

  University of Massachusetts Amherst

  4 shared

Labs

• Jordan Kodner's LabPI

  Computational approaches to child language acquisition and their broader implications

Education

• Ph.D., Computer Science

  University of California, San Diego

  2008

• B.S., Computer Science

  University of California, San Diego

  2003