Tokenization is a necessary component within the current architecture of many language models, including the transformer-based large language models (LLMs) of Generative AI, yet its impact on the model's cognition is often overlooked. We argue that LLMs demonstrate that the Distributional Hypothesis (DH) is sufficient for reasonably human-like language performance, and that the emergence of human-meaningful linguistic units among tokens motivates linguistically-informed interventions in existing, linguistically-agnostic tokenization techniques, particularly with respect to their roles as (1) semantic primitives and as (2) vehicles for conveying salient distributional patterns from human language to the model. We explore tokenizations from a BPE tokenizer; extant model vocabularies obtained from Hugging Face and tiktoken; and the information in exemplar token vectors as they move through the layers of a RoBERTa (large) model. Besides creating sub-optimal semantic building blocks and obscuring the model's access to the necessary distributional patterns, we describe how tokenization pretraining can be a backdoor for bias and other unwanted content, which current alignment practices may not remediate. Additionally, we relay evidence that the tokenization algorithm's objective function impacts the LLM's cognition, despite being meaningfully insulated from the main system intelligence.

Over the past decade, the gender pay gap has remained steady with women earning 84 cents for every dollar earned by men on average. Many studies explain this gap through demand-side bias in the labor market represented through employers' job postings. However, few studies analyze potential bias from the worker supply-side. Here, we analyze the language in millions of US workers' resumes to investigate how differences in workers' self-representation by gender compare to differences in earnings. Across US occupations, language differences between male and female resumes correspond to 11% of the variation in gender pay gap. This suggests that females' resumes that are semantically similar to males' resumes may have greater wage parity. However, surprisingly, occupations with greater language differences between male and female resumes have lower gender pay gaps. A doubling of the language difference between female and male resumes results in an annual wage increase of $2,797 for the average female worker. This result holds with controls for gender-biases of resume text and we find that per-word bias poorly describes the variance in wage gap. The results demonstrate that textual data and self-representation are valuable factors for improving worker representations and understanding employment inequities.

Large Language Models (LLMs) like ChatGPT reflect profound changes in the field of Artificial Intelligence, achieving a linguistic fluency that is impressively, even shockingly, human-like. The extent of their current and potential capabilities is an active area of investigation by no means limited to scientific researchers. It is common for people to frame the training data for LLMs as "text" or even "language". We examine the details of this framing using ideas from several areas, including linguistics, embodied cognition, cognitive science, mathematics, and history. We propose that considering what it is like to be an LLM like ChatGPT, as Nagel might have put it, can help us gain insight into its capabilities in general, and in particular, that its exposure to linguistic training data can be productively reframed as exposure to the diegetic information encoded in language, and its deficits can be reframed as ignorance of extradiegetic information, including supradiegetic linguistic information. Supradiegetic linguistic information consists of those arbitrary aspects of the physical form of language that are not derivable from the one-dimensional relations of context -- frequency, adjacency, proximity, co-occurrence -- that LLMs like ChatGPT have access to. Roughly speaking, the diegetic portion of a word can be thought of as its function, its meaning, as the information in a theoretical vector in a word embedding, while the supradiegetic portion of the word can be thought of as its form, like the shapes of its letters or the sounds of its syllables. We use these concepts to investigate why LLMs like ChatGPT have trouble handling palindromes, the visual characteristics of symbols, translating Sumerian cuneiform, and continuing integer sequences.

Recent studies suggest social media activity can function as a proxy for measures of state-level public health, detectable through natural language processing. We present results of our efforts to apply this approach to estimate homelessness at the state level throughout the US during the period 2010-2019 and 2022 using a dataset of roughly 1 million geotagged tweets containing the substring ``homeless.'' Correlations between homelessness-related tweet counts and ranked per capita homelessness volume, but not general-population densities, suggest a relationship between the likelihood of Twitter users to personally encounter or observe homelessness in their everyday lives and their likelihood to communicate about it online. An increase to the log-odds of ``homeless'' appearing in an English-language tweet, as well as an acceleration in the increase in average tweet sentiment, suggest that tweets about homelessness are also affected by trends at the nation-scale. Additionally, changes to the lexical content of tweets over time suggest that reversals to the polarity of national or state-level trends may be detectable through an increase in political or service-sector language over the semantics of charity or direct appeals. An analysis of user account type also revealed changes to Twitter-use patterns by accounts authored by individuals versus entities that may provide an additional signal to confirm changes to homelessness density in a given jurisdiction. While a computational approach to social media analysis may provide a low-cost, real-time dataset rich with information about nationwide and localized impacts of homelessness and homelessness policy, we find that practical issues abound, limiting the potential of social media as a proxy to complement other measures of homelessness.

While quantitative methods have been used to examine changes in word usage in books, studies have focused on overall trends, such as the shapes of narratives, which are independent of book length. We instead look at how words change over the course of a book as a function of the number of words, rather than the fraction of the book, completed at any given point; we define this measure as "cumulative word-time". Using ousiometrics, a reinterpretation of the valence-arousal-dominance framework of meaning obtained from semantic differentials, we convert text into time series of power and danger scores in cumulative word-time. Each time series is then decomposed using empirical mode decomposition into a sum of constituent oscillatory modes and a non-oscillatory trend. By comparing the decomposition of the original power and danger time series with those derived from shuffled text, we find that shorter books exhibit only a general trend, while longer books have fluctuations in addition to the general trend. These fluctuations typically have a period of a few thousand words regardless of the book length or library classification code, but vary depending on the content and structure of the book. Our findings suggest that, in the ousiometric sense, longer books are not expanded versions of shorter books, but are more similar in structure to a concatenation of shorter texts. Further, they are consistent with editorial practices that require longer texts to be broken down into sections, such as chapters. Our method also provides a data-driven denoising approach that works for texts of various lengths, in contrast to the more traditional approach of using large window sizes that may inadvertently smooth out relevant information, especially for shorter texts. These results open up avenues for future work in computational literary analysis, particularly the measurement of a basic unit of narrative.

Well curated, large-scale corpora of social media posts containing broad public opinion offer an alternative data source to complement traditional surveys. While surveys are effective at collecting representative samples and are capable of achieving high accuracy, they can be both expensive to run and lag public opinion by days or weeks. Both of these drawbacks could be overcome with a real-time, high volume data stream and fast analysis pipeline. A central challenge in orchestrating such a data pipeline is devising an effective method for rapidly selecting the best corpus of relevant documents for analysis. Querying with keywords alone often includes irrelevant documents that are not easily disambiguated with bag-of-words natural language processing methods. Here, we explore methods of corpus curation to filter irrelevant tweets using pre-trained transformer-based models, fine-tuned for our binary classification task on hand-labeled tweets. We are able to achieve F1 scores of up to 0.95. The low cost and high performance of fine-tuning such a model suggests that our approach could be of broad benefit as a pre-processing step for social media datasets with uncertain corpus boundaries.

Medical systems in general, and patient treatment decisions and outcomes in particular, are affected by bias based on gender and other demographic elements. As language models are increasingly applied to medicine, there is a growing interest in building algorithmic fairness into processes impacting patient care. Much of the work addressing this question has focused on biases encoded in language models -- statistical estimates of the relationships between concepts derived from distant reading of corpora. Building on this work, we investigate how word choices made by healthcare practitioners and language models interact with regards to bias. We identify and remove gendered language from two clinical-note datasets and describe a new debiasing procedure using BERT-based gender classifiers. We show minimal degradation in health condition classification tasks for low- to medium-levels of bias removal via data augmentation. Finally, we compare the bias semantically encoded in the language models with the bias empirically observed in health records. This work outlines an interpretable approach for using data augmentation to identify and reduce the potential for bias in natural language processing pipelines.

In an effort to better understand meaning from natural language texts, we explore methods aimed at organizing lexical objects into contexts. A number of these methods for organization fall into a family defined by word ordering. Unlike demographic or spatial partitions of data, these collocation models are of special importance for their universal applicability. While we are interested here in text and have framed our treatment appropriately, our work is potentially applicable to other areas of research (e.g., speech, genomics, and mobility patterns) where one has ordered categorical data, (e.g., sounds, genes, and locations). Our approach focuses on the phrase (whether word or larger) as the primary meaning-bearing lexical unit and object of study. To do so, we employ our previously developed framework for generating word-conserving phrase-frequency data. Upon training our model with the Wiktionary---an extensive, online, collaborative, and open-source dictionary that contains over 100,000 phrasal-definitions---we develop highly effective filters for the identification of meaningful, missing phrase-entries. With our predictions we then engage the editorial community of the Wiktionary and propose short lists of potential missing entries for definition, developing a breakthrough, lexical extraction technique, and expanding our knowledge of the defined English lexicon of phrases.