AAAL 2026: Invited Colloquium

Identity, ideology, and capital: Investing in agentive generative AI practices

Ron Darvin, The University of British Columbia

The proliferation of generative AI (GenAI) has significant implications for language and literacy education, reshaping traditional understandings of authorship, creativity, and learner agency. This paper examines GenAI practices through the conceptual lenses of identity, ideology, and capital as articulated by Darvin and Norton’s (2015) model of investment, and augmented by notions of sociomateriality (Fenwick, 2015). By foregrounding the interaction between human learners and algorithmically driven platforms, this paper highlights the complex distribution of agency and power across human and nonhuman actors. Learners' access to material, cultural, social, linguistic and semiotic resources can shape dispositions towards GenAI and practices. At the same time,  platform designs, algorithmic processes, and datasets are not neutral but material manifestations of ideologies and power structures that can steer interactions, where both learners and tools are positioned in different ways. By applying a critical lens to understand human-AI interactions, this paper underscores that learners’ identities and their capacity to exert agency are deeply intertwined with their ability to recognize, interrogate, and navigate these embedded ideologies. To invest in agentive GenAI practices, learners need to  develop critical digital literacies that discourage them from passively consuming AI-generated content and instead enable them to recognize how power operates in these interactions.  By actively negotiating their own meanings and intentions, learners can transform their GenAI practices into meaningful learning opportunities.

How does interaction with LLM chatbots shape human understanding of culture?

David Wei Dai, University College London
Zhu Hua, University College London
Guanliang Chen, Monash University

Against the proliferation of Large-Language-Models (LLM)-based Artificial Intelligence (AI) products such as ChatGPT and Gemini, and their increasing use in professional communication training, researchers, including applied linguists, have cautioned that these products (re)produce cultural stereotypes due to their training data (Dai & Zhu, 2024). However, there is a limited understanding of how humans navigate the assumptions and biases present in the responses of these LLM-powered systems, described by Jones (2024) as “culture machines” and the role humans play in perpetuating stereotypes during interactions with LLMs. In this article, we use Sequential-Categorial Analysis, which combines Conversation Analysis and Membership Categorization Analysis, to analyse simulated interactions between a human physiotherapist and three LLM-powered chatbot patients of Chinese, Australian, and Indian cultural backgrounds. Coupled with analysis of information elicited from LLM chatbots and the human physiotherapist after each interaction, we demonstrate that users of LLM-powered systems are highly susceptible to becoming interactionally entrenched in culturally essentialized narratives. We use the concepts of interactional instinct and interactional entrenchment to argue that whilst human-AI interaction may be instinctively prosocial, LLM users need to develop Critical Interactional Competence (CritIC) for human-AI interaction through appropriate and targeted training and intervention, especially when LLM-powered tools are used in professional communication training programmes.

Digital accents, homogeneity-by-design, and the evolving social science of written language

AJ Alvero, Cornell University
Quentin Sedlacek, Southern Methodist University
Maricela León, Southern Methodist University
Courtney Peña, Stanford University

Human language is increasingly written rather than just spoken, primarily due to the proliferation of digital technology in modern life. This trend has enabled the creation of generative AI trained on corpora containing trillions of words extracted from text on the internet. However, current language theory inadequately addresses digital text communication's unique characteristics and constraints. This paper systematically analyzes and synthesizes existing literature to map the theoretical landscape of digital language evolution. The evidence demonstrates that, parallel to spoken language, features of written communication are frequently correlated with the socially constructed demographic identities of writers, a phenomenon we refer to as ``digital accents.'' This conceptualization raises complex ontological questions about the nature of digital text and its relationship to identity. The same line of questioning, in conjunction with recent research, shows how generative AI systematically fails to capture the breadth of expression observed in human writing, an outcome we call “homogeneity-by-design.'' By approaching text-based language from this theoretical framework while acknowledging its inherent limitations, social scientists studying language can strengthen their critical analysis of artificial intelligence systems and contribute meaningful insights to their development and improvement.

Bias and stereotyping: Human and artificial Intelligence

Okim Kang, Northern Arizona University
Kevin Hirschi, University of Taxes – San Antonio

Much research has demonstrated listeners’ biases toward L2-accented speech, i.e., perceiving accented utterances as less credible (Lev-Ari & Keysar, 2010), less grammatical (Ruivivar & Collins, 2019) or less acceptable for certain professional positions (Kang et al., 2023), due to their bias and stereotyping issues (Kang & Rubin, 2009). However, AI technologies have emerged as a viable alternative to human perception because they allow for a better understanding of how biases and stereotypes are induced through design, implementation, and training data. For example, in facial recognition, AI recognition has long been known to perform poorly for visual minorities (Hardesty, 2018). In AI systems for speech recognition, AI systems have shown biased and subpar performance in capturing African American Language (Martin & Wright, 2023), and male speakers were more correctly transcribed by YouTube’s automatically generated captions than female speakers (Tatman, 2017). Furthermore, Chinese first language (L1) speakers were less accurately processed than Spanish and Indian L1 speakers of English (Bae & Kang, 2024). The implications of such social biases and their persistence in AI have not yet been fully understood by Applied Linguistics scholars or computer scientists who investigate AI ethics. Accordingly, in this presentation, we will discuss issues in social biases that impact users of different language varieties and relate these to the currently limited research on biases in AI. We will focus on stereotyping research amongst human listeners in various social and educational contexts as well as review research that investigates these biases in AI. We propose that bias found in social perception research can serve as an agenda for how to investigate AI models’ classification, processing, and generation tasks. The presentation will conclude with specific recommendations and future directions for research and pedagogical practices.

Addressing GenAI biases for ELT from a Global Englishes perspective

Seongyong Lee, University of Nottingham Ningbo China 
Jaeho Jeon, University of Alabama
Jim McKinley, University College London
Heath Rose, University of Oxford 

This study explores the capability of ChatGPT, a large language model (LLM)-based generative AI (GenAI) tool, to support Global Englishes Language Teaching (GELT), addressing the critical gap between GenAI’s potential and its inherent biases that limit representation of diverse English varieties. LLMs typically overrepresent dominant English varieties and favor standardization through algorithmic design, undermining the pluricentric principles of GELT. Thus, it aims to develop effective GenAI-GELT instructional modules using customized ChatGPT models. Through Design and Development Research methodology, we developed three models: a Basic Model using single-step prompting, Refined Model 1 employing multi-step prompting, and Refined Model 2 integrating ELF corpora as external datasets with advanced prompt engineering techniques. The Models were assessed across three pedagogical roles: content curator for World Englishes materials, evaluative feedback provider for ELF writing, and conversation partner facilitating ELF interaction. Findings revealed that while basic prompting defaulted to native-speaker norms, the integration of advanced prompt engineering with external corpus data substantially improved outputs aligned with GELT principles, though standardization biases persisted. The study highlights three critical implications: the importance of teacher-GenAI collaborative design processes, the necessity of teachers' GELT-specific professional knowledge, and their essential role in orchestrating complementary resources to maximize GenAI's pedagogical effectiveness while mitigating inherent biases in language education.