Exploring AI for intercultural communication: open conversation

Zhu Hua, David Wei Dai, Adam Brandt, Guanliang Chen, Giuliana Ferri, Spencer Hazel, Chris Jenks, Rodney H. Jones, John P. O’Regan, Shungo SuzukiApplied Linguistics Review 2024 · DOI

TL;DR

This paper is a high-level synthesis of expert perspectives on the intersection of Artificial Intelligence and Intercultural Communication (ICC). It argues that while AI offers unprecedented tools for language learning and standardized interaction, it risks reinforcing "Anglo-centric" biases and eroding human agency unless researchers move from being passive spectators to active shapers of the technology.

What Problem Does This Paper Address?

The authors tackle the "speed and depth" of AI’s expansion into human sociality. The core challenge is that AI development is currently dominated by commercial interests and technical optimization, often ignoring the sociolinguistic nuances of culture.

The paper identifies a looming crisis: AI models often act as a "factory reset" for language, defaulting to a bland, "standard" English that erases regional identities (like Singlish or Chinglish) and reinforces existing power structures. The authors seek to bridge the gap between technical AI development and the human-centric insights of applied linguistics to ensure AI facilitates—rather than distorts—meaningful intercultural exchange.

Methods

As a "discussion forum" piece, the paper does not employ a traditional laboratory experiment. Instead, it uses a structured qualitative inquiry (an "open conversation"). Ten experts from diverse backgrounds—including Conversation Analysis (CA), sociolinguistics, and learning analytics—respond to five key questions regarding their research experiences, perceived challenges, and the future of the field. This method allows for a multi-perspectival "epistemological mapping" of a rapidly evolving landscape that traditional empirical studies might struggle to capture in real-time.

Datasets

Not discussed in the paper. The authors do not introduce a new dataset; however, they provide a critical critique of existing Large Language Model (LLM) training data, noting it is heavily biased toward "Anglo-centric" data and "dominant linguacultural capitals."

Key Insights and Ideas

  • AI as a "Systematic Interlocutor": In professional training (e.g., nursing), AI can provide a standardized, fair experience for learners that human actors cannot replicate due to "interactional variability."
  • The "Standardization" Trap: There is a deep concern that generative AI is "linguistically and culturally illiterate." By defaulting to "standard" forms, AI may inadvertently perform a "massive toxicity" by distorting intercultural relations into a monochrome, algorithmic version of reality.
  • Artificial Sociality: Researchers are moving beyond studying communication through machines to communication with machines. This requires a "theory of mind" for AI—understanding how we imagine AI "thinks" and how those imaginaries change our own sense of intelligence and humanity.
  • Four Modes of Communication: The paper suggests a framework for future research based on four interfaces:
  • Human-Machine-Human (mediation)
    1. Human-Machine (interaction)
    2. Machine-Human (perception)
    3. Machine-Machine (automated cultural knowledge production).
  • The "Oppenheimer" Moment: Several contributors warn of the "malevolent use" of AI insights for surveillance or warfare (citing the "Lavender" program), urging researchers to ask if their work could be manipulated for harm.

Key Results

While qualitative, the "results" of this conversation establish a roadmap for the field:

  • Consensus on Agency: ICC researchers must stop being "spectators" and start "channeling energy into shaping the technology" through techniques like prompt engineering and bias mitigation.
  • Methodological Shift: The authors advocate for using Membership Categorization Analysis (MCA) and Conversation Analysis (CA) to study "culture in action" within AI interactions, identifying how AI represents (or misrepresents) social categories like "lawyer" or "trustworthy person."
  • Policy Call: A strong recommendation for "design justice" and government intervention to ensure the "AI public square" isn't entirely dominated by market motives.

Limitations and Future Work

  • Technological Lag: The authors acknowledge that the "accelerated advancement" of AI makes it nearly impossible for individuals or researchers to keep up using traditional academic cycles.
  • The "Black Box" of Culture: Current AI struggles with the "nuances of different cultural contexts," often failing to grasp cultural norms and values that vary across communities.
  • Future Directions: The authors propose exploring "digital interculturality"—how AI alters how we make sense of sensitive topics like war, religion, and migration. They also suggest investigating whether AI might eventually prove better than humans at identifying "distributional variations" in how different cultural groups behave, provided they are prompted correctly.

Summary updated

Abstract

Abstract AI is not new. What is new, however, is the speed and depth of its expansion in almost every aspect of our lives. This discussion forum is dedicated to exploring new frontiers and agendas for language and intercultural communication research. In this concluding piece, we invite the contributors to share insights on five key questions: their experiences (Question 1), the challenges and opportunities that we face (Question 2), the strengths and skills afforded by intercultural communication and applied linguistics (Question 3), considerations when collaborating with AI developers and user groups (Question 4) and the future landscape of intercultural communication (Question 5). Through these inquiries, we hope to amplify the contributors’ voices and experiences, often difficult to fit in academic writing, but crucial for contextualizing their epistemological stances in their work. We seek to broaden the discussion, drawing out a bigger picture of pressing issues, and exploring future prospects.
BibTeX
@article{hua2024exploring,
  author = {Zhu Hua and David Wei Dai and Adam Brandt and Guanliang Chen and Giuliana Ferri and Spencer Hazel and Chris Jenks and Rodney H. Jones and John P. O’Regan and Shungo Suzuki},
  title = {{Exploring AI for intercultural communication: open conversation}},
  journal = {Applied Linguistics Review},
  year = {2024},
  doi = {10.1515/applirev-2024-0186},
}
Figure