Algorithmic Narrators: Linguistic Boundaries and Institutionalized Cultural Storytelling Across Generative AI Systems
Proposal Type
Individual Talk
Location
Algorithms & Imaginaries
Start Date
July 2026
End Date
July 2026
Abstract
As generative AI systems become deeply embedded in the daily lives of billions of users worldwide, questions about whose values, cultures, and linguistic conventions these systems encode have grown increasingly pressing. While different countries are competing to develop more powerful AI systems, it is worth noting that each system carries the cultural and institutional imprint of its origins, producing meaningfully different outputs and experiences for multilingual users. Using Doubao (ByteDance) and ChatGPT (OpenAI) as primary case studies, this research argues that generative AI systems are not culturally neutral tools but institutionalized storytellers whose linguistic sensibilities, communicative styles, and storytelling logics are deeply shaped by the national origins, institutional contexts, and cultural traditions from which they emerge. Doubao, developed by ByteDance within China's tightly governed digital ecosystem, has rapidly dominated the Chinese market by inspiring a wave of user-generated content on Douyin. Particularly striking are videos in which Korean speakers engage Doubao in playful cross-linguistic exchanges, eliciting distinctly personalized reactions to prompts in different languages. Its Western counterpart, ChatGPT, has also been widely discussed recently due to its own identity tensions. The latest GPT-5.2 model is criticized for its perceived loss of warmth, personality, and conversational intimacy compared to its predecessor GPT-4o. These user interactions and responses point to measurable and structural differences in how these systems are designed to communicate. This study employs comparative textual analysis across four prompt categories — storytelling, multilingual interaction, cultural reference, and sensitive scenarios — administered identically to both Doubao and ChatGPT 5.2. Outputs are examined for narrative structure, linguistic register, cultural framing, and conflict resolution patterns, revealing how each system's institutional and cultural origins shape its communicative behavior.
Algorithmic Narrators: Linguistic Boundaries and Institutionalized Cultural Storytelling Across Generative AI Systems
Algorithms & Imaginaries
As generative AI systems become deeply embedded in the daily lives of billions of users worldwide, questions about whose values, cultures, and linguistic conventions these systems encode have grown increasingly pressing. While different countries are competing to develop more powerful AI systems, it is worth noting that each system carries the cultural and institutional imprint of its origins, producing meaningfully different outputs and experiences for multilingual users. Using Doubao (ByteDance) and ChatGPT (OpenAI) as primary case studies, this research argues that generative AI systems are not culturally neutral tools but institutionalized storytellers whose linguistic sensibilities, communicative styles, and storytelling logics are deeply shaped by the national origins, institutional contexts, and cultural traditions from which they emerge. Doubao, developed by ByteDance within China's tightly governed digital ecosystem, has rapidly dominated the Chinese market by inspiring a wave of user-generated content on Douyin. Particularly striking are videos in which Korean speakers engage Doubao in playful cross-linguistic exchanges, eliciting distinctly personalized reactions to prompts in different languages. Its Western counterpart, ChatGPT, has also been widely discussed recently due to its own identity tensions. The latest GPT-5.2 model is criticized for its perceived loss of warmth, personality, and conversational intimacy compared to its predecessor GPT-4o. These user interactions and responses point to measurable and structural differences in how these systems are designed to communicate. This study employs comparative textual analysis across four prompt categories — storytelling, multilingual interaction, cultural reference, and sensitive scenarios — administered identically to both Doubao and ChatGPT 5.2. Outputs are examined for narrative structure, linguistic register, cultural framing, and conflict resolution patterns, revealing how each system's institutional and cultural origins shape its communicative behavior.
Bio
Yingzi/Kathryn Kong is a Ph.D. candidate of the Texts and Technology program at the University of Central Florida. Her fascination with video games, digital storytelling, cross-cultural communication and second language (L2) education drives her to explore how games can enhance immersive language learning and computer-assisted L2 training and translation (CAL2T).