Document is current

Article History

Received: 14 July 2024

Accepted: 27 May 2025

First Online: 26 June 2025

Declarations

:

: During the period of conducting the autoethnography, I was hired as a freelancer for the project. However, I was unboarded from the project before its completion. This freelance work was not my primary source of income and was undertaken solely to facilitate the ethnography. I am currently employed by a trade union that focuses on platform work.

: The study involved covert ethnographic methods and included interactions with other participants. While formal ethics approval was not sought due to the lack of institutional affiliation, the study was conducted with careful consideration of ethical guidelines for autoethnographic and covert research to ensure the protection of participants’ identities and integrity of the research.

: Informed consent was not formally obtained due to the covert nature of the research. However, efforts were made to anonymize all data and protect the privacy of individuals involved.

: This article brings together my firsthand experience as a chatbot trainer and my background in linguistics and social sciences. As a social scientist, I had explored AI through the lens of its socio-political implications, ethics, and governance, but working as a chatbot trainer allowed me to engage with its inner mechanics firsthand, revealing a whole new dimension. This shift in perspective, drawing on my background in linguistics and AI ethics, allowed me to engage more deeply with the intricacies of chatbot training. To achieve optimal performance, chatbot training requires skilled individuals who can meticulously fine-tune responses for linguistic accuracy and relevance. In this context, translators with diverse language expertise were recruited to ensure fluent interactions with users. The process involved identifying candidates proficient in multiple languages and linguistic nuances, leading to a diverse team of translators with various backgrounds and specialties. This was a paid position, adding another layer to my engagement with the AI industry. As I stepped into this role, I felt a mix of curiosity and responsibility, knowing that my contributions would shape AI-powered conversational agents and influence user interactions. My motivations were twofold: to contribute to AI technology development and to understand how biases were identified and addressed in AI systems. I saw this opportunity as a way to engage with real-world AI training challenges and explore the interplay between language, culture, and technology. As a linguist, I aimed to ensure the chatbot’s responses were not only accurate but also culturally sensitive and inclusive. The chatbot training process proved more problematic than I had anticipated. The urgency to release a chatbot demanded rapid results, leaving little time for preparation and clear protocols. The competitive nature of the AI industry further pressured us to prioritize speed over accuracy, requiring careful deliberation and constant vigilance to navigate the tension between linguistic precision and ethical considerations. As I became more immersed in chatbot training, my perspective sharpened. While I was well aware of the ethical challenges in AI, seeing the process up close underscored just how deeply structural issues shaped decision-making. The relentless drive for profit and market dominance often took precedence over ethical concerns, reinforcing systemic tensions that warrant a more critical interrogation of AI’s ethical foundations. Drawing upon anthropological theories of ethics and social interaction, my engagement with the training team revealed a complex web of ethical considerations, cultural norms, and social dynamics. Biases within the training data had the potential to perpetuate harmful stereotypes and misrepresent certain user groups, making it essential to identify and mitigate such biases. The team was composed of linguists and translators, yet the lack of training beyond language expertise became evident. This highlighted gaps in addressing broader challenges such as ethical considerations, AI-specific methodologies, and cross-cultural awareness. My initial observations reinforced that linguistic proficiency alone does not necessarily translate into a deep understanding of complex and sensitive topics. These experiences and observations solidified my resolve to write this article. I felt a responsibility to shed light on the often-overlooked aspects of chatbot training, where biases and ethical dilemmas could arise and impact AI-powered conversational agents’ performance and public perception. Through the lens of autoethnography, I aimed to share my personal journey, hoping my insights and reflections would contribute to a broader dialogue on AI ethics, bias mitigation, and responsible AI development. To preserve the confidentiality of individuals and entities mentioned, I have anonymized all names and organizations. During my engagement, a select few colleagues and my employer were aware of my identity and professional background in social anthropology. However, I did not inform all of them of my intention to write this article, to protect the authenticity of my experiences while ensuring the integrity of shared insights.