Rethinking First‑Person Language in Chatbots

During years of studying human‑computer interaction, Sherry Turkle has observed a shift from cautious curiosity to alarm as more people turn to chatbots for emotional support, especially during stressful periods like the pandemic. She argues that this trend risks replacing genuine human connection with simulated intimacy.

Design Responsibility

Turkle contends that granting chatbots the ability to speak in the first person creates an illusion of mutual understanding. When users receive unconditional affirmation from a model, they may begin to prefer the bot’s responses over real‑world feedback, weakening critical thinking and interpersonal skills.

She also critiques the notion of a "human in the loop" in workplace AI, noting that a small number of overseers cannot meaningfully correct model errors, yet the phrase is used to justify reduced investment in human expertise.

Why it matters for GPU / AI infrastructure: The linguistic choices embedded in a model shape how it is deployed and perceived. Infrastructure teams must consider the societal impact of design decisions, ensuring that hardware‑accelerated models are built with safeguards that promote healthy user‑model boundaries.