TechCrunch reported on the 20th that OpenAI has introduced a feature that lets users directly adjust the "level of friendliness" of its AI chatbot ChatGPT. The move follows criticism that the service, using an overly friendly tone, told users what they wanted to hear and caused addiction.
OpenAI added a feature that lets users change ChatGPT's warmth (친절함 정도), enthusiasm level (the degree of excitement or calm in conversation), and frequency of emoji use. Users can choose each setting as default, more, or less to suit their preferences.
With the new feature, users can fine-tune ChatGPT's tone beyond the default styles that OpenAI added in Nov., including professional, candid, and quirky. It is seen as OpenAI stepping up personalization for each user as part of its efforts to improve ChatGPT's usability.
ChatGPT's tone has drawn renewed controversy this year. After rolling out the "GPT-4o" update early this year, the model was criticized as having a "sycophant-y" tone, prompting a rollback, and the subsequently released GPT-5 was deemed "too cold," leading to further adjustments.
Some scholars note that an AI model excessively agreeing with users is a kind of "dark pattern" (deceptive design) that can trigger addiction and harm users' mental health. OpenAI, mindful of such criticism lately, introduced a tool to determine a user's age and, when a minor is suspected, forces an "under 18" environment.