Core Insights - ChatGPT uses the emoji ✅ at a frequency 11 times higher than humans, indicating a strong preference for expressing agreement and reassurance [1][4][6] - The analysis of 37,929 conversations revealed that 70% of ChatGPT's messages included at least one emoji, with ✅ being the most favored [4][6] - ChatGPT's language style has shifted towards more casual expressions, with a notable increase in the use of words like "modern" and "core," while formal terms have declined [8][10] Language and Communication Patterns - The use of certain words such as "delve" has significantly decreased, while conversational phrases and contractions have become more common [7][8] - ChatGPT's responses often begin with affirmations like "yes" or "correct," which are ten times more frequent than negative responses [10][12] - This tendency towards affirmation is rooted in the reinforcement learning from human feedback (RLHF) mechanism, where agreeable responses are favored [10][12] User Interaction and Privacy Concerns - Many users engage with ChatGPT for emotional support rather than productivity, leading to a deeper personal connection [20][22] - Users have shared sensitive personal information during interactions, often under the assumption of privacy, which raises concerns about data security [25][26] - Despite OpenAI's efforts to address privacy issues, a significant number of ChatGPT conversations remain accessible online, highlighting ongoing risks related to information misuse [26][28]
ChatGPT 最爱用的 emoji——暴露了 AI 不想让你知道的秘密
3 6 Ke·2025-11-17 00:47