gulf of envisioning

预见鸿沟：Subramonyam 等人提出的概念，指用户在使用 LLM 等概率性 AI 系统时，难以提前想象/预判模型将生成什么内容、输出能否匹配自身需求。

English Definition

"A concept (Subramonyam et al.) describing the difficulty users face in predicting what a probabilistic AI system like an LLM will generate and how well that output will align with their goals."

用法说明 · 针对中文母语者

常与 gulf of execution、gulf of evaluation (Norman) 对照使用，强调 LLM 时代特有的输出不可预见性。

常见错误

·与 Norman 的'执行鸿沟/评估鸿沟'混淆——后者关注人机交互中的执行与反馈

真实用例 · 3 条

"Because outputs are probabilistic and unpredictable, users cannot know in advance what the model will generate or how well it will align with their goals, a difficulty Subramonyam et al. term the "gulf of envisioning""

论文Pipeline extracted — review needed·在引言中引出 LLM 交互中新的等待动态

"Subramonyam et al. [53] highlight a related "gulf of envisioning" in LLM use"

论文Pipeline extracted — review needed·在相关工作部分再次提及此概念

"Because users lack direct visibility into model processes, they imbue observable cues such as delays with explanatory weight"

论文Pipeline extracted — review needed·解释为何延迟成为用户推断模型行为的关键线索

由 pipeline 自动采集，待人工 review