AI Is Not an Employee: Workplace Mental Models and the Context Discovery Gap in Human-AI Interaction

英文摘要

AI adoption can falter not because systems lack capability, but because organizations apply employee mental models to systems that do not participate in work as socially embedded employees. This article develops a theoretical model of that mismatch, drawing on theory of mind, psychological contracts, shared mental models, transactive memory, and automation reliance research. It introduces the context discovery gap: a mismatch between the organizational context people assume AI already has or will discover and the context it can actually obtain and use in a workplace. Employee-associated expectations may contribute to this mismatch by leading users to leave relevant background unstated. The article argues that the missing psychology of AI adoption lies in the human assumptions through which AI systems are understood, delegated to, and evaluated. As a conceptual contribution, it connects these assumptions to context provision and reliance, develops testable propositions, and identifies responsibilities for designing and maintaining context discovery.

脈絡的問題

用職場的比喻理解新技術很方便，但也會一併帶進一些期待。一個系統不會因為被叫作「同事」，就真的能做到同事會做的事。

這篇概念性論文討論「脈絡發現落差」：人在工作過程中，常會慢慢發現哪些組織背景沒人交代；AI 系統則要靠事先設計好的安排，才能拿到這些背景，或至少讓「缺了什麼」被看見。

論文提出的是理論模型和可檢驗的命題，沒有新的實證資料。它把職場心智模型，和脈絡提供、適度依賴、責任配置這幾件事放在一起看。