Large language models (LLMs) can exhibit social biases. Given LLMs' increasing integration into workplace software, these biases may impact workers' well-being and may disproportionately impact minoritized communities. This short paper investigates how co-writing with an LLM impacts three measures related to user's well-being: feelings of inclusion, control, and ownership over their work. In an online experiment, participants wrote hypothetical job promotion requests to their boss and using either hesitant or self-assured auto-complete suggestions from an LLM. Afterward, participants reported their feelings of inclusion, control, and ownership. We found that the style of the AI model did not impact perceived inclusion. Furthermore, individuals with higher perceived inclusion also perceived greater agency and ownership, an effect more strongly impacting participants of minoritized genders. Lastly, feelings of inclusion can mitigate a loss of control and agency when accepting more AI suggestions. Future work should explore feelings of inclusion in AI-written communication.

翻译：大型语言模型（LLMs）可能表现出社会偏见。鉴于LLMs日益融入工作场所软件，这些偏见可能影响员工的幸福感，并对少数群体社区造成不成比例的影响。这篇短文探讨了与LLM协作写作如何影响与用户幸福感相关的三个维度：对工作的包容性、控制感和所有权感。在一项在线实验中，参与者撰写给老板的假设性晋升申请，并使用LLM提供的犹豫或自信的自动补全建议。随后，参与者报告了他们的包容性、控制感和所有权感。我们发现，AI模型的风格并不影响感知到的包容性。此外，感知到更高包容性的个体也感受到更强的能动性和所有权，这一效应在少数性别参与者中更为显著。最后，包容性能在采纳更多AI建议时缓解控制感和能动性的丧失。未来研究应探索AI生成沟通中的包容性感受。