The paper explores the unintended impacts of alignment procedures on Large Language Models (LLMs) in terms of global representation, focusing on English dialects, multilingualism, and global opinions. The authors find that current alignment procedures create disparities between English dialects and global opinions, improve multilingual capabilities, and increase alignment with US opinions relative to other regions. They discuss design decisions that lead to these unintended impacts and provide recommendations for more equitable preference tuning. The study uses a variety of datasets and models to evaluate the effects of alignment on downstream tasks, highlighting the need for transparency and careful consideration of the diverse user base when aligning LLMs. The paper also includes a new dataset of opinionated questions about countries and discusses the limitations and ethical considerations of their work.The paper explores the unintended impacts of alignment procedures on Large Language Models (LLMs) in terms of global representation, focusing on English dialects, multilingualism, and global opinions. The authors find that current alignment procedures create disparities between English dialects and global opinions, improve multilingual capabilities, and increase alignment with US opinions relative to other regions. They discuss design decisions that lead to these unintended impacts and provide recommendations for more equitable preference tuning. The study uses a variety of datasets and models to evaluate the effects of alignment on downstream tasks, highlighting the need for transparency and careful consideration of the diverse user base when aligning LLMs. The paper also includes a new dataset of opinionated questions about countries and discusses the limitations and ethical considerations of their work.