Data visualization practitioners often lack formal training, resulting in a knowledge gap in visualization design best practices. Large-language models like ChatGPT, with their vast internet-scale training data, offer transformative potential in addressing this gap. To explore this potential, we adopted a mixed-method approach. Initially, we analyzed the VisGuide forum, a repository of data visualization questions, by comparing ChatGPT-generated responses to human replies. Subsequently, our user study delved into practitioners' reactions and attitudes toward ChatGPT as a visualization assistant. Participants, who brought their visualizations and questions, received feedback from both human experts and ChatGPT in a randomized order. They filled out experience surveys and shared deeper insights through post-interviews. The results highlight the unique advantages and disadvantages of ChatGPT, such as its ability to quickly provide a wide range of design options based on a broad knowledge base, while also revealing its limitations in terms of depth and critical thinking capabilities.

翻译：数据可视化从业者往往缺乏正规训练，导致在可视化设计最佳实践方面存在知识缺口。以ChatGPT为代表的大语言模型，凭借其海量的互联网训练数据，在填补这一缺口方面展现出变革性潜力。为探索这一潜力，我们采用了混合研究方法。首先，我们分析了可视化设计论坛这一数据可视化问题库，将ChatGPT生成的回复与人工回复进行对比。随后，通过用户研究深入探讨了从业者将ChatGPT作为可视化助手的反应与态度。参与者携带自己的可视化作品与问题，以随机顺序接收人类专家与ChatGPT的反馈。他们填写体验问卷，并通过后续访谈分享更深入的见解。研究结果揭示了ChatGPT的独特优势与不足，例如其能基于广博知识库快速提供多样化设计选项，同时也暴露出其在深度与批判性思维能力方面的局限性。