Reflection on Data Storytelling Tools in the Generative AI Era from the Human-AI Collaboration Perspective

To lower the barrier of telling appealing and effective stories, researchers have spent considerable efforts to build AI-powered tools to facilitate their creation and communication with different strategies of human-AI collaboration [29]. In these tools, AI collaborators are often powered by heuristic-based methods [57], traditional machine learning models [13], or smaller-scale deep learning models [36]. Compared to the previous techniques for AI collaborators, the recently emerging large-scale generative AI models, including the text-to-image models [62] and large language models (LLMs) [69], can achieve better performance on various data storytelling-related tasks, such as data analysis [20] and text generation [72], and enhance the communication between humans and AI with conversations. These advantages indicate their potentials to be game-changers in the research direction of human-AI collaboration for data storytelling, including improving the experience of collaborating with AI and diversifying the collaboration patterns between humans and AI [29]. After two years of the public access of these models, it is a critical time point to reflect how this research discipline progresses in the new era of large-scale generative AI models and identify future opportunities. To achieve the goal, it is essential not only to focus on how these generative AI techniques are applied in existing tools, as explored in a previous survey [17], but more importantly, to compare the human-AI collaboration patterns in the latest tools in the generative AI era with those in earlier ones. Only through this comparison can we understand the shift in human-AI collaboration paradigms, identify the value of these powerful techniques in enhancing human-AI collaboration, and propose future research directions.