Teaching Data Storytelling in the Age of AI

AI is changing how we write, but it doesn't replace the creative, messy, and iterative process of storytelling, especially when working with data. Writing isn't just about getting words on a page; it's about exploration, structure, and finding meaning. AI can be a great brainstorming partner, helping students generate ideas, refine drafts, and push past writer's block, but it shouldn't do all the heavy lifting. Inspired by "Teaching with AI" (Bowen & Watson), I've been experimenting with ways to incorporate AI into the classroom while keeping students actively engaged in their own creative process. This talk will explore strategies for integrating AI in a way that enhances, rather than replaces, the essential thinking and storytelling that make data communication effective. 

Being able to find the story in the data can seem like an elusive learning goal. What makes a good story, what is the process of discovering it, and how do we help students tell it well? In this talk I will share lessons learned from working with creative writers and students to design an interactive data exploration app to inspire creative writing about the environment. A key feature of this app is the accompanying writing prompts that help guide users towards features of the data that may be particularly generative. I will discuss the relationship between what makes a good writing prompt and what makes a good story and share writing prompts that can be transferred to a variety of statistics and data science classroom settings. 

Now that Large Language Models (LMMs) like ChatGPT are so adept at code, there is a tension between our desire to teach reproducible research methods and our desire to have students produce their own authentic work. One possible solution is to transition computing tasks to Graphical User Interfaces (GUIs). While a GUI like Excel or Tableau makes it harder to track reproducible work, it is also harder for students to rely fully on an LMM. A LMM can produce a list of instructions, but students need to follow those instructions appropriately for their own data. I have reintroduced GUIs into my data science classes alongside the code-based solutions I have long taught, and will discuss the strengths and drawbacks of this approach. 

Statistical writing is a key part of the undergraduate statistics curriculum, as it both helps students learn to reason about complex ideas and teaches them an important professional skill. As large language models (LLMs) offer increasingly sophisticated tools for automating writing, we must ask what effect they have on
the development of students as statisticians and statistical writers. We describe the results of several experiments comparing the grammatical and rhetorical style of students, experts, and LLMs, showing significant differences in how LLMs manage information, express confidence, and structure sentences. The
results suggest that LLMs do not effectively adapt their style to the genre and audience of their writing, and we explore the implications this has for teaching. If students offload the work of writing to a tool that does not write like a data scientist, how can we teach them to communicate like an expert data scientist? 

The *magnum opus* in the field of computer programming is surely Knuth's unfinished and provocatively-titled *The Art of Computer Programming*. Yet while computer programming can create art, few would characterize computer code itself *as* art, since it generally lacks the expressiveness and emotional connection that defines art. However, Knuth and many others do recognize computer code as a medium for expressing ideas, similar to the way we think about writing. Explicitly linking the two, Hermanns (2017) reminds us that "programming is writing is programming," and Vee (2017) goes so far as to argue that the ability to interpret computer code is becoming a new form of literacy. The latter argument dovetails with Nolan and Temple Lang (2010)'s call to integrate computing into the curriculum: if indeed the ability to read and write computer code is a necessary skill for producing high-quality statistics and data science work, then we should teach more of it. But all of these arguments were made before the recent advances in generative artificial intelligence, which is now capable of fitting and interpreting statistical models, analyzing data, and writing functioning computer code. With all this in mind, we introduce the notion of "artisanal programming" and ponder its place in higher education in the near future.