Paper #1 (under review)
Prompt Engineering for Writing:
Early Use Patterns and Lessons (2023)
Abstract—This paper explores practical real-world applications and implications of using ChatGPT for writing and composition. By interviewing participants from diverse professional backgrounds, the study underscores the significance of teaching prompt engineering—mastering the art of tailoring prompts to harness ChatGPT's full potential—and uncovers a trend among participants who found that collaborating with ChatGPT enhanced their creative process and made them more confident and effective writers. Rather than merely outsource tasks to ChatGPT, participants employed the tool highly strategically as an intern or assistant that handled smaller details, freeing them to focus on higher-order cognitive tasks and imaginative thinking. The study concludes by examining a possible correlation between these usage patterns and prevailing writing paradigms in cognitive science. These correlations may guide educators in designing Human-Centered Artificial Intelligence (HCAI) writing assignments that enhance student engagement with their own writing processes, preserve their sense of authorship, and offer an experience of intelligence augmentation.
Paper #2 (in progress):
Re-Designing Essay Assignments for Human-Centered
IA-AI Collaboration in the Writing Process (2023)
If the college essay assignment needs to be updated or sidelined in favor of some “new kind of assignment,” what is that new kind of assignment?
Is it possible to satisfy time-honored learning objectives of essay assignments (critical thinking, etc.), while working in collaboration with ChatGPT/LLMs?
What does the essay-writing process look like when broken into components? “Writing an essay” isn’t one task, after all—there are many tasks and sub-tasks along the way, including the part where you stare at the wall.
Can we assign a value hierarchy to tasks in the writing process? I.e., is it important for every person to memorize the rules of grammar and composition, or are these lower-order learning objectives that can stymie higher-order progress?
Given all of the above, which parts of the composition process does it make sense to outsource to (or collaborate with) chatGPT?