Wordflow: A Novel Google Docs Add-on for Social Prompt Engineering
Author(s)
Chakravarthy, Aishwarya
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
Because large language models (LLMs) are trained on immensely large datasets from the Internet and beyond, they are highly useful for natural language tasks such as summarization, question-answering, and creative writing. The process of asking an LLM questions, called “prompting”, can enhance the accuracy of outputs. In spite of this, the art of prompting can often seem unintuitive to the average user, who is unfamiliar with NLP techniques. Recent research has focused on utilizing LLMs to serve as AI-writing assistants by allowing users to prompt LLMs; however, there is still a need to help users achieve the best prompts possible for their goals. To address this research gap, we introduce Wordflow, an AI-writing assistant presented as a Google Documents add-on that utilizes social prompt engineering to help users unlock the full potential of popular LLMs such as GPT and Gemini Pro. Within Wordflow, users can prompt LLMs and insert results directly into their documents, save prompts for local viewing, and share prompts to the larger community of users. Users can also search community prompts by topic to add prompts to their personal prompt library. The impact of Wordflow compared to prior research is its ability to combine social engineering and the power of LLMs to help users achieve their writing goals. A pilot study with ten participants was conducted to evaluate the efficacy of Wordflow; the results indicated that Wordflow is highly effective compared to current standard writing environments.
Sponsor
Date
Extent
Resource Type
Text
Resource Subtype
Undergraduate Research Option Thesis