Generative AI for Crowdfunding: Promises and Pitfalls

Jump to: navigation, search

Title: Generative AI for Crowdfunding: Promises and Pitfalls

Mentor: Dr. Larry Zhiming Xu

Approach: The study will use large language models (e.g., ChatGPT or equivalent) to generate crowdfunding campaign messages based on human-generated messages in existing fundraising campaigns. Leveraging natural language processing (NLP) techniques, it will compare AI- and human-generated text in terms of linguistic characteristics and evaluate whether AI might outperform humans in certain aspects. Randomized experiments will also be conducted to incorporate human judgment in assessing the feasibility of using generative AI for crowdfunding message writing and strategy implementation.

Summary: Crowdfunding has been widely used for entrepreneurial and nonprofit fundraising on a global scale. However, existing literature has shown that most crowdfunding messages are poorly crafted and lead to undesirable fundraising outcomes. With generative AI tools such as ChatGPT becoming readily available, crowdfunding platforms are envisioning the future of incorporating such services for facilitating fundraising management and improving crowdfunding outcomes. Inevitably, these promises may come with pitfalls, raising new moral and ethical concerns over the limitations of generative AI. Hence, this study is among the first of its kind to explore the opportunities and challenges that generative AI has created for crowdfunding management.

Student activities: The REU fellow(s) will perform the following major tasks: 1) Retrieving data from publicly available webpages (e.g., GoFundMe, Kickstarter); 2) Generating content using large language models; 3) Preparing data for appropriate analysis; 4) Assisting in performing the analysis on a large text corpus consisting of both AI- and human-generated crowdfunding project descriptions; 4) Getting familiar with NLP analytical procedures such as sentiment analysis, multinomial language models, and distributed representation of words; 5) Getting familiar with corresponding data visualization tools and techniques. 6) Assisting in writing and publishing academic articles.

Student background: Students should be familiar with web scraping/crawling techniques and can retrieve and store unstructured data from the web. Basic knowledge of natural language processing and machine learning is preferred.