Strategic Information Disclosure and Recommendation for Crowdfunding

Description

Crowdfunding is a novel fundraising model that enables creators to raise funds from a crowd of contributors through online platforms (e.g., Kickstarter and Indiegogo). In return, the creators will use the funding to develop their innovative projects (e.g., digital arts, smartwatches, and board games), put them into production, and usually reward the contributors with the products. As crowdfunding usually involves the fundraising of innovative projects that have not existed in the market yet, there is a high degree of information asymmetry between creators and contributors. The online platforms alleviate it by releasing pledging status on its webpage and facilitating creators’ communication with contributors through product updates. Because such information plays an important role in shaping contributors’ perception towards the project success likelihood, it gives rise to an active research area of strategic information disclosure [1]. In this line of research, an honest yet strategic platform, which releases truthful information at the optimized timing for revenue maximization, is considered. So far, most of these studies are limited to empirical research, which lacks the theoretical foundation for the rigorous algorithm design. Thus, in Task 1, we aim to theoretically understand the impact of the information disclosed to the contributors’ pledging behavior, and develop a practical pricing and information disclosure algorithm. In addition, the online platforms create value for the creators and contributors by reducing their matching costs. Due to a large number of projects and contributors in a stochastic setting, the optimal matching problem aiming to maximize the project success ratio in the platform is a daunting task. Thanks to the recent advancement in artificial intelligence (AI), it is possible to improve the matching efficiency through a recommendation system. Unfortunately, the existing solutions are usually based on poorly motivated heuristics [2] that can lead to overfunding problems [3], while the academic research on crowdfunding matching services is limited and impractical. Thus, in Task 2, we aim to fill in this research gap on crowdfunding recommendation by first establishing the theoretical foundation on the allocation of advertisement space to projects. For practical implementations, we plan to design a real-time recommendation system based on individual contributors’ preferences by using the contextual multiarmed bandit framework. In summary, with our preliminary results [4, 5, 6], the proposed research is highly feasible. These research efforts will establish the theoretical understanding of strategic information disclosure and recommendation and lead to the design of practical algorithms in crowdfunding.