Artificial intelligence's (AI) increasing presence in our daily lives, driven by its ability to process vast amounts of data and create personalized interventions, has coincided with a growing need to encourage prosocial behaviors in online communities. As digital platforms continue to shape social interactions, understanding how AI can promote positive behaviors and enhance prosocial outcomes in these spaces has become crucial. This thesis explores the intersection of AI and social science, examining how humans and AI can collaboratively encourage prosocial behaviors, with a specific focus on content moderation in video-sharing platforms.

The research employs a multi-faceted approach, beginning with a comprehensive literature review to establish a foundational understanding of human-AI interaction and prosocial behaviors in online communities. Building on this, the study utilizes technology probes and qualitative methods to gain insights into user engagement and behavior, particularly exploring how interactive features like Danmaku (bullet screen comments) can promote prosocial behaviors and strengthen social connections on video-sharing platforms.

A significant component of this research involves the development and evaluation of novel systems for human-AI co-creation in content moderation. The study introduces DanModCap, an innovative tool leveraging generative AI to moderate Danmaku comments and foster positive online interactions. To support effective human-AI teamwork for prosocial outcomes, new approaches incorporating shared goals, clear communication, and mutual respect were designed and implemented. Additionally, the research explores visually impactful interventions, such as Impact Captions, aimed at enhancing prosocial behavior and engagement on video-sharing platforms. These additions aimed to evoke emotional responses, improve user understandings of complex issues, and encourage positive actions.

My research also evaluated how human-AI co-creation affected prosocial behavior. I measured changes in attitudes, behaviors, and prosocial outcomes using both quantitative and qualitative methods to understand the effects of these interventions. These methods highlighted the unique benefits of human-AI collaboration in promoting positive social change and identified areas for improvement. I considered potential biases and inequalities in human-AI collaboration, aiming to ensure that the benefits were fairly distributed and would truly benefit society as a whole. By addressing these concerns, I developed more inclusive and effective ways to use AI for social good. The entirety of the thesis combined insights from user engagement studies, tool development, and impact assessments. The findings have implications for designers, policymakers, and researchers who use AI for societal benefit, offering new perspectives on how collaborative technologies can create positive change in our connected world.

By synthesizing insights from user engagement studies, tool development, and impact assessments, this thesis offers valuable implications for designers, policymakers, and researchers utilizing AI for societal benefit. It presents new perspectives on how collaborative technologies can foster positive change in our interconnected digital world, particularly within the domain of content moderation. Ultimately, this research contributes to the creation of more empathetic, cooperative, and altruistic online communities, paving the way for a fairer and more just digital society.