Collaborative Research: Social Media Co-Pilot: Enhancing Teens’ Digital Literacy and Cyber Safety Education with AI-based Conversational Intervention

在线环境中网络风险的普遍存在要求进行数字扫盲和网络安全教育，以预防和减轻可能对青少年造成严重和持久后果的危害。基于人工智能（AI）的会话代理在提高数字素养和网络安全教育方面具有巨大潜力，因为它们可以提供个性化，动态和即时的学习指导和反馈。这项研究的目标是通过社交媒体共同试点教育青少年在互联网上的安全，负责任和道德行为，这是一种基于人工智能的创新对话干预，嵌入在互动和现实的在线学习环境中。通过动态分析学习者的输入，Co-Pilot将识别“可教时刻”，并提供个性化的会话反馈，以加强亲社会行为，并阻止在线的危险行为。有能力根据情况承担不同的会话角色（例如，更有知识的同伴、需要立即帮助的同伴、成人导师等），社交媒体合作试点将引导学习者在面对常见的社交媒体风险和陷阱时进行更深入的反思和负责任的行动，例如在线攻击，侵犯隐私，网络钓鱼和诈骗。这一项目将使所有青年受益，但也将为弱势和得不到充分服务的社区提供额外的、有针对性的支助，从而缩小数字鸿沟。青年的数字素养和网络安全技能的提高可以为STEM素养奠定基础，并提高他们追求STEM职业的准备程度。PI将采用尖端的自然语言处理（NLP）技术来开发驱动社交媒体Co-Pilot的会话代理。他们将与青少年和教育工作者一起反复完善社交媒体联合试点，他们的形成性反馈将为工具的逐步改进提供信息，并提供个性化的现场对话和对各种网络安全挑战的批判性反思。Co-Pilot将使用大型语言模型进行开发，其中包括来自不同种族和民族的青少年和教育工作者的广泛投入，以确保其与不同学习者的相关性，以及其有效性和可扩展性。一旦建成，PI将通过随机现场实验评估社交媒体联合试点对青少年网络安全知识和行为结果的影响。通过利用教育理论来改进社交媒体合作试点的设计和评估，该项目将为基于AI的会话代理作为体验式学习工具的有效性提供新的知识。它还将促进对以人为本的人工智能设计的理解，以及在教育领域定制大规模NLP模型的新方法，同时确保设计和应用的可靠性，透明度和公平性。该奖项反映了NSF的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估而被认为值得支持。

The prevalence of cyber risks in online environments calls for digital literacy and cyber safety education to prevent and mitigate harms that can lead to serious and long-lasting consequences for teens. Artificial intelligence (AI)-based conversational agents have great potential to improve digital literacy and cyber safety education because they can provide personalized, dynamic, and immediate learning guidance and feedback. The goal of this research is to educate teens on safe, responsible, and ethical conduct on the internet through Social Media Co-Pilot, an innovative AI-based conversational intervention embedded in an interactive and realistic online learning environment. By dynamically analyzing learners’ input, the Co-Pilot will identify “teachable moments” and offer personalized, conversational feedback to reinforce prosocial behaviors and discourage risky actions online. With the capacity to assume different conversational roles appropriate to the situation (e.g., a more knowledgeable peer, a peer needing immediate help, an adult mentor etc.), Social Media Co-Pilot will guide a learner into a deeper reflection and responsible actions when confronting common social media risks and pitfalls, such as online aggression, privacy violation, phishing, and scams. This project will benefit all youth, but will provide additional, customized support to vulnerable and underserved communities, thus reducing digital divides. Youth’s improved digital literacy and cyber safety skills can provide a foundation for STEM literacy and enhance their readiness to pursue STEM careers. The PIs will employ cutting-edge natural language processing (NLP) techniques to develop the conversational agent that drives Social Media Co-Pilot. They will iteratively refine Social Media Co-Pilot with teens and educators, whose formative feedback will inform incremental improvement of the tool and provide personalized in-situ conversations and critical reflection on various cybersecurity challenges. Co-Pilot will be developed using large language models with extensive input from racially and ethnically diverse teens and educators to ensure its relevance to different learners, as well as its effectiveness and scalability. Once built, PIs will evaluate the effects of Social Media Co-Pilot on youth’s cyber safety knowledge and behavioral outcomes through a randomized field experiment. By leveraging educational theories to inform improvements in the design and evaluation of Social Media Co-Pilot, this project will contribute new knowledge about the effectiveness of AI-based conversational agents as experiential learning tools. It will also advance the understanding of human-centered AI design and novel ways to customize large scale NLP models in the education domain, while ensuring reliability, transparency, and fairness in design and application.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.