CAREER: Conversational Intuition: Modeling Latent Interactional Dynamics to Improve Online Discussions

这项研究将开发计算工具和理解，可以赋予在线讨论平台关于正在进行的对话的人工直觉。虽然在线下世界中，我们已经习惯了微妙的对话信号，这些信号可以帮助我们将自己的意图和期望与对话者的意图和期望保持一致，但在网上，我们还没有对这种(通常是沉默的)元对话信号产生敏锐的敏感度。该项目将开发一个计算框架，用于(1)在面向目标的在线讨论中重建这一丢失的元对话渠道的各个方面；(2)检测错位和对话失败的警告迹象；以及(3)开发旨在保持处于风险中的对话的机制。这项工作引入了新的方法来跟踪整个对话过程中对话者的对话行为的平衡。它提出了新的假设，将这些互动平衡的动态与目标导向对话的结果联系起来，并在大规模对话数据的预测环境中和通过现实世界目标导向讨论平台上的实验干预来检验这些假设。对元对话渠道的转变理解和捕捉它的新计算模型将在许多方面产生更广泛的影响。首先，该项目将提供建立在线讨论平台的组织可以采用的新知识，以最大限度地提高其有效性。其次，它为对话的计算处理起关键作用的学科提供了新的研究方法。第三，该项目将通过发布数据集、代码和工作应用程序来增加研究基础设施。第四，该项目将开发一门新的跨学科本科课程，将研究活动产生的代码、数据和知识整合在一起。最后，将有一个广泛的外展计划，旨在通过实践互动活动增强弱势高中生的能力，突出在线团队合作的优势和潜在陷阱。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

This research will develop computational tools and understanding that can endow online discussion platforms with an artificial intuition about ongoing conversations. While in the offline world we are attuned to subtle conversational signals that help us align our intentions and expectations with those of our interlocutors, online we have not yet developed an acute sensibility for such (often muted) meta-conversational signals. This project will develop a computational framework for (1) reconstructing aspects of this lost meta-conversational channel in online goal-oriented discussion; for (2) detecting warning signs of misalignment and conversational failure; and for (3) developing mechanisms aimed at keeping at-risk conversations on track. This work introduces new methods for tracking the balance of conversational behavior of interlocutors throughout a conversation. It proposes new hypotheses connecting these dynamics of interactional balance to the outcomes of goal-oriented conversations and tests these hypotheses in prediction settings on large-scale conversational data and through experimental interventions on real-world goal-oriented discussion platforms. A transformed understanding of the meta-conversational channel and new computational models for capturing it will have broader impacts along many dimensions. First, the project will provide new knowledge that organizations building online discussion platforms can adopt to maximize their effectiveness. Second, it provide new research methods for disciplines where computational treatment of conversations plays a key role. Third, the project will add to research infrastructure through the release of datasets, code, and working applications. Fourth, the project will develop a new interdisciplinary undergraduate course that will integrate the code, data, and knowledge generated by the research activities. Finally, there will be an extensive outreach program aimed at empowering disadvantaged high-schoolers through hands-on interactive activities highlighting the strengths and potential pitfalls of online teamwork.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.