Influencer marketing has become a very popular tool to reach customers. Despite the rapid growth in influencer videos, there has been little research on the effectiveness of their constituent elements in explaining video engagement. We study YouTube influencers and analyze their unstructured video data across text, audio and images using a novel "interpretable deep learning" framework that accomplishes both goals of prediction and interpretation. Our prediction-based approach analyzes unstructured data and finds that "what is said" in words (text) is more influential than "how it is said" in imagery (images) followed by acoustics (audio). Our interpretation-based approach is implemented after completion of model prediction by analyzing the same source of unstructured data to measure importance attributed to the video elements. We eliminate several spurious and confounded relationships, and identify a smaller subset of theory-based relationships. We uncover novel findings that establish distinct effects for measures of shallow and deep engagement which are based on the dual-system framework of human thinking. Our approach is validated using simulated data, and we discuss the learnings from our findings for influencers and brands.

翻译：影响者营销已成为接触消费者的热门工具。尽管影响者视频快速增长，但对其构成要素在解释视频参与度方面的有效性研究甚少。我们研究YouTube影响者，并利用一种新颖的"可解释深度学习"框架分析其非结构化视频数据（涵盖文本、音频和图像），该框架兼具预测与解释双重目标。我们的预测方法分析非结构化数据发现，"说了什么"（文本内容）比"如何说"（图像呈现）以及声学特征（音频）更具影响力。在完成模型预测后，我们通过分析同一非结构化数据源来衡量各视频元素的重要性，从而实施解释方法。我们消除了多个虚假和混杂关系，并识别出基于理论的较小关系子集。基于人类思维的双系统框架，我们发现了浅层参与度与深层参与度的独特效应。我们的方法通过模拟数据进行验证，并讨论了研究结果对影响者和品牌的启示。