To Explain or Not To Explain: An Empirical Investigation of AI-based Recommendations on Social Media Platforms

Abstract

Abstract Artificial intelligence integration into social media recommendations has significant promise for enhancing user experience. Frequently, however, suggestions fail to align with users’ preferences and result in unfavorable encounters. Furthermore, the lack of transparency in the social media recommendation system gives rise to concerns regarding its impartiality, comprehensibility, and interpretability. This study explores social media content recommendation from the perspective of end users. To facilitate our analysis, we conducted an exploratory investigation involving users of Facebook, a widely used social networking platform. We asked participants about the comprehensibility and explainability of suggestions for social media content. Our analysis shows that users mostly want explanations when encountering unfamiliar content and wish to be informed about their data privacy and security. Furthermore, users favor concise, non-technical, categorical representations of explanations along with the facility of controlled information flow. We observed that explanations impact users’ perception of the social media platform’s transparency, trust, and understandability. In this work, we have outlined design implications related to explainability and presented a synthesized framework of how various explanation attributes impact user experience. In addition, we proposed another synthesized framework for end user inclusion in designing an explainable interactive user interface.

Ähnliche Arbeiten