I will give an overview of our work assessing various interventions against misinformation and \fake news" on social media (for a review, see [7]). I will start by brie y discussing the limitations of two of the most commonly discussed approaches: warnings based on professional fact-checking, which are not scalable and which we nd may increase belief in, and sharing, of misinformation which is not agged [4]; and emphasizing publishers, which is (surprisingly) ine ective because untrusted outlets typically produce headlines that are judged as inaccurate even without knowing the source [2]. I will then focus on two more promising approaches. First, we show that most users do not want to share misinformation, but may wind up doing so anyway because the social media context directs their attention towards other, more salient factors. Therefore, we show using survey experiments and a Twitter eld experiment that shifting users' attention towards accuracy increases the quality of news they subsequently share [5, 7, 3]. Second, we show that crowds of laypeople produce judgments that are highly aligned with professional fact-checkers when assessing the trustworthiness of news sources [6, 3] and the accuracy of individual articles [1], indicating that using crowdsourcing to identify misinformation is a promising approach.
Rand