JMIR Publications

ABSTRACT

Background:

Misinformation detection and spread prediction frameworks have focused predominantly on linguistic and social characteristics of misinformation and neglected features of the information environment where misinformation emerges and spreads.

Objective:

In this study, we embraced uncertainty features of the information environment and developed a novel environmental uncertainty perception framework (EUP) for misinformation detection and spread prediction on social media.

Methods:

The framework involves uncertainty of the information environment at four scales: physical environment, macro-media environment, micro-communicative environment, and message framing. We evaluated the EUP on existing real-world COVID-19 misinformation datasets.

Results:

Experimental results demonstrated that the EUP alone obtained comparably good performance (e.g., detection accuracy = 0.753; prediction accuracy = 0.71) to the state-of-the-art baseline models (e.g., BiLSTM: detection accuracy = 0.733, prediction accuracy = 0.707; BERT: detection accuracy = 0.755, prediction accuracy = 0.728). Also, the baseline models obtained improved accuracy after cooperating with the EUP by an average of 1.98% and 2.4% for the misinformation detection and spread prediction tasks, respectively.

Conclusions:

Our findings not only suggest the efficacy of informational uncertainty in misinformation detection and spread prediction but also offer theoretical and practical contributions to the field.