Misinformation Detection for Safe Transportation Systems

Investigator(s):

• Arlei Silva, Rice University, ORCID # 0000-0003-1792-0076 (PI)

Project Description:

This project is focused on developing a methodology for the detection of misinformation regarding transportation systems on social media. A novel dataset combining Twitter and traffic sensor data will be constructed and shared with the broader community. This will be the first large-scale public dataset for research on this topic. A comprehensive characterization of the dataset, correlating traffic events and tweets will be performed. One key question to be investigated is whether misinformation related to transportation follows similar patterns to misinformation in other domains, such as vaccines and politics. Results from the characterization will be applied to identify potential fake tweets related to traffic events. This analysis will be combined with a manual inspection to generate a set of labeled tweets that will be used for training semi-supervised methods for misinformation detection related to transportation. The proposed approach will be compared against alternatives from the literature. The main findings will be summarized in the project report and in at least one publication. Software, datasets, and metadata produced through the project will be made publicly available.

We summarize the major objectives of this project as follows :

1. Create and characterize a dataset correlating social media posts and traffic patterns
2. Propose a model for identifying and tracking the spread of misinformation on transportation systems
3. Evaluating the proposed model using real datasets.

Figure 1 Misinformation detection for road traffic content. We wil combine text-based approaches with traffic data from sensors to effectively detect fake content

Figure 1 illustrates provides a high-level view of our approach.  This project will engage in transformative research by developing a new methodology for the detection of misinformation about transportation systems. We will focus on content shared on social media (e.g., Twitter) that functions as disinformation attacks to manipulate road traffic. Analytical and simulation models for the spread of misinformation in traffic systems will be investigated. These models will combine state-of-the-art models for the spread of information and for road traffic. The final part of the project will focus on the detection of misinformation. Approaches based on content, spreading pattern, and their combination will be compared in terms of detection accuracy based on a small sample of manually labeled posts. Novel multi-modal solutions that take into account traffic patterns will also be proposed. In particular, the proposed approaches will rank potentially false or ill-intended posts in terms of their potential impact on traffic.