WeVerify testing ran from December 2019 and will continue until September 2021. In total, nine test cycles have been scheduled in order to evaluate the tools and single components from a user, i.e. fact checker’s, perspective.
In these test cycles, single components (e.g. Twitter Social Network Analysis Tool) and tools such as the browser plug-in that combine several components have been tested. Staff from the user partners of the project (AFP, EU Disinfo Lab and Deutsche Welle) that verify user-generated content on a daily basis have validated this technology to provide feedback on its usefulness, and also suggest improvements.
As WeVerify approaches the end of its second project year, four test cycles have been completed so far. This article outlines the results focused on three major developments of the project: the InVID/WeVerify Browser plug-in, the Database of Known Fakes, and the Professional Workbench.
Five more test cycles are scheduled for the final year.
InVID/WeVerify Browser Plug-in
The initial version of the browser plug-in was developed during the EU-funded project InVID (invid-project.eu). The tool is designed to be a verification “Swiss army knife,” helping journalists save time and be more efficient in their fact-checking tasks.
Over the course of WeVerify testing and development, the browser plug-in has been enhanced with more modules designed to facilitate the fact checker’s work. The Twitter Social Network Analysis Tool and the Digital Companion are examples of two such additions.
The Twitter Social Network Analysis Tool is an attempt to democratize social network analysis on Twitter to empower journalists and fact-checkers in their investigations and debunking tasks while tackling disinformation. By reducing complexity and providing automated visualizations that can be embedded into a fact-checking report, users can quickly make sense of content propagation on Twitter and gradually improve their understanding of social network analysis.
The Digital Companion is designed to allow non-professional users, such as private persons with an interest in verifying content on social media platforms, to undertake first steps to identify content manipulation. When using this tool, the user is guided through initial verification steps, helping them critically question whether a piece of content is what it claims to be.
During the final project year, WeVerify will go one step further by implementing the open-source micro-targeted Debunking Bot. This will help users who have encountered misinformation explore debunks, fact checks and credible content.
Database of Known Fakes (DBKF)
The blockchain-based Database of Known Fakes aims to create and maintain a consistent, distributed database of manipulated content. The web-based user interface enables journalists and verification professionals to add newly identified fake content and search for already known fakes. The DBKF is currently in its beta version and will be made publicly available over the next few months.
The Professional Workbench builds upon the commercial tool Truly which provides journalists with a working environment where they can collaboratively verify online content. It is a web-based collaboration and verification platform.
The platform offers easy aggregation of user-generated content, real time collaboration, a clear verification workflow, integrated semi-automated verification tools and shared verification tasks and information between teams. This collaboration can occur both within the same organization (such as news desks) or even across departmental and organizational boundaries (such as between multiple media organizations). The platform is continuously updated with new features, such as deepfake detection, to tackle new challenges encountered by professionals working in the field.
Feedback from Users
Feedback from those who have tested the tools and modules shows that the project is well on track. In general, users have reported appreciation for the tools’ ease of use, specifying that they require little-to-no technical knowledge and make certain aspects of analyzing social media data clear and visual.
Staff from Deutsche Welle’s social media desk see a need for increased understanding of how misinformation is disseminated on Twitter and other social media channels.
One of the most important elements of WeVerify, according to users, is that features are increasingly interlinked, making the browser plug-in a ‘one-stop-shop’ for much of their work. The Twitter Social Analysis Tool has been named a real standout with the search giving a quick overview of what is happening in specific circles of Twitter while also providing a variety of ways to analyze this information. The ability to export interaction graphs to conduct a deeper network analysis is very useful for journalists who want more technical freedom.
Users report that the Database of Known Fakes saves time and effort by creating an archive of false claims, preventing duplicate fact-checking work. The Professional Workbench helps journalists work together on more complex fact-checks.
Users have also identified a few drawbacks, which are mostly due to part of the technology not being fully operational yet. Reporting of bugs and shortcomings has provided the developers with information needed to further improve the technical solutions.
One wish users had is the ability to search links to see where they have been shared online. Alongside integration with the other tools, this addition would ensure users spend much more time with WeVerify when conducting their initial research.
The key highlight mentioned by testers is the intuitive GUI, which is key for journalists who are not tech-savvy.
In the final project year, the frequency of test cycles will be increased. More external users from other media outlets and human rights organizations will be included in test sessions to improve the quality of WeVerify’s tools and solutions.
Authors: Jordan Wildon (DW) and Tim Koch (DW)