Glassbox is here! Let’s open the newsrooms to the community
In mid-2020, Noé Dominguez and I [Pepe Flores] started working on the development of Glassbox, a protocol for news publishers to open the history of changes of any Wordpress-based site. The intention was to give audiences the possibility of seeing stealth edits in the news.
After a year (and a pandemic), we have achieved a stable and ready-to-use plug-in.
The development process of Glassbox gave us a deeper understanding of the dynamics of newsmaking. For the software modeling, we conducted several in-depth interviews with key informants: journalists, publishers, content creators, academics, digital security experts, and members of the Wikimedia and Wordpress communities.
Before the start of the project, we had been documenting the systematic use of stealth edits, that is, substantial changes in a text that are not informed to the readers. Now, thanks to these new conversations, we obtained a wider knowledge about the editorial practices that encourage these edits.
Our first finding was that stealth edits happen mainly during the first three or four hours after publication, usually to correct typos or minor mistakes. The text is normally edited within 24 to 48 hours since its publication, and rarely after 72 hours. Another common edit is the change of titles during the day for search engine optimization (SEO).
In breaking news, we detected a peculiar practice. Some media publish unedited press releases from news agencies, but at the same time, assign a journalist to write their own version. Once the content is ready, the media delete the press release and substitutes it with the new text. Readers are never notified about these changes.
If a piece is receiving attention, editors prefer to update it instead of dividing the coverage into multiple posts. For example, journalists tend to update a post to reflect the reaction of a source or to complement the information. These comments are usually obtained several hours after publication. Often, this happens as a consequence of the work dynamic in the newsroom, since journalists continue the coverage of other colleagues.
On the other hand, the motivations of stealth edits may be political, especially in media heavily financed by official propaganda. Authorities also suggest sources, leak information, and change titles; even call the publishers or media owners to ask them to soften the content.
Some media outlets consider that deleting certain contents is legitimate, for example, if the information is wrong. However, larger media prefer to preserve those texts and publish new posts instead. Other media outlets are ashamed of old posts, especially those who publish rumors or unconfirmed news to increase web traffic. These media delete posts to prevent people from finding them on Google.
Similarly, our research found that several media have a punitive work culture, so they were afraid that transparenting stealth edits may be used to harass journalists. The informants also warned that the visualization of changes should not interfere with their website layouts.
On the other hand, informants considered Glassbox as a useful tool to train young journalists in the newsroom, to have a better understanding of misinformation through machine-learning processes, or for academic research or digital preservation.
For the Wikimedia community, the integration of Glassbox with a Wikibot seems extremely useful for breaking news coverage. Same as the journalists, Wikipedians update articles about last-minute events, and they never know if a reference was modified.
The Wordpress community helped us understand how Glassbox could be naturally integrated with the internal dynamic of their system. We learned that Wordpress automatically preserves the last 100 revisions of a post, although those are only visible to the author and the website administrator. These revisions are consultable through queries and visible in HTML format. Thanks to this feedback, we also decided to use JSON to maximize compatibility with the Wordpress API.
All these insights were vital to model the software according to the needs of the different stakeholders in the newsmaking process.
As a result, Glassbox was designed to leverage the change log functionality in Wordpress, adding the capability of making these revisions public if active. This way, the plug-in has two main exit points: one that allows you to see the changes made to a post in the last 72 hours, and another that allows you to consult all the changes made to a certain post.
The last step in Glassbox integration with Glassbot, our bot for Wikipedia. We have chosen Pywikibot because of its compatibility with MediaWiki. The bot will check if a page used as reference has Glassbox installed; if so, it will consult if there are any changes in the last 72 hours, and will notify the community through a message in the Discussion page of the article. This complement is in the final stage of development.
For me and Noé, this has been a long but fascinating journey. We expect that both the findings of our research and the plug-in itself contribute to a proactive attitude towards transparency in digital media. Besides, our work is available under a GNU Affero General Public License because we want anyone to reuse and modify this beautiful project.
Let’s open newsrooms to the community!
