Could Blockchain Be a Solution to the Issue of Misinformation Online?

The New York Times R&D team built an experimental prototype to find out.

A version of this story was published in Times Insider, which delivers behind-the-scenes insights into how our journalism comes together.

When a news photo makes its journey from a camera to publication, it passes through many devices, technologies and people. Each step in this journey is designed to ensure the truthfulness and accuracy of the person or event depicted in the photo. However, once the photo makes it to the internet, it can end up anywhere in a matter of minutes — sometimes without its original context or with purposely misleading information.

The R&D team at The New York Times has been exploring a number of different solutions to address the issue of misinformation online. One of those explorations was The News Provenance Project, which experimented with blockchain technology.

A prong of The News Provenance Project was to conduct research around how readers make judgments about the photos they see online. To help with that research, we built a prototype that leveraged blockchain to surface a news photo’s contextual information on a simulated social media feed. We wanted to see whether visible contextual information, such as the photographer’s name and the location depicted in the photo, could help readers better discern the credibility of news photos in their social feeds.

Wait — why blockchain?

Blockchain allows multiple organizations to read from and write to a shared database. It tracks these changes in a log of actions, otherwise known as a “ledger.”

We wanted to see if it was possible to record everything that happens to a photo, from capture to publication, in the form of photo metadata and display that information on social media platforms. Blockchain’s ability to preserve the full history of a photo separately from the image file itself made it a compelling technology for our explorations around tracking provenance.

For our prototype, we created a private network of theoretical news organizations and a simulated social media platform that shared ownership of a database and a ledger. The news organization that were part of the network could make changes to photo metadata in the database. The ledger ensured that there was a transparent record of the time and author of the changes.

How we built our blockchain system

Our blockchain system was built by a team of developers from IBM Garage using IBM’s Blockchain Platform. To architect the system, we created a network model to ensure the right members of the network had the appropriate permissions for the metadata. We also created a data model that would help us track the necessary metadata fields to understand the history of a news photo.

The network model

The scope of a news photo’s journey is large, and there are numerous possibilities for network membership across all of its touch points. A photographer’s camera could be a network member that sends photo metadata through mobile networks into the database, with no human intervention required. A photo editor could also be a network member and their photo edits would be reflected in the metadata.

For our prototype, we focused on a simple scenario where a news organization was aware of a photo’s origin information and was responsible for publishing the photo in an article. We constructed a blockchain network that involved three network member organizations, or “nodes”: two theoretical news organizations (Local Gazette and National News) and one social media organization (which we called “the Social Media Platform”).

With these nodes, we were able to explore a few interactions on the blockchain: two different news organizations could create photo instances and update their associated metadata; the news organizations could publish the same photo in different articles; and the social media platform could access all photo history and publishing information.

In blockchain systems, interactions are written in software and stored in smart contracts. These smart contracts can additionally include checks to ensure that changes made to the database are truthful and accurate, avoiding the need for human input.

For our prototype, we kept it very simple: all possible interactions between network members were included in a single smart contract. Only one other organization needed to approve interactions and every interaction was automatically approved. For example, anytime Local Gazette wanted to create a new photo record on the blockchain, it had to get permission from National News. But, National News automatically approved the transaction without actually checking for truth or accuracy.

The data model

After conducting 34 user research interviews with people with different political perspectives, we came up with a list of contextual information that the people in our interviews found most useful about a photo, such as who took it and where it was taken.

We mapped this contextual information to IPTC photo metadata fields, which are encoded in photo files and are used in the photo industry to preserve copyright and licensing information.

This exercise helped us construct a data model for storing metadata in our blockchain database in the form of a photo record.

The people we interviewed indicated that in addition to learning about the origins of a photo, they wished to understand the publishing history of a photo, such as where the photo had been published and what captions had been written for it.

Because wire services, like the Associated Press or Reuters, often work with multiple news organizations, a single photo might be published in many different news articles. With this in mind, we created a separate publishing record in our data model that would enable us to track publishing information around a photo.

What we learned

When testing our prototype with users, we found that it effectively helped them make informed judgments about photos in a social media feed. Yet, there is more research and exploration that needs to be done.

While blockchain technology could be a viable tool in helping preserve the origin of news photos, there are a number of challenges that would need to be addressed before it could be widely used.

Content on the blockchain must be reviewed by network members.
In our prototype, we set up simple smart contracts, where changes made to a photo’s record were approved automatically. In a real-world scenario, smart contracts would be required to validate any changes made to a photo’s record to ensure credibility. Encoding those validation rules would be complex for certain metadata fields, like the description of an event in a caption.

Photos on the internet must be matched to their original versions on the blockchain.
For our prototype, we assumed the social media platform would be able to associate a photo on its feed to a photo in the blockchain through an ID. In reality, the social media platform would need to associate a photo without any knowledge of whether that photo already exists on the blockchain. The photo on social media might have been heavily edited, making it even more difficult to find its instance on the blockchain.

Bhaskar Ghosh, a student at Columbia University who conducted research for The News Provenance Project, investigated perceptual hashing and computer vision as potential mechanisms for associating photos on social media with photos on the blockchain. However, Ghosh noted that those mechanisms would require further refinement.

Blockchain networks must be accessible to news organizations.
This prototype was the R&D team’s first attempt at using blockchain technology for media provenance and it required significant operational and technical overhead. In order for a blockchain solution to become a reality, news organizations with varying financial and technical resources need to be able to participate. Finding ways to lower the barriers to entry is an essential component of any future explorations.