First Quarter Updates from Tattle of the New Year! 📝

New years tend to be new beginnings for most, and that's what we had in store as well as we brought onboard two new full-time staff members - Kaustubha and Aurora!

Read on to find out about the project that kept the team churning for 3 months and other engagements!

2024 is election year

In 2024, half of the world's population are heading to the polls. With the seemingly unending rise of Gen AI, the tech landscape is abuzz with activity, and AI applications are expanding and regulations evolving. Tattle has been exploring the potential harms of this tech surge, particularly on civic agency.

In December 2023, we strategized ways to tackle harms from digitally manipulated images, realizing that detecting synthetic media requires subjective evaluation and trusted networks, rather than a foolproof tool.

Collaborating with the Misinformation Combat Alliance, we're addressing this issue through a WhatsApp tipline (Deepfakes Analysis Unit) for querying media manipulation levels, providing assessment reports for each item.

Our focus is on the collaborative platform powering the Deepfakes Analysis Unit, facilitating collective verification among experts, using our open-source media anlaysis engine Feluda (a certified DPG now), and our expertise in scalable social software.

The DAU WhatsApp tipline launched on March 25th, and we will be sharing developments from the project.

You can contribute to open issues of the collaborative platform here.

Keeping in tune with the hype of the season, we also participated in multiple events centred around deepfakes, fabricated content and their impact on the Indian elections.

You can check them out here and here.

Tech work

One of the core components of Feluda is "operators". They are modules that help us analyse media items - text, image, video, audio etc. They act as plugins that are only loaded if specified.

We added new operators for analysing audio files, images as well as one for finding similar media items.

To see the operators in action and process media items in them, workers come to our help. Workers are used to deploy operators at scale and use other core components of Feluda such as Store and Queue. A worker can help us search for similar media items too.

Currently, we have written workers for audio, video and media matching.

Along with this we also implemented security fixes and integrated new security tools for code scanning and added a security policy. We also setup release management procedure for Feluda ensuring smoother deployments, automation of releases and testing while merging code.

We also scaled our cloud infrastructure to handle large number of requests.

You can read more about our security improvements and policy here.

But that's not all there is to talk about...

Tattle participated in a roundtable consultation on the gendered nature of the information disorder in India.

We attended the roundtable held by CIS and POV in Mumbai, and had some interesting discussions whereby intersectional perspectives were brought forth in an attempt to understand the impact of gender upon information disorders.

Kaustubha also contributed to our blog with a piece on the prevalence of cross-platform abuse and the varied responses that have come up to tackle the issue. Look out for part 2.

The organisation also hosted an in-person sprint for six full-time staff for the first time! The two weeks were busy as the team ramped up work on the DAU tipline.

Finally, the DFL cohort formally wrapped up in Jan. We presented our work on the dataset expansion as well as our attempts at figuring out a compensation framework for people who contribute to the Uli dataset.

Stay tuned to further updates from us in the next edition of the newsletter!

Regards,
Yash.