Papers, reports, datasets, and abstracts from Tattle's work on misinformation, online harms, and civic tech.
A journalistic collaboration to understand the infrastructures that enable production of non-consensual synthetic imagery.
Quantifies India’s online betting apps ecosystem by linking social media promotion to user harm
A report we worked on with RATI foundation that examines how AI-generated content, popularly known as ‘deepfakes’, is impacting and reshaping online harassment. Drawing from cases reported to Rati's helpline Meri Trustline, the report reveals a concerning trend: while media & headlines often center on celebrities and politicians targeted through AI, a more personal crisis is also unfolding. Ordinary survivors are being targeted through images that are artificially generated but possess the capacity for real harm. The incidents are rarely revealed to close family circles, let alone feature in larger discourse.
Book chapter reflecting on building an AI model for detection of gendered abuse in India by centering and working with those affected by gendered abuse, as a part of Uli.
Addressing the need for automated detection of online gender-based violence, this paper presents a dataset on gendered abuse in Hindi, Tamil, and Indian English, created through a participatory approach involving experts from the LGBTQIA community in South Asia. This paper was accepted at the Workshop on Online Harms and Abuse at NAACL 2024
A shared task conducted at ICON 2023 based on the Uli Dataset to build NLP systems that detect online gender-based violence
Reflecting on the interdisciplinary collaboration of researchers and activists in developing Uli, this paper emphasizes the integration of qualitative methods and the concept of intercurrence (an occurrence within an occurrence) to navigate the complexities and unexpected outcomes in interdisciplinary research.
Highlighting the co-design efforts of a diverse team, this essay explores the visual design process for Uli, and the creation of an alternate visual culture to challenge dominant social media narratives.
This essay discusses the development and challenges of the Uli tool, a user-facing browser plug-in using machine learning and participatory approaches to detect and mitigate online gender-based violence in Indian English, Hindi, and Tamil on Twitter.
To understand the impact of incentives on online information-sharing behavior, this study created a mock social media platform and found that both financial and social incentives increased the sharing of true information, with demographic factors such as age, education, and political ideology also playing a role.
Based on analysis of Covid-19 relief groups on WhatsApp, this report contends that the second wave in India showed a new facet of Information Disorder that challenged existing conceptions of misinformation response.
This paper presents a novel dataset that can be used to prioritize check-worthy posts from multi-media content in Hindi. It is unique in its 1) focus on user generated content, 2) language and 3) accommodation of multi-modality in social media posts. This Dataset was accepted at ICWSM 2021.
ICWSM 2021 Workshop on Information Credibility & Alternative Realities in Troubled Democracies
Presentation at the Annual Conference of Platform Governance Research Network
Report on responsible data collection from chat applications.
Abstract presentation at Workshop on Comparative Approaches to Disinformation, Berkman Klein Center, Harvard University
Presentation at Conference for Truth and Trust Online