ALEJANDRO CUEVAS


Work in Progress




Generative Propaganda
With Madeleine Daepp, Robert Osazuwa Ness, and Joyojeet Pal
Description
We characterize the use of Gen AI to manipulate public opinion through fieldwork in two countries characterized by their abundance of computational propaganda. 
CategoryLLMs, User Study, Qualitative, Disinformation


Why Do People Buy YouTube Accounts?
With  Manoel Horta-Ribeiro and Nicolas Christin
Description
We measure sales of social media accounts and investigate their uses after they are sold. We find that many accounts are used for potentially illicit ends, such as political propaganda or financial scams. We then develop a technique to identify similar accounts in the wild.

CategoryUnderground Markets, Mixed Methods, Social Media


Anec-Doctoring: Multilingual Multicultural Disinformation Red-Teaming

With Saloni Dash, Madeleine Daepp, Bharat Nayak, and Dan Vann
Description
We develop an AI red teaming approach to automatically generate adversarial prompts across languages and cultures. Our goal is to develop techniqes to mitigate disinformation that scale globally and realistically reflect how adversaries (mis)use models.
CategoryLLMs, Red Teaming, Disinformation


What Are Telegram Bots Used For?

With Haoxiang Yu and Nicolas Christin
Description
We collect a large dataset of Telegram bots using snowball sampling as well as other public sources. We then interact with the bots to obtain additional information about their uses. Our goal is to characterize the uses of Telegram bots, ranging from the benign to the malicious.
CategoryUnderground Platforms, Mixed Methods, Large-scale Measurement


Characterizing The Illicit Uses of Generative AI Models

With Nicolas Christin
Description
We document and characterize the emerging and evolving use cases of Generative AI models by exploring how people in underground communities use, monetize, and distribute Gen AI tools and models.
CategoryUnderground Platforms, Mixed Methods, LLMs


©Dos mil veinticuatro. Paraguayo.