Crowdsourced accountability reporting shines a brighter light on Big Tech

"More recently, reporters have requested user-exported data from TikTok and OpenAI to answer key questions about how tech users interact and what they're shown. User-exported data contains detailed and well-formatted data based on each user's history, and is likely an artifact of compliance to data privacy laws in Europe and California. Importantly, this data allows reporters to report on real behavior on tech platforms, as opposed to creating fictitious accounts and mimicking user behavior."

"I predict an uptick in crowdsourcing-based investigations, especially for tech accountability. As access to official APIs and research agreements decreases, crowdsourcing emerges out of necessity. Newsrooms such as ProPublica and Correctiv have had teams specializing in engagement reporting and crowdsourcing for years now. Reporters have built browser extensions, set up nationally representative user panels, and shared web forms to figure out Uber and Lyft's cut from rides and how Facebook continued to recommend partisan political groups despite telling lawmakers they had stopped."

"These stories lay the groundwork of what's possible, but crowdsourcing is not without its limitations. Recruitment, non-representative samples, and the time needed to maintain these projects are unavoidable. I am hopeful that technologists will build tools to help streamline data intake. However, participation will be crowdsourcing's main bottleneck. Projects will need to be centered around addressing sources' needs to truly succeed."

Expect growth in crowdsourcing-based investigations for tech accountability as official API access and research agreements decline. Newsrooms have long used engagement reporting and crowdsourcing tools such as browser extensions, representative user panels, and web forms to uncover platform practices. Reporters are increasingly requesting user-exported data from platforms like TikTok and OpenAI; that data contains detailed, well-formatted histories likely tied to privacy-law compliance and enables analysis of real user behavior rather than simulated accounts. Crowdsourced projects face unavoidable challenges: recruitment, non-representative samples, and maintenance time. Success requires tooling for data intake, centering projects on sources' needs, and transparent communication about data use, anonymization, and risks.

Read at Nieman Lab