Online ads are ubiquitous on the modern web. Hundreds of billions of dollars are poured into the ads industry each year to more effectively capture the attention and clicks of consumers. This domain is not a new one for researchers and past works have revealed violations of privacy, sought to understand methods for online ad analysis and their challenges, provide evidence of significant number of ads with problematic content, and much more. However, ads are neither archived or have been systematically studied on a large scale. The fleeting and programmatic nature of the ad ecosystem makes ads difficult to study, and while attempts at automated ad collection have been made, they took place within a short span of time and the resulting data is inaccessible to the broader research community.
This talk proposes a publicly available, continually updated dataset of ads crawled from thousands of popular websites and a search interface for interacting with the dataset. Automated visual analysis is employed to extract basic features from each ad, such as text or advertised brand(s), and allow ads to be organized using these features in our interface. Also proposed are some research questions to investigate and ad policy recommendations based on the tools. Overall, it is hoped that these contributions can promote accessible, interdisciplinary ad research and start the push towards higher quality, less harmful ads on the web.
CITP is running a Works in Progress (WiP) Seminar on Thursdays (unless noted otherwise) from 11:30 a.m. to 12:30 p.m.
Please visit the website for the full schedule.