We are asking participants to come up with ideas and approaches for providing public accountability for targeted advertising online. The hackathon will provide participants access to existing tools and data sets as one possible starting point. Any of the following approaches fall within the scope of the challenge:
- Designing strategies and tools for making sense of the large volumes of data collected by automated ad collection. For example, we have over 300,000 ads collected by the Australian Ad Observatory, complete with metadata about the ads and information about the demographics of people who received the ads. Such strategies might focus on any of the following:
- Detecting false, misleading, and harmful ads;
- Discovering bias and discrimination in ad targeting;
- Detecting predatory advertising
- Designing tools and approaches for capturing and archiving ads and sponsored content from social media platforms, including TikTok, to see how people are being targeted.
- Improving on existing tools provided at the Hackathon, including tools that collect ads from Twitter, Facebook, and from Google searches.
- Developing future strategies for ad accountability.
Profesor Mark Andrejevic, Dr Abdul Karim Obeid, Professor Daniel Angus, and Professor Jean Burgess.
Social media platforms are transforming how online advertising works and, in turn, raising concerns about new forms of discrimination and predatory marketing. This article published in The Conversation talks about the rise of ‘dark ads’, consequences of these advertising practices, and the need for public interest research.
Robards, Brady (Primary Chief Investigator (PCI))Carah, Nic (Chief Investigator (CI))Elliott, Karla (Chief Investigator (CI))Tanner, Claire (Chief Investigator (CI))Roberts, Steven (Chief Investigator (CI))Dobson, Amy (Chief Investigator (CI))Savic, Michael (Chief Investigator (CI))
A project working with young people to examine the promotion of alcohol, junk food, sugary drinks, and gambling they see through digital platforms and apps.
Profesor Mark Andrejevic and Professor Bronwyn Carlson
The internet has provided advertisers with the ability to fly below the radar of public accountability. This is because online ads are visible only to targeted individuals on their personal devices. However history indicates that public accountability is crucial because advertisers have an established record of using harmful stereotypes and targeting vulnerable populations. This article provides examples of racist advertising and stereotyping, predatory advertising and why research is important in addressing these issues.
THE AUSTRALIAN AD OBSERVATORY PROJECT
The ADM+S Australian Ad Observatory project aims to shed light on ‘Dark Ads’, by working with members of the public, to assist in providing accountability for online advertising. This project asks participants to install an extension on their Web browser that collects all the ads they see when they go on Facebook. The extension only collects the ads (which are identified as “sponsored” content). It does not collect any personal information other than the demographic information supplied when installing the extension.
Project website The Australian Ad Observatory – ADM+S Centre
Background paper: The Ad Observatory Project
ADM+S 2022 Hackathon – Authenticated Dashboard Link
Login Credentials: Individual credentials to be emailed out to individual participants
ADM+S 2022 Hackathon – All-Time Dataset
Link to JSON files: https://cloudstor.aarnet.edu.au/plus/s/0PkEncmqiLVCSam
Link to Media files: https://cloudstor.aarnet.edu.au/plus/s/gu4hBaUmW2t8BSc
ADM+S 2022 Hackathon – Political Dataset
Link to JSON files: https://cloudstor.aarnet.edu.au/plus/s/0HR3T15VQtDtYYy
Link to Media files: https://cloudstor.aarnet.edu.au/plus/s/ZVm4LIC3APPXKcQ
ADM+S 2022 Hackathon – One Week Dataset
Link to JSON files: https://cloudstor.aarnet.edu.au/plus/s/emNiOo3ZRDdmfrM
Link to Media files: https://cloudstor.aarnet.edu.au/plus/s/3J14T27ORCubsXl
THE TWITTER AD COLLECTOR
This project allows us to track ads that appear in Twitter feeds. Unlike the Facebook ad project, however, it does not rely on user data donation. We have created dummy accounts that can be set up to follow accounts with particular sets of characteristics and collect the resulting ads that appear in the Twitter feed. For example, the account can be set up to follow and respond to accounts of conservative political candidates and columnists; or progressive ones (by liking). The ads that are collected are fed into a combined dashboard available here: http://dasdd-env.eba-vjkephmj.ap-southeast-2.elasticbeanstalk.com/ads.
The dashboard allows the ads to be sorted by accounts (“bots”), and also by the designated filters. There is room here to consider what additional filters might be added. There is also the possibility of creating new categories of accounts, and new dummy accounts.
We have an existing set of ads collected during the trail stage of the Twitter project. This data is based on four bot, two that are programmed to look as if they are based in the US and two in Australia. In each case, one bot has been programmed to follow accounts on the political right and one on the left. This information is viewable in the dashboard if you click on “bots.” The dashboard combines data from the Google bots project and from the Twitter project, so you can toggle between data sets by going to “data source” in the lower left-hand corner.
Source code for the Dashboard and Twitter bots: https://github.com/FIT4002-DASDD
Dashboard to view dataset:
Dark Ad Scraping (eba-vjkephmj.ap-southeast-2.elasticbeanstalk.com)
THE AD TRANSPARENCY PROJECT USING GOOGLE BOTS 2020
This project explores how ads might be targeted to them based on the search terms they use and the sites they visit. We wanted to see if, for example, searching for politically coded terms like “Make America Great Again,” might yield particular types of ads. We created Google profiles and used them to search on designated search terms. The automated profiles collected ads from the top 10 sites yielded by the search. We also programmed them to visit particular web sites (mainly left and right wing publications) to collect the ads they encountered on these sites.
The data collected in the pilot stage of the project are presented in the combined dashboard: http://dasdd-env.eba-vjkephmj.ap-southeast-2.elasticbeanstalk.com/ads. These can be filtered by the characteristics of the Google account (whether it was asked to visit conservative or progressive sites).
This link includes prerequisites, instructions for running the project in local, running the project in a docker container and running the project in AWS
These are links to the ads collected during the project. To access these files you will need to download ALL the files first (each file is 1GB except the last one that is 518MB). Go to www.7-zip.org to open the files.