Algorithm Audits for Transparency and Accountability

During AI World Society (AIWS) conference at Harvard University Faculty Club on September 23 2019, the AIWS Social Contract 2020 was initiated and promoted by Michael Dukakis Institute for Leadership and Innovation (MDI). The AIWS Social Contract 2020 is the Principles of Politics for peace, security, and better world with deeply applied AI, Internet, in which citizens have rights to be involved directly and effectively in political decisions.

To contribute to AIWS Social Contract 2020, Dr. Christo Wilson, Associate Professor at Northeastern University and Fellow at Berkman Klein Center for Internet and Society at Harvard University, presented on an important topic about “Algorithm Audits for Transparency and Accountability”. At the digital age, the data volume is increasing exponentially, intelligent systems are increasing ubiquitous and level of sophistication is also increasing rapidly. This causes a multiple high risks including privacy violations via ubiquitous surveillance, entrenched discrimination, predatory commerce as well as biased and manipulation information curation. Therefore, the regulation and enforcement are essential from data collection (what can be collected, by whom, for what purpose) to the system capabilities (facial recognition, deepfakes, inference of sensitive attributes) for good outcomes in the society (no discrimination, unbiased presentation of information).

Within a strong expertise on security, privacy, and transparency on the web, Professor Christo developed an innovative scientific measurement technique as Algorithm Auditing. The method is developed to understand when an algorithmic system is present, data collection and sharing practices, as well as behavior of the system. In particular, Algorithm Auditing uses controlled experiments to understand whether black-box algorithmic systems are unfair or discriminatory, understand online tracking and develop techniques to improve online privacy, as well as measure and improve public key infrastructures like SSL/TLS and DNSSEC. This research has brought the fruitful results such as comprehensive “maps” of information sharing between online advertisers and trackers, incorrect prices sent to passengers by Uber, price discrimination by e-commerce and travel websites, gender discrimination in resume search tools used by recruiters, racial and gender discrimination on “gig-economy” services.

Professor Christo’s presentation contributed to an important research methodology and solution on Transparency and Accountability, which is one of fundamental components on AIWS Ethics and Practice Index. Hence, the AIWS conference is an elevated platform to share its commitment to the constructive and development of AI with a great contribution and collaboration from prestigious universities, think tanks, non-profits, firms and other entities for promoting ethical norms and practices in the digital age.

Download the presentation here