Assessing the Algorithmic Impact of Amazon Surveillance

This was the capstone project for the Human-Centred Data Science component of my Master’s degree. It was a group project in which we used the Canadian Government’s Algorithmic Impact Assessment Tool to assess Amazon’s recent Surveillance initiative for their delivery van fleet. This paper was collectively planned, and I owned half of the Results section, the Discussion & Reflections section, and the final formatting and assembly of deliverables. The abstract of this paper is as follows -

This paper investigates the impact of Amazon's decision to install AI-powered surveillance cameras in their fleet of delivery vans and provides recommendations for process improvement. The algorithmic impact assessment tool (AIAT) developed by the Government of Canada was used to evaluate the system's impact, but conflicts arose due to misalignment with private sector needs and limited access to information. Despite these limitations, the algorithm was found to have a high impact on the rights, health, and economic interests of individuals and communities, scoring 73/107 on the AIAT. The paper highlights the risks posed by the system and recommends including a human-in-the-loop, peer review, explanation requirements for decisions, and training on the applications of Amazon's algorithm to improve the system's fairness and transparency.

The final paper, along with any appendices that have their own documents, can be found here -