Algorithms are all around us. Based on large amounts of data and through automated decision-making processes, they increasingly exercise power over us. As a way of governing this power, we should be able to hold these algorithms and those responsible for implementing them accountable. This research posed the following research question: How can algorithmic accountability be implemented by design?

A literature review of the theory related to algorithmic accountability was executed, which led to a framework of five aspects that are required for achieving it: responsibility, explainability, debatability, auditability and sanctionability. Within a design case study, the implementation of these five aspects was investigated.

Based on the insights from research and design activities in this research, I conclude that the following guidelines will structure the process of designing accountability mechanisms:

(1) Understand the context in which the algorithm is created and employed;
(2) anticipate unwanted developments or effects of the system;
(3) design context-specific accountability mechanisms based on these findings.