ORCAA is a consulting company that helps companies and organizations manage and audit their algorithmic risks.
ORCAA audits existing algorithms for accuracy, bias, consistency, transparency, fairness and timeliness, following a 4-step approach. First, a deep dive into the high level design of the algorithm, including the assumptions on the type, quality, and quantity of data, the nature of data pre-processing, the algorithm itself, the structure and assumptions around the objective function and the optimization to that, the way the algorithm was trained, and the approach taken to monitor and update the algorithm. Second, matching the design to the implementation; this is done at the code level if it’s available and via black box testing if not. Third, the execution audit, in which we kick the tires of real-world constraints of the algorithm. In the fourth step we write up the process and include a “to-do” list of improvements and suggestions.
Company Type: Enterprise
Region: US & Canada
Industry Category: Consulting Services
Fighting Type of Bias: Algorithmic bias Data-driven bias
Product: Algorithmic Auditing & Training
