Respondus is committed to the responsible use of computer vision technology. Algorithms used with Respondus Monitor are systematically tested for fairness in regard to age, gender, skin tone and other factors.

Developing Computer Vision Algorithms

The computer vision team at Respondus combines the rigor of science with the creativity of experienced practitioners. The following processes help to ensure fairness within the Respondus Monitor system:

  • Identify algorithms best suited for the unique data being analyzed
  • Train, tune and evaluate algorithms using real, extensively-annotated data
  • Use supplemental algorithms to address unique environments and outliers
  • Perform comprehensive quality-assurance testing and evaluation using a diverse team
  • Iterate the above process until a release candidate is achieved
  • When possible, publish the results to ensure transparency and interpretability

Algorithm Fairness and Respondus Monitor Proctoring

The following whitepaper provides an in-depth look at Respondus’ process for evaluating facial detection algorithms for fairness.

Algorithm Fairness and Respondus Monitor Proctoring: A study using Casual Conversations