The Visibility component of our model indicates the degree to which affected individuals are aware of how models or algorithms are being used. Specifically, this refers to the end users of models and algorithms. Models with high levels of Visibility make it easy for end users to access clear information about the system, how their data is used, their user rights, and other user-relevant content.

The Integrity component of our model indicates whether the model is fair to everyone regardless of identity. In order to achieve a high score on model Integrity, various aspects of the system must meet our requirements pertaining to end user information autonomy, whether biases have been identified and mitigated, documented fair and inclusive practices, whether the longer-term impacts are positive, and more.

The Optimization component of our model indicates how well the model was built. Various factors about the model inputs and outputs impact the overall Optimization score, including the data quality used to train models, training of the models themselves, decisions made in the design process, model criteria and other factors central to model development.

The Legislative Preparedness component of our model indicates how well the algorithm and surrounding systems are prepared to meet the requirements of both current and upcoming legislation. A variety of factors impact the Legislative Preparedness score, a few of which include organizational accountability measures, documentation of company policies, adherence to standards, and awareness of the regulatory landscape.side of a div block.

The Effectiveness component of our model indicates how well the model works. This is an indicator of how well the algorithm actually functions and whether this aligns with the claimed or intended results. A model with high levels of Effectiveness will have thorough documentation pertaining to system validation, how identified issues and risks are managed, and alignment with the organization.

The Transparency component of our model indicates how clearly the system’s algorithm is understood by internal users. This encompasses factors that underlie how the model works and how to interpret model output. A model with a high score on Transparency will have a system with thorough documentation, it will be accessible to all users, and it will allow for thorough monitoring of system use, among other factors.