Bias Prevention and Responsible AI

Workforce data and practices are full of bias.

Visier’s mission is “To reveal the human truth that helps business and employees win together.”. This includes being sensitive to and exposing bias where it exists.

Workplace data contains bias, as shown by decades of academic research¹. Visier measures such bias periodically and has published research reports on this topic for nearly a decade². The data for this work comes from the organizations that use the Visier solution, containing over 20 million employee records from more than 10,000 organizations.

Visier products are built with bias mitigation in mind.

Visier partners with reputable providers of third-party Large Language Models (LLMs) to facilitate the functions of our AI tech systems. For example, LLMs allow Vee to translate a natural language input into prompts that Visier People understands or a natural language output back to the user in direct response to a user’s chat input³. These LLMs leverage bias mitigation techniques such as specific data acquisition and pre-processing, algorithmic fairness and transparency, and monitoring and evaluation⁴.

Vee is taught Visier’s proprietary Blueprint, which has been specifically curated to address employee populations in the same manner as Visier People. Vee’s training data is not generic, and aggregate data is not used to train Vee.

Visier products are designed to support organizations, helping them to discover bias in their workforce and enabling people to make better informed decisions, which will reduce the risk of bias perpetuating.

• The Visier solution is built to provide insights into the workforce. Exposing and measuring bias is a core competency for Visier products.
• Viser enables customers to become active in mitigating bias in their organizations.
• The Visier solution provides data-driven insights to executives and human resources professionals to help them make effective business decisions related to their employees and candidates.
• The Visier solution is highly configurable by customers, allowing them to implement their own compliance requirements and organizational policies around bias
• The Visier solution includes deep documentation about how it works, so that users have full transparency about the insights they are getting from the solution.

Customer Responsibility.

Customers may configure Visier models with data attributes of their choosing, i.e. tenure, direct manager, age, etc. Therefore customers can specifically configure the models, improve accuracy and reduce biases as needed.

Each customer is responsible for confirming that its source data is free from bias.

Footnotes