Explainable AI for Talent Management
Sep. 2020 - Dec. 2020, 3 months
Jiajun Dai (UX Designer) / Yi Wang (Data Analyst) / Dianne Kim (UX Researcher)
UX&UI Design / User Testing / User Interview
We build an AI model to predict employee attrition. Such model would help HRs better manage talents and define strategies to reduce attrition costs. However, most HRs don't have machine learning backgrounds so they are likely to have misunderstanding and trust issues on the AI prediction results. By adopting principles of trustworthiness, transparency, and ethics and testing several explanation options with real users， I designed the interfaces to help company management without ML backgrounds to better understand the prediction results and take actions to retain talents.
How might we explain the model in a trustworthy and non-expert-friendly way to help HRs fairly and effectively manage talents?
Team Attrition Analysis
Show the history and the trends in team attrition, thus helping HRs notice potential risks in an early stage
Personal Attrition Analysis
Integrate AI attrition prediction with employee work performance to help HRs effectively see the 'why' behind the model
Improve the explanation transparency and fairness by displaying intuitive figures and using plain language, helping HRs understand the 'how' behind the model
Customize what-if scenarios to explore potential solutions to retain talents and further understand the model limitations