Canadian company uses machine learning to promote DEI in the hiring process
Toronto-based software company, Knockri has developed an AI-powered interview assessment tool to help companies reduce bias and strengthen diversity, equity, and inclusion (DEI) in the hiring process.
Knockri’s interview assessment tool uses natural language processing (NLP) to assess only the transcript of an interview, ignoring non-verbal cues including facial expressions, body language, or audio tone. Additionally, race, gender, age, ethnicity, accent, appearance, or sexual preference apparently did not impact a respondent’s score.
To achieve “objective scoring,” Faisal Ahmed, co-founder and chief technical officer (CTO) of Knockri, says the company takes a holistic and strategic approach in training its model, including constantly trying data, training and new and different tests. , which covers a wide range of representations in terms of race, ethnicity, gender and accent, as well as occupational roles and choices. After training the model, the company performs quality checks and negative impact analysis to analyze the scoring patterns and ensure that quality candidates do not fall through the cracks.
Although he works with clients with high hiring volume like IBM, Novartis, Deloitte and the Canadian Department of National Defence, Ahmed says their model is not able to analyze all jobs in the world. . “Once we have new clients, new geographies, new roles, or even new experience levels to work with, we’ll wait to get an update on that, compare, retrain, then push the scores. We are very transparent about this with our customers.”
To ensure that the data fed into the AI is not itself biased, Ahmed adds that the company avoids using data from past hiring practices, such as reviewing resumes or successful hires. ten years ago, as they may have recruited using biased or discriminatory practices. practices. Instead, says Ahmed, the AI model is guided by industrial and organizational psychology (IO) to focus only on identifying the type of work behaviors or activities needed for specific jobs. For example, if a customer service role requires empathy, the model will identify behaviors from candidates’ past experiences and words that reflect that specific trait, Ahmed explains.
He recommends clients use Knockri early in the interview process when there is a reasonably high volume of applications, and the same experience, scoring criteria, and opportunities can be deployed for all candidates.
Ahmed says their technology is intended to help companies lay the foundation for fair and equitable candidate assessment, and is not intended to replace a human interviewer. Decisions made by Knockri are reviewed by a human being, and later stages of the interview process will inevitably involve human investigators.
“We won’t solve all your problems, but we will get you back on track,” concludes Ahmed.
