How SeekOut Diversity Classifiers Work

Article author
Ed
  • Updated

SeekOut Diversity features are designed to help recruiting teams take action to improve organizational diversity by finding and including highly qualified, diverse candidates in the recruiting pipeline. 

Diversity Filters

SeekOut includes filters for finding Female, Black, Hispanic, Asian, and Veteran candidates. These filters find candidates who are likely members of those demographic groups.

SeekOut's Classification Methods

SeekOut has created machine learning classifiers which use a large number of attributes found on candidate profiles – including names, educational history, group membership, terms used on their profile, and others – to make a best guess of the diversity status of each candidate.

Using tagged training data, the classifier builds a model which predicts diversity status based on the attributes of the candidate's profile. We use test data to verify that the models are performing well and regularly retrain the models.

We aim to maximize precision (did we correctly match the candidate's diversity status) and recall (did we find most of the available candidates). Our current precision and recall is above 90% meaning our classifiers are very accurate and we find the vast majority of candidates for each diversity category.

Comparison with Other Approaches

Without a tool like SeekOut, recruiters typically resort to manual efforts which introduce additional bias and find many fewer diverse candidates.

A common method for finding female candidates in the US, for instance, is to run a search for one of the top 50 female first names. This approach results in a small subset of the total pool of female candidates and those that are returned are predominantly Caucasian since the top 50 names in the US are predominantly European names. By contrast, SeekOut uses the top one million names in it's female classifier (along with many other attributes), so it finds many more female candidates with much more diverse backgrounds.

Reducing Unconscious Bias

In addition to diversity filters, SeekOut provides a mode to redact candidate photos, names, and more information to reduce unconscious bias. This mode helps recruiters and their colleagues focus on candidate skills and experience.

Customization Opportunities

SeekOut's diversity filters can be customized to meet the needs of our enterprise customers. These customizations can include deploying new diversity filters for additional hard-to-find candidates (e.g., Native Americans or candidates with disabilities), customizing our existing filters, or even removing or limiting diversity filters to adhere to corporate policies.

Was this article helpful?

Comments

0 comments

Please sign in to leave a comment.