Our grading feature lets you quickly increase the accuracy of Vervoe's predicted score for your company, by telling you which questions to grade. This targeted grading method will focus your team's manual scoring efforts on questions that will optimize your candidate's predicted scores.
Manually grading candidate responses as Excellent, Average, or Poor trains the Vervoe AI model to value the aspects in an answer that you value as an organization. The process is anonymous, so you'll only see responses and not the individual candidate names. We've also optimized the process, so only a selection of responses that the AI considers different enough is given to you to grade. With Excellent, Sufficient, Average, and Poor grades, our AI will be able to verify the types of responses you find essential and replicate them for the next time.
Once at least 20 candidates have completed your assessment, navigate to the Optimize tab.
You’ll see questions highlighted in blue as ready to grade. Clicking on a Ready to Grade question will open the question view. Once a grade has been selected the next question will be automatically loaded.
The responses you’ll be given to grade are the ones our AI would most like to learn about from you.
For example: Will you mark a response with spelling errors as bad? Or is the content of the response more important regardless of spelling?
You will notice that questions have four different status types:
Ready to Grade: There are enough responses for this question, so you can manually start grading.
Grading not Required: This applies to an auto-graded multiple choice question or one that has been marked as do not grade by the employer
Optimized: This question has already been optimized and no further grading is required
No Responses: Candidate(s) have not answered this question
Once all the "Ready to Grade" questions have been scored the AI will start optimizing. This can take up to 2 hours.
Once the first round of grading is finished, the system may require a few more responses to be graded to finish training the AI. This is when a new status, Needs more answers, will be added.
You will see the same grading panel, with the addition of the status block showing the % and number of additional questions you need to grade to optimize the assessment.
Once the model has been fully optimized, your status block will show 100% and your "Grading Required" badge will be replaced with a blue "Optimized" symbol.
What do I need to do to use active learning?
Set aside some time to train your model by grading the suggested questions. Once we know the value you place on certain responses your model will grade future answers the same. For each question, we’ll be looking for enough excellent, average, and poor grades to fully train the AI, so you’ll be presented with a series of responses from each. You may need to grade some more responses the next time you log in if there’s not enough from each of the categories.
What will happen to the candidates once I grade?
To reduce bias candidate names have been removed, so you’ll only be grading the quality of the response.
Once the optimization is complete you’ll see a variation in the candidate scores which will relate to the more accurate grading model. As always candidates will be unaware of the grade allocated to them or the additional marking. You’ll now have a list of top performers not only for job-ready skills but also the things you value as an organization!
What's the difference between using the active learning grading model and manually grading candidate's responses from their candidate card?
Manual grading improves your predicted score accuracy. Using our active learning model will reduce the number of questions you need to grade in order to reach peak optimization. It will also save you countless hours on manually grading each candidate response while allowing you to still find the right candidate. The grading feature will only surface the responses that need your review, so you won’t waste time.