The Meaning in Hiring

Abstract

Artificial Intelligence (AI) has permeated every part of our world. It discovers new molecules, recommends what to watch and informs many business decisions everyday. There it has also become part of hiring, where recruiters are in need of more efficiency. The digital age has caused the amount of applications per position to skyrocket, while organisations have noticed that in the last decades an increasing amount of their value generation is directly connected to their people. Vendors market AI systems as intelligent workers that can help human resources (HR) departments find the best people efficiently. Different systems help with analytics, writing, assessments or video interviews, where participants answers questions from an AI system on their computer.

With the implementation of AI systems, there are often ethical problems involved. Biases are often hidden in the data or algorithm, that may cause people to be unfairly treated by an AI system. But AI interviews have another ethical problem that has gotten relatively little attention: autonomy over self-representation.

During an interview, you always want to show your best side and focus on your strengths and best experiences. You know better than anyone else what you can do and therefore it is important that you are able to represent yourself. But AI interviews interfere with that self-representation because they make assumptions on what you mean before they present that information to a recruiter. Also, by nature AI systems can only work directly with quantitive data, so how are you sure that your meaning of ‘teamwork’ was properly processed?

This project uses a novel approach of Value Sensitive Design in combination with
a different framework for generative prototypes to find a solution to this problem. Generative prototypes focus on generating hypotheses to further understanding. Here they were used with provocation in multiple iterations to elicit the values that people have about self-representation in hiring.

Those findings were synthesised into a new process that helps applicants maintain autonomy over self-representation through conveying feedback so they understand how well they are doing, through steering the interpretation of their answers and by keeping regular interviews to ensure the right nuance still arrives at the recruiter. This process was evaluated with another generative prototype, which informed the final three design requirements for AI interview systems in hiring: integrating feedback into the interview, managing expectations and assumptions and building in options for escalation.

In this way, the first steps are made for designing better AI systems that respect the autonomy over self-representation of applicants.