Time to get practical! Follow along with the use case described in this section carefully, and try to put into practice the learnings gained from earlier chapters.
Here’s the situation: your colleague John has created a proposal for an AI-based solution, and you’re helping him to evaluate its trustworthiness.
Your employer is looking to hire a lot more people in the next years, and there are limited resources available, often resulting in hurried decisions and delays. Applicant screening is currently done manually and takes a lot of time.
The solution is to use AI to screen applicant resumes to automatically identify the most promising candidates to interview.
John explains that the system will be trained on historical data of the applications to various positions in your company. The model would be trained to recognize applicants similar to the ones that have been hired previously – or it could even consider their job performance.
John has heard somewhere that fairness should be considered in algorithmic decision-making, especially when related to recruitment. He asks you if bias is a concern in this particular case.
John suspects that some of the historical hiring decisions in the company might have been affected by gender, ethnicity, and other factors unrelated to professional qualifications. He suggests that the system would exclude these prohibited characteristics in the data input. That way, the decisions would be based only on skills, background, and previous employment.
You tell John that there are, however, ways to detect bias in the system.
“Okay,” John says. “But what are we supposed to do with that information? How can we get rid of that bias?”
“Okay, got it,” John says. “Sometimes an applicant asks why they got rejected. What should we do in those cases?”
“Well, is it a lot of work to get the explanations?” John asks. “Sounds quite straightforward as they’re computer-made decisions and all.”
“Oh… that sounds like more work than I thought,” John says. “Is there something else we should take into account?”
“I don’t understand,” John admits. “How would that happen in practice?”
“Okay, we don’t want to let that happen. I guess we need to take security seriously,” John sighs. “Is that all?”
“Whoa, this all is a lot to take in. How can I even get started when there are so many things to consider?” John wonders.
