about this website
Artificial Intelligence (AI) along with many other technologies stands at the forefront of innovation. While its potential benefits cannot be denied, it is imperative that we handle such a powerful tool carefully. One particular application of AI is in talent acquisition. As recruiters read through hundreds of job applications and resume every day, it can be quite exhausting and time-consuming; however, AI has the capability to streamline the process while also removing any human bias. Nonetheless, if it is not implemented correctly the ramifications can be huge. This website looks at Amazon’s AI recruiting bot that showed bias against women as well as explores all the different factors that contributed to it.
Amazon's AI
Since 2014 Amazon has been building programs for reviewing job applications. This experimental AI recruiting bot was meant to weed out applicants and find the best ones. However, in 2015 Amazon’s machine learning specialists found out that it wasn’t selecting applicants in a gender-neutral way. The model was biased because when it was trained with applications submitted over the past 10 years, it developed an unintended bias towards men since the applicants were predominantly male. Whenever the model saw words like “women” on the resume, the applicant would be penalized.
The chart to the left was retrieved from Reuters. It shows the proportion of male and female employees across different tech companies. In Amazon’s case 40% of the employees are females and 60% are males. This goes to show how important it is to consider the dataset you are training your models with. Fortunately, this bot was only experimental and the tool was supposedly “never used by Amazon recruiters to evaluate candidates.”
The Problem Or The solution?
A major obstacle that AI recruiting faces is that the industry has been saturated with white males for decades and consequently the historical data that the models are built on will cause an inherent bias. Nevertheless, many companies are trying their best to hone the technology. There is a startup called HireVue which analyzes applicant’s speech and facial expressions during video interviews. Goldman Sachs uses their tool to decide which department the candidates would be the best fit for.
For AI recruiting to succeed there has to be transparency. New legislation has to be passed as there is very little oversight for AI hiring tools. This will ensure that the companies are held accountable. There are already steps being made toward this; for instance, New York passed a bill that required annual bias audits of AI use in the city. Mistakes are an inevitable part of innovation however it is important to understand the impact of such a tool. This is why companies need to tread lightly and ensure these hiring tools aren’t discriminatory.
AI
Recruitment
I hope this website has provided you with a new understanding of AI hiring tools and both their benefits and dangers.