Machine Learning in Recruiting

In 2014, Amazon began building computer programs to review job applicants’ resumes with the aim of mechanising the search for top talent (Dastin, 2018). Clearly, Amazon have thrived under the influences of new technology, and can thank warehouse automation, automatic pricing, and people falling in love with Alexa for their rise to the top of the e-commerce industry. Recruiting remained one of the relatively untapped lines of business, so why not try and enact the same changes there?

The reason became very clear very quickly. Amazon’s recruitment mastermind had a bias against women. Now, this seems like a commercial catastrophe in this day and age, and I agree, but the reason in happened is relatively simple. It tried to automate hiring with a machine learning algorithm, but upon testing it realised that it merely perpetuated the tech industry’s bias against women. Simply put, it trained the machine on historical data, and so the historical preferences stayed true (O’Neil, 2018). Perhaps the machine works perfectly in line with historical hiring preferences, but until there is enough balanced data available, it won’t recruit effectively. It is an exciting prospect; a machine picking the people, but it seems to be a few short steps away from perfection.

Amazon focused on CV and written application screening, but that is not the only area where waves are being made. Launchpad have begun deconstructing and analysing video interviews to aid the recruitment process. Each video is transcribed and the candidate’s language and speech patterns are analysed. Algorithms can measure key personality and aptitude indicators like sentiment and lexical complexity to determine not only what it is they’re saying, but also how effectively they communicate their ideas (Hamilton, 2018). That sounds mind-blowing, but it doesn’t stop there. The system then looks at the manner in which candidates communicate. The actual audio files are analysed for things like pitch variation, pacing, pauses and other speaking qualities that could be indicative of performance level. Finally, the video recording itself is observed, looking at facial expressions, movements, eye tracking and more. In short, something as simple as raising an eyebrow a millimetre following a certain question or statement could make you more likely to be a good leader, and Launchpad are aiming to harness that power.

Looking at the recruiting objectively, it seems like automation is the way to go. With an accurate, wholly representative training dataset, the time and cost needed to develop a hiring machine is sure to pay off. It is fascinating to see how much money company’s are spending on recruiting, and that if that can be curbed in one fell swoop with substantial investment, I struggle to see why it shouldn’t be. In the future, perhaps the same amount of information could be learned from one digital interview that requires no staff or resources to host as from a full 12 hour assessment day; that is a win for all parties involved.

Regulatory Changes

With the introduction of the EU General Data Protection Regulation (GDPR), recruiting changed overnight. As the focus on the technology side of recruiting increases, it is imperative for companies to remain within the realm of the new laws. How they gather, store, and most importantly use applicant’s data is now scrutinised in much greater detail (Girdler, 2018). The most important considerations relevant to recruiting are as follows:

1. Companies need to have a legitimate reason to process data. This one is rather straightforward, but it means that data collected must be relevant to the job posting. It also means a recruitment decision must be made in a timely manner.
2. Consent, consent, consent! A recruitment company must have explicit consent from the applicant to process their data. This will become more relevant if applications are being screened and compared through a machine that actively learns from the applicant’s data.
3. Companies need to be completely transparent about their data processing. If an applicant is not being considered for the position but their information is being used to train the model on who not to hire, the candidate must be informed (Bika, 2018).

The introduction of this new regulation may be a thorn in the side of technological advancement in the recruiting industry for some time, but it remains to be seen how companies are dealing with it.

Targeted Advertising and Postings

The final area the world of recruitment is changing that I will discuss is in advertisement and targeting postings and listings. The power of Google and other advertising giants can be used to ensure that the right people are being exposed to the right listings. If I have an interest in working in a hedge fund, there is little to no point in paying to expose me to an advertisement for a listing in a nursery. By analysing cookie data and search and internet usage history, you can discover a persons interests. This targeted advertising extends the reach of the job posting and can lure candidates who aren’t actively looking for an open position on a job board. Rather than investing substantial time and money to plan an advertising campaign, design ads, and buy media, the employer simply signs up to have the targeted ad generated from their job opening (Rossheim, 2018).

If this is combined with techniques from above, a company could develop a checklist of traits and characteristics that make a good hire, and target people who have usage history that is indicative of the same characteristics. Thus begins a matching game on a global scale. If the full power of technology could be used in recruiting, it could transform it from an exhausting back and forth conversation into an exciting game of cat and mouse, and once again everyone is happier. There is no doubt significant investment would be needed, and perhaps the technology is still too infantile to be worthy of such investment. However, there is no doubting that the day where a fully automated recruitment process is the norm is nearly upon us.

Devin Connolly
Novus Opus

References

Bika, N. (2018). GDPR compliance guide for recruitment | Workable. [online] Recruiting Resources: How to Recruit and Hire Better. Available at: https://resources.workable.com/tutorial/gdpr-compliance-guide-recruiting [Accessed 27 Dec. 2018].

Dastin, J. (2018). Amazon scraps secret AI recruiting tool that showed bias against women. [online] U.S. Available at: https://www.reuters.com/article/us-amazon-com-jobs-automation-insight/amazon-scraps-secret-ai-recruiting-tool-that-showed-bias-against-women-idUSKCN1MK08G [Accessed 27 Dec. 2018].

Girdler, S. (2018). The 4 trends changing recruitment, and the opportunities that they provide for background screening | Onrec. [online] Onrec.com. Available at: http://www.onrec.com/news/opinion/the-4-trends-changing-recruitment-and-the-opportunities-that-they-provide-for [Accessed 27 Dec. 2018].

Hamilton, W. (2018). Predicting Right-Fit Hires: The Impact of AI and Machine Learning on Recruitment. [online] Launchpadrecruits.com. Available at: https://www.launchpadrecruits.com/insight-articles/ai-machine-learning-recruitment [Accessed 27 Dec. 2018].

O’Neil, C. (2018). Bloomberg – Amazon’s Gender-Biased Algorithm Is Not Alone. [online] Bloomberg.com. Available at: https://www.bloomberg.com/opinion/articles/2018-10-16/amazon-s-gender-biased-algorithm-is-not-alone [Accessed 27 Dec. 2018].

Rossheim, J. (2018). Technology’s Impact on the Recruiting Landscape | Monster.com. [online] Monster Hiring Resource Center. Available at: https://hiring.monster.com/hr/hr-best-practices/recruiting-hiring-advice/attracting-job-candidates/new-recruiting-strategies.aspx [Accessed 27 Dec. 2018].