A recruiter contacted me recently via email and encouraged me to apply for several entry-level roles in communications. Why communications, when my background clearly reflects a long career in HR? And why were those positions located in Georgia, when I live in New York City? I had no idea.
Curious, I replied to the email explaining that I was not looking for employment, and subsequently learned that the email had been sent by an algorithm that determined I was an entry-level millennial living in Savannah. This inaccurate profiling was formed by scans of my recent Twitter and Facebook posts, which described my new (pro-bono) role handling communications for a first-time senate candidate in Georgia.
Artificial intelligence (AI) has taken employers’ ability to identify potential hires to a more robust level than traditional applicant tracking software (ATS) that simply parses candidates’ resumes for keywords. One compelling feature of AI recruitment tools is that they help companies reach out to passive candidates whom employers would otherwise miss. Many of them do so by using chatbots that proactively introduce themselves over text, email or social media to candidates that visit an organisation’s career page, but don’t necessarily apply to a job.
But despite their potential, there are pitfalls that come with using AI for recruiting—my recent experience being one. Here are some other AI recruiting pitfalls and solutions to avoid them:
Pitfall: Developers of AI recruitment tools tout the technology’s automated objectivity as a way to reduce human bias against older workers, minorities or religious groups. Developers even claim to improve diversity by teaching AI to seek candidates from under-represented demographics. But since AI learns by perceiving patterns in past behaviours, there may be hidden biases in your company’s hiring that an AI solution will inevitably pick up.
The Society for Human Resource Management (SHRM) cautions that “if a company’s highest performers historically have been identified as white males between 30 and 40 years old—because those individuals were frequently promoted into next-level jobs—that bias can inadvertently become built into algorithms that learn from talent management patterns.”
Solution: Work with a vendor who is willing to customise a recruitment tool to avoid unconscious biases, and who will program the tool to actively diversify your candidate pool (regardless of past patterns).
Pitfall: There’s potential for AI to misunderstand insights garnered from social media. Consider the individual who posts or re-tweets something reprehensible in order to expose how objectionable he or she finds it; what AI learns from this could affect that person’s reputation.
Solution: Talk to your employment attorney. In some states, privacy laws prohibit using information found on social media from hiring decisions. Find out how to separate certain data pools such as social media, and potentially even omit them from your AI tool’s hiring considerations.
Pitfall: Recruitment tools use closely-guarded proprietary algorithms that aren’t subject to regulatory oversight, leaving HR in the dark about how the technology truly works. This can make companies susceptible to unexpected lawsuits or can hurt their reputation.
To complicate matters, multinational companies must comply with the EU’s General Data Protection Regulation (GDPR) as of May 25, 2018. This regulation stipulates how individuals’ personal data is collected, and includes information gathered from digital footprints and factors that reference, among other things, one’s cultural or social identity. Compliance with GDPR may affect how even U.S.-based companies recruit.
Solution: New York City has instituted a task force to monitor the use of AI systems. Stay abreast of similar developments in cities where your company has a presence.
There’s a great deal of promise behind AI from a recruiting standpoint, but there’s a great deal of risk as well. Before AI recruitment becomes as ubiquitous as ATS software, HR leaders will need to consider all of the challenges that come with implementing the technology—and get ahead of them.
This article originally appeared in ReWork.