Employers are dropping facial recognition software — what’s next?

From opening up your phone by using face ID to scrolling through an algorithmically determined social media feed, artificial intelligence is deeply embedded into our day-to-day lives – so it’s hardly surprising AI is also rapidly becoming an integral part of our work set-up.

Its prevalence is such that 83% of US companies now use AI in their HR processes, and with the pandemic driving recruitment and interviewing online, attention has turned to how machine learning might help recruiters filter potential candidates.

But the use of AI in recruitment has not been without its problems. In early 2021 leading software company HireVue announced it was rolling back software that analyzed candidates’ facial expressions in video interviews to assess their suitability for the role.

HireVue, which provides video interview software used by companies including Unilever and Hilton, decided it would no longer use facial expressions as part of its algorithmic assessment amid concerns that the process could be vulnerable to bias. Yet HireVue still believes AI has an important role to play in recruitment.

“When you consider the time constraints, budget cuts and other challenges presented to hiring teams by the COVI-19 pandemic, it’s easy to see why recruitment is one of the leading sectors to embrace the benefits of AI,” says Lindsey Zuloaga, chief data scientist at HireVue and an expert in the use of AI in recruitment.

“There are a number of AI-powered innovations such as assessment technologies and various degrees of gamification that are gaining significant momentum. For instance, work simulations and skills testing can put candidates closer to job-relevant tasks that they will need to complete in the role, giving employers greater insight into potential performance and giving candidates a chance to determine their interest in a role.”

AI before the hiring process starts

HireVue continues to use some AI technology such as Natural Language Processing (NLP) in their video interviews.

NLP uses an algorithm to make sense of human language, evaluating the competencies of each candidate based on their verbal responses to questions against the requirements of the role and organization.

When asking a standard question the technology can identify, understand, and analyze key components of each candidates’ answer, which HireVue claims can allow them to rate their response.

Yet Zuloaga believes that some of the most exciting uses of AI in recruitment could start before the hiring process even begins and extend well beyond a role’s start date.

“AI technologies are being used to match candidates with jobs that they didn’t apply for directly, but show an aptitude for based on assessments,” she says. “Even post-hire, AI can predict employee turnover and give leaders a chance to intervene and save top-performing employees from leaving.”

Chatbots, recommender engines, and resume scanning tools are currently some of the most popular uses of AI in recruitment, but companies are continuing to innovate in order to ease the workload of recruiters and help them to make smarter decisions.

“We need to reduce bias and work more efficiently,” says Kristina Angeltvedt, CEO and co-founder of Nixa.io a European-based company that creates innovative technology solutions to the challenges that HR professionals face. And AI serves both.

“In the future, I believe that all of the most time-consuming tasks like job listing, sourcing, screening, and assessment will be replaced by AI. Training AI on selecting candidates that fit your organization’s current and future needs is also going to be a significant part of how the technology develops. I think it will take time to develop reliable systems that can do this, but this will bring the biggest successes to the process.”

Data bias is still a problem

Developing reliable systems that use AI to assess candidates is certainly at the forefront of HR technologists’ minds.

As well as the abandonement of facial recognition technology, the past few years have seen a number of high-profile misses in the race to incorporate machine learning into hiring processes.

In 2018 Amazon scrapped a system that reviewed applicants’ resumes after finding that it discriminated against women, and a 2019 study suggested Facebook’s employment advertisements were skewed depending on the race and gender of the user.

In spite of an argument that AI can remove human biases from the application process, data bias is also a major problem that the industry is working hard to overcome. For HR professionals, concerns also remain about how applicant screening systems are implemented, and what data is used.

“When using AI correctly it can act as a personal assistant to recruiters – music to many of our ears!” says Zena Alana, head of talent and people operations at Superscript which regularly uses AI in its recruitment process.

“But AI applicant screening systems can also be problematic. Using AI for screening CVs is essentially ticking boxes, and we shouldn’t be putting people in boxes or basing their suitability solely on what their application says. The best way to ensure your AI is operating reliably and without bias is to only use objective data, like cognitive aptitude.”

Building trust in recruitment

Andreea Wade is an AI and machine learning portfolio director at iCIMS, a company that creates AI-driven recruitment platforms and solutions.

She believes HR technology products should match the way that consumer products such as Netflix and Amazon allow users to interact with them.

She argues that people are now accustomed to intuitive user interfaces and personalized recommendations, and expect the same level of intelligence from application processes. But she also sees AI ethics as a major hurdle, suggesting that the increased familiarity that people have with some of the problems of machine learning means that building trust in HR AI will be crucial.

“We’ll certainly see more explainable AI in the future,” she says. “As new data regulations come into play it will be important for HR professionals to educate colleagues and candidates to ensure that technologies are accepted. Candidates and employees will need assurance that new technology meets all compliance expectations, will reduce bias and treat them fairly, and will keep their personal information secure.”

Like Zuloaga, Wade believes that some of the most exciting uses of AI being developed are career pathing products that aid talent advancement and facilitate internal mobility, using recommender engines and predictor engines to match existing staff with upcoming roles. She also says that personalized machine/human engagement products such as chatbots are becoming much more sophisticated and that early diversity and inclusion product developments are showing promise.

“I’m excited to watch AI solutions advance in bridging talent gaps and rerouting candidates, coupled with efforts to drive intentional design for diversity,” she says. “These areas have huge potential for disruption.”

The disruption of recruitment processes that Wade describes offers opportunities as well as challenges. As HR integrates AI into how its employees are hired, engaged with, and promoted, building trust and eliminating biases will remain a hot topic.

The end of facial recognition interviews doesn’t represent the end of AI in recruitment, but rather a chance for companies to innovate and develop in response to their problems. The way recruiters and technologists respond could redefine hiring processes for years to come.