The Journal The Authority on Global Business in Japan

Artificial intelligence (AI) is seeping into every aspect of our lives. From smartphone assistants and photo sorting to autonomous cars and medical diagnoses, code that is capable of adapting, learning, and im­­prov­ing itself is changing our world. If AI can help you do your job better, shouldn’t it also be able to land you that job in the first place?

The eyes that assess you at your next interview might just be driven by ones and zeroes. If that sounds like science fiction, it’s time to get up to speed on the state of AI. Since IBM’s Watson stole Ken Jennings’ crown on the US trivia show Jeopardy! in 2011, AI has advanced by leaps and bounds. Today, Watson is having a strong impact on numerous industries—including recruitment of players for sports teams.

Take that recruitment angle, apply it to the workplace, and you find a tool that could revolutionize the hiring process. Watson is just one of many AI projects, and ingenious researchers and programmers are applying the technology in diverse areas—including human resources (HR).

CHANGING CHANNELS
Smartphones, messaging apps, voice over internet protocol (VoIP), and video teleconferencing have ushered in a new era of communication. And with these technologies has come a shift in preferences. While older managers and candidates may still prefer to pick up the phone, younger generations do not. Text has become the dominant method of interaction, and it has opened the door to some clever ways for companies to automate and streamline contact with customers and job applicants.

“If we look at what is happening in the market, we will notice that communications are shifting to messaging apps, where they get automated with chatbots,” said Arik Akverdian, chief executive officer of San Francisco-based VCV Inc., developer of a recruitment tool powered by AI.

Chatbots are AI-driven programs that can hook into messa­­ging platforms and carry on conversations in ways that convin­cingly mimic human interaction. Their presence is growing rapidly, and a December 2016 survey of companies conducted by software developer Oracle found that 80 percent of respon­dents planned to use chatbots by 2020.

“In recruitment, communications with candidates is mo­­ving in this direction as well,” Akverdian explained. “The number of candidates is greater than the number of recruiters in one company, so communications have to be automated.”

The earliest chatbot may have been ELIZA, the natural-language program developed by Joseph Weizenbaum at the MIT Artificial Intelligence Laboratory in 1964. The concept has come a long way since then, and modern versions have been in use since 2013, when chatbots were added to the popular messaging app WeChat. Facebook made the technology part of its Messenger app in 2016.

With text firmly established, sound is driving the next evolution.

“We have worked in this field for several years, and audiobots are one of the most popular formats among our clients today. This is a bot that calls a person, recognizes their speech, and asks and answers questions. HR has begun to use this format not only for external candidates, but also for internal employees who use it to conduct surveys.”

Alexander Jenner, sales team manager at Computer Futures, also cited the shift in communication preferences. “We often use chat messaging services such as LINE, Facebook Messenger, WeChat, and WhatsApp to establish initial contact and set up a meeting. But we still meet face to face or use the phone to communicate for the first meeting. This is an important part of building a connection, and I don’t see that changing.”

Computer Futures is part of the London-based SThree Group, a global recruitment organization specializing in science, technology, engineering, and mathematics industries.

Once establishing a more personal connection through a phone call or face-to-face meeting, Computer Futures relies heavily on messaging for ongoing communication.

How receptive a candidate is to this approach depends not only on age but also industry. “As you might expect, people hailing from the IT industry—especially tech startups—are very comfortable using chat to communicate,” Jenner said. “And we communicate with many HR staff and chief tech­nology officers using chat services rather than phone. It’s part and parcel of the shift in how we all communicate in today’s society.”

FACE FOR SUCCESS
AI can do more than take the conversational load off HR. When combined with facial recognition technology and machine learning, it can also be used to automate the initial phase of the interview process. By analyzing how a candidate answers questions on what are essentially recorded video resumes, this combination of technology can identify the best candidates before human screeners get involved.

This is where Akverdian’s company comes in. The VCV AI-Powered Robot Recruiter, as they call it, can accomplish in 45 minutes what would take 21 hours using traditional methods. According to the company, the average recruiter spends 21 hours to select three people from 250 resumes for a face-to-face interview. But VCV uses a four-step method to run through this process in less than an hour by:

1. Scanning hundreds of thousands of resumes in minutes.

2. Contacting matching candidates with a choice of chat or phone call.

3. Using an audiobot to call candidates and explain the details of the position.

4. Analyzing video responses using facial recognition and predictive analytics.

“The idea behind this technology is to evaluate several factors,” Akverdian explained. “What does the candidate say? How do they say it? How closely do they match the profile of a successful employee?”

This is the same thing a person does when they watch a video or meet the candidate face to face. It’s a process that is often subjective. “Our task,” he added, “is to make the technology evaluate the candidate objectively and, at the same time, use a hundred more criteria than a human can use.”

TOO IMPERSONAL?
The idea of removing the element of human intuition from the process may sound alarms. There’s a reason we talk about having a “gut feeling.” As AI systems such as VCV roll out, is there cause for concern? Can technology give everyone a fair shake?

Nancy Ngou, associate partner at EY Advisory and Consulting Co. Ltd. and a member of the American Chamber of Commerce in Japan (ACCJ) Board of Governors, thinks so.

“Being interviewed by an AI, rather than a human, while impersonal resulting in more stiff responses, could provide benefits beyond time saved, speed, and consistency,” she told The ACCJ Journal. “One benefit could be that a candidate is less nervous speaking to a machine, because they can avoid judgmental human facial reactions. An additional benefit is that bias based on someone’s appearance can be eliminated.”

Back at Computer Futures, Jenner finds some potential risk in recent findings. “There are obvious privacy and bias issues surrounding the use of AI in screening, and already tests conducted by Amazon have shown that automated systems can quickly develop screening biases,” he said. “So, as a society, we do need to proceed with caution. However, I think it’s inevitable that the technology will be used more and more in the screening process.”

As for facial recognition, Ngou warned that, if used in the screening of candidates, one of the benefits of using AI could be negated, depending on the purpose. “If facial recognition is used to measure a candidate’s reactions to questions, it could be beneficial,” she said. “But it should not be used to make a prediction about the candidate based on their facial features. Hiring should be based on an individual’s skills and abilities, not their appearance.”

DIVERSITY AND INCLUSION
If AI develops a screening bias, as has been found in some tests, could its use undermine a company’s diversity and inclusion efforts?

Ngou says that some ways to mitigate these potential risks “include being transparent about the data utilized and the logic behind the decision, conducting periodic bias reviews of the data and outcomes, and, importantly, ensuring that a diverse group of people design, develop, and test the system.”

Akverdian agrees that ensuring diversity is one problem that must be solved when developing AI for screening. “It is important to understand that the system will work as it is taught, and it is very important to incorporate the principle of diversity into the training model,” he explained. “But in which situation is there a greater chance of selecting diversity? When one person evaluates another or when technology evaluates a person? I suppose it is the latter.”

That seems unlikely. The idea of intuition is a difficult one to shake, and we all like to think that we are free of bias. But that belief itself is a bias. Ngou believes AI can indeed do better than natural intelligence.

“AI can reduce the bias a human interviewer may have about things unrelated to the job or position. For example, a human interviewer may make assumptions about someone’s ability based on the candidate’s appearance, or may be biased towards a candidate who lived in the same town as they did,” she explained. “If the algorithm is written as such, it will make predictions based on the best available relevant data, not the irrelevant information a human interviewer may gather during an interview.”

At the same time, Ngou added, if the best available data is not reviewed for historical bias, AI will learn from the biased data and will exacerbate the historical bias. “If the hiring manager had historically hired individuals from their alma matter, the algorithm may favor people from that same school over other preferred schools.”

Jenner returned to his earlier statement about Amazon tests. “This is a sensitive issue, and we have already seen how AI built on machine learning inputs picks up the same biases that humans have. They are just reflecting what exists in the wider world.

“But I am sure that AI systems could be programmed to encourage diversity and reduce unconscious bias in screening and hiring, which would be a positive step.”

He acknowledges, however, that the technology is in an early stage and is controversial.

LOOKING AHEAD
How does it all come together? Christopher Reilly, director at the SThree Group, sees the tech as a way of ultimately improving relationships between recruiters and candidates rather than something that diminishes the human element.

“I hope that AI and machine-learning technology will give us, as recruiters, the ability to improve by automating time-consuming tasks such as resume sourcing, screening, formatting, and sending,” he said. “Instead of spending our time on these things, we can focus more of the consultative aspects ultimately leading to more placements for us and better experiences for our customers—whether they be candidates or clients.”

And as Ngou concluded, “With the recognition that AI is only as smart as the algorithm written, companies are looking to hire more diverse teams to better insulate the programming from inherent bias.”

When it comes to choosing the best person for the job, our artificial creations may know better than we do. At the least, they can help minimize bias and find the best matches of talent and company.

Christopher Bryan Jones is Editor-in-Chief of The ACCJ Journal. Originally from Birmingham, Alabama, he has lived in Japan since 1997.
In which situation is there a greater chance of selecting diversity? When one person evaluates another or when technology evaluates a person?