When Tong Lin, on the verge of graduating from a college in Chengdu, capital of the southwestern province of Sichuan, received several notifications for artificial intelligence (AI) job interviews, she realized just how integral AI-driven interviews have become to the recruiting strategy of many companies this year.

Despite having experienced more than 10 AI interviews, she still finds it somewhat surreal to talk to AI recruiters, even though she's become accustomed to the general process.

"Interacting with a human recruiter offers a more engaging experience because their reactions are instinctive, which helps you adjust your behavior," she told technology news platform 36Kr. "Conversely, during AI interviews, although the interviewers look human, you're still communicating with an impassive robot, without getting immediate feedback from facial expressions and body language essential for adjusting your response in real time."

Still, she understood the rationale behind this preference for AI-assisted interviews. With a record 11.79 million college graduates entering the Chinese job market this year—increase of over 200,000 from last year—the recruitment job is especially daunting for human resource personnel.

"Many of my peers are sending out resumes en masse," she said. "This results in an overwhelming number of applications for companies, especially larger ones. AI technology greatly improves their ability to efficiently sift through the resumes and identify the best candidates."

Evolving trend

AI technology has already been applied in the recruitment process for years. In 2016, British consumer goods giant Unilever pioneered the integration of AI technology into its first round of interviews. The move, which used over 15,000 data points to evaluate candidates, ignited widespread discussion. By 2017, several companies in China had also adopted AI interview systems for their recruitment processes.

Han Shaowen, Executive Director of Beijing Avocado Technology Co. Ltd., a company founded in 2019 that enhances recruitment processes with the use of AI, explained the capabilities of these early AI systems. "At that time, the AI interviewer could already interpret human expressions, micro-expressions and voice patterns using natural language processing and computer vision technologies, along with semantic analysis capabilities," he told China News Weekly magazine.

After years of development, AI interviewers have evolved, now being able to more accurately and objectively assessing job seekers in simulated interview scenarios, efficiently screening candidates. They have advanced considerably, offering precise matches for positions by leveraging big data, he said.

Gao Yaxin, a recent college graduate from Beijing, received her first in-person interview notification in late April after having participated in five AI interviews with different companies. Before her first AI interview, she looked for tips and tricks online, and practiced extensively. Many AI systems provide two to three opportunities to record a candidate's responses, a feature that alleviates some of the stress associated with the job interview process.

"Generally, the questions asked by AI interviewers are quite similar, including a brief self-introduction, your views on the position you're applying for, and the advantages you think you can bring to the role," she told newspaper China Youth Daily. "I noticed that some people even posted scripts online for answering these questions. I wonder if this leads to interviewees giving similar responses based on similar scripts."

After several AI interviews, she became more at ease and was able to engage with the AI in a more relaxed manner. "It can cut down on time and travel expenses for job interviews," she continued. "The key is to make sure you have a stable Internet connection."

Gao went on to create a social media account to share her job-hunting experiences and insights into the AI interview process, amassing over 5,000 followers. Xiaohongshu, a super popular Chinese lifestyle and e-commerce platform, has over 50,000 posts related to AI interviews, with topic views surpassing 50 million. "Whether we embrace it or not, AI interviews have arrived and are set to become the predominant method for conducting job interviews in the future," she said.

"AI interviewers have taken over the monotonous work of initial candidate screenings in recruitment and have halved the time spent on preliminary interviews," Lin Jing, a senior human resources manager at a telecommunications company in Beijing, told Xinhua News Agency. She has been leveraging AI for first-round interviews since 2021, which has markedly decreased the demand for interviewers during campus recruitment periods.

She elaborated that AI interviewers efficiently gather, sort and classify applicants' basic information, subsequently producing evaluative reports. Companies then utilize these reports to select an appropriate number of candidates, corresponding to the number of job openings, into the subsequent round of face-to-face interviews. 

Before the use of AI interviewers, it was necessary to have a dozen or more human interviewers working in tandem; now, merely two or three suffice.

Room for improvement

"Many enterprises have just started experimenting with AI interviews this year and still need to accumulate more experience," Wang Ran, an AI scholar at Huazhong University of Science and Technology, told China News Weekly.

He noted that during the review of interview videos, candidates with above-average experience and abilities sometimes received lower recommendation scores due to poor image recognition, diverted eye contact, and unnatural expressions.

Liu Wei, Director of the Human-Computer Interaction and Cognitive Engineering Laboratory at the Beijing University of Posts and Telecommunications, highlighted the opacity of AI interview scoring criteria by referencing a vlogger's experiments.

The content creator participated in an AI interview on two distinct occasions to test the system's response to environmental changes. Initially, they chose a plain white wall as a backdrop; and during the second session, they opted for a bookshelf background. Remarkably, just this change in background led to a 15-percent increase in the AI interviewer's favorable impression of the vlogger.

"This is because AI algorithms have blind spots and lack the nuanced insight of a human interviewer," Liu told Xinhua. "Interviewees who may not excel academically might still be adept at employing tactics that can potentially deceive AI. However, these same tactics are unlikely to fool a real human interviewer."

The multiple recording opportunities provided in AI interviews can also increase psychological pressure for some candidates. One netizen, set to graduate from college this year, said online that she often needed two attempts for each question, feeling that she always performed better the second time. Consequently, what should have been a 14-minute interview (seven questions, two minutes each) ended up taking her two to three times longer to complete.

Then there are the concerns regarding privacy, with candidates questioning whether their facial recognition data and self-introductions will be deleted afterward.

"Though some industry insiders claim that data generated from AI interviews are typically not externally accessible or easily breached, relying solely on the self-regulation of companies is not enough," Wang said.

He stressed that the regulation of AI ethics lags behind technological developments, particularly in legal compliance. "As AI interviews become more widespread, targeted data security management regulations should be prioritized, and AI interviews should advance within a regulated framework," he said. BR

(Print Edition Title: You're Hired!)

Copyedited by Elsbeth van Paridon

Comments to yuanyuan@cicgamericas.com