Bookmark Ctrl+D Save this page as a bookmark to fully understand the latest information, which is convenient and fast. You can also download desktop shortcuts. Click Download | Sina Technology | Sina homepage | Sina Navigation

Is AI interview reliable?

   Welcome to follow the WeChat subscription account of "Sina Technology": techsina

Wen/Tao Tao

Source: combustion dimension (ID: chain)

"Please share an experience of achieving challenging goals" "Please share one of the most complex emergencies you have ever dealt with" "Please share a market analysis you have ever done"... This is the experience of Qiu Qiu, a new undergraduate majoring in finance, who participated in an autumn recruitment interview in September.

In the face of rigid voice and template questions, the AI interview system without eye contact but with a time limit for answering questions, Qiu Qiu, who has never experienced an AI interview, is a little confused.

"The company I applied for is Anta, a sports brand. During the interview, each question is limited to 6 minutes. During this time, I need to record the answer first, then convert the system voice into text, and smooth the text answer. But when I open my mouth, my mouth is full. Compared with real people, it is difficult to say neatly when facing a blank."

In fact, it's not just autumn. This autumn, a large number of new graduates have to deal with the AI interview when applying for jobs. On the Little Red Book, many young people share their experience in participating in AI interviews, including "AI interview admission rate", "AI interview routine", "AI interview skills", etc. The AI interview enterprises involved also cover the fields of FMCG, foreign capital and banking.

 Figure/AI Interview related notes source/screenshot of Xiaohongshu Figure/AI Interview related notes source/screenshot of Xiaohongshu

"In the past two years, a large number of enterprises will use AI recruitment in the first round of interviews, because the external environment has changed." Angela, who once worked as HR of a FMCG brand, told Burn Times, "The epidemic has become a physical factor that plagues recruitment, and the imbalance in the supply of the job market has led to the accumulation of a large number of resumes, while HR has no time to screen."

At this time, the AI interview, which breaks through the time and space constraints of traditional recruitment and reduces the cost of manpower and time, becomes a "favorite". The main "AI Talent" of station B's UP mentioned in his article that in the small-scale interview scenario, AI talent intelligent recruitment can save at least 65% of the interview time for each round of interviewers; In large-scale interview scenarios, the time of the interviewer can be saved by more than 95%.

In addition, the advantages of AI interview are relatively objective and fair. "Using AI interview can avoid the subjective factors of recruiters interfering with the interview results too much and reduce such factors as appearance discrimination and regional discrimination, which can be set in the algorithm," said Tian Juanjuan, executive director of Zhilian Recruitment Evaluation Research Institute.

Or because of this, the global AI interview market is expanding. According to the data of Facts&Factories, the global AI interview market will reach 610 million dollars in 2021, and is expected to reach 890 million dollars in 2028, with a compound annual growth rate of 6.5%.

However, the possible drawbacks of AI interview are also worthy of attention.

   Some insiders said that the artificially set AI system is a double-edged sword, which can not only increase objective factors, but also strengthen prejudice At the same time, AI systems at the initial stage of intelligent learning sometimes even form AI interview bias involuntarily according to "self judgment". On several social platforms, Ran found that many new graduates who had participated in AI interviews were suspicious that they were dismissed because of their lack of education or gender discrimination.

"In my opinion, the information asymmetry of the AI interview is the main reason for the interviewer to question." Liu Wei, director of the Human Computer Interaction and Cognition Engineering Laboratory of Beijing University of Posts and Telecommunications, said to Ran Siyuan, "AI interview manufacturers should not only serve the human resources department of enterprises, but also open up the communication channel between candidates and AI interview system."

In addition, the lack of insight in AI interview and the poor sense of interview experience brought to the applicants are also the problems that AI interview was talked about.

Liu Wei said that AI is still difficult to reach the human level in terms of cognitive level, and the future development trend of the recruitment industry should be human-computer integration.

   Interviewer trapped in AI system

Even without experience, Qiuqiu was not surprised to learn that she was dismissed by Anta's AI interview.

"Interviews that lack eye contact are not easy to adapt to. In addition, AI interviews usually examine the personality characteristics of interviewees through their looks and tone, which makes me more uncomfortable in interviews." Qiuqiu said when referring to AI interviews.

Also experienced the elimination of AI interview, Lucy, a law master who just graduated this year, "The AI interview that brushed me off required only one minute to answer each question. Time was very tight, unlike in the real person interview, there was no time limit, so I was brushed off before I could finish answering."

Some netizens on the social networking platform also complained, "After the interview with AI, I was thin and broken, my mouth ran in front, and my mind was chasing after me, so I took it!"

In addition to being nervous and time pressed, the candidates also complained about the visual effect of the AI interview. "Except for the candidates with high appearance, most of them are distorted and ugly on the AI interview platform," Lucy told Ran Ziyuan.

Lucy also said that according to her understanding, AI interviews focus not only on the candidates' words, looks, appearance, but also the surrounding environment

   The technical problems involved in the AI interview system are also complained about by many students Lucy Tell Ran Siyuan, "Some AI systems are not smooth enough and are easy to get stuck, which will directly affect the scores of interviewees. Therefore, for interviewers, at least find equipment with stable network and good radio effect to complete the AI interview, and try to avoid problems caused by hardware."

   However, the "routine" questions raised by AI interviewers during the interview process will gradually "mechanize" the answers of candidates after a large number of interviews ”。

In this autumn's recruitment, Lucy participated in interviews with 10 companies, three of which were AI interviews. Qiu Qiu, who has participated in 30 interviews, has experienced 7 AI interviews, involving enterprises in different fields including Carlsberg, Kimberly Clark, Uniqlo, Shanghai Rural Commercial Bank and so on.

Qiu Qiu told Ran Ziyuan that many AI interview questions for FMCG brands and banks are similar, similar to "What difficulties have you encountered in your work and how did you solve them?".

On the social platform, many candidates also said, "Have you ever had communication difficulties in your team collaboration? How did you solve them? What was the result?" and other routine questions are often encountered in AI interviews. After getting familiar with the process, prepare a template answer to the regular question, recite it carefully, and you may pass the exam.

 Figure/AI Interview Template Source of Questions and Answering Skills/Screenshot of Little Red Book Figure/AI Interview Template Source of Questions and Answering Skills/Screenshot of Little Red Book

Qiu Qiu, who has gradually found a pattern, said that after preparing some AI interview question banks of FMCG brands, banks and other enterprises shared by netizens on Xiaohongshu, she also slowly summarized some experience in answering questions. "With these skills, I really don't fear the camera anymore."

"After having experience, I passed the other six AI interviews successfully. But at the same time, I feel more and more like a machine repeating answers." Qiu Qiu said helplessly.

It is understood that Qiu Qiu has experienced three different AI forms: voice answer, AI virtual person image question and answer, and HR dubbing in the entire seven AI interviews.

"I remember very clearly that the company of HR dubbing was Standard Chartered Bank Shanghai Branch." Qiu Qiu said to Ran Ziyuan that the diversity of AI interviews is beyond his imagination, "but I prefer real person interviews. AI interviews always make me feel that there is a lack of humanism in recruitment."

   What are the advantages of AI interview?

Although the interviewees complain a lot about the AI interview, in recent years, the AI interview has been favored by enterprises.

   Behind this, or because AI interview can reduce costs and increase efficiency, be free from physical constraints, investigate multiple contents, and reduce artificial recruitment bias

In fact, the reason for the initial birth of AI interview is its high efficiency and low cost of recruitment.

AI interview first started in 2009. At that time, the financial crisis swept the world, many of the world's top 500 enterprises began to large-scale layoffs, and the job market competition intensified. A large number of jobs are in short supply, but enterprises are forced to reduce costs and increase efficiency, forcing the Human Resources Department to complete recruitment with as few manpower as possible. AI interview came into being.

However, with the economic recovery and the slowdown of the market demand for AI interviews, AI interviews have not been rapidly popularized in the world in the next 10 years.

However, since 2020, under the influence of the epidemic, especially the arrival of the post epidemic era, AI interviews have come to the front again due to the cost reduction and efficiency increase needs of the Human Resources Department.

Zeyu, head of L'Oreal (China) campus recruitment, said that L'Oreal introduced the "HRTPS" AI text analysis system in the recruitment of trainees in 2020. And HRTPS can just meet the needs of efficient and diversified primary screening. "

The relevant person in charge of Liepin's AI interview product "Multifaceted" also revealed that the multi-faceted AI interview can quickly conduct more than 4000 interviews within 24 hours, which is about equal to the workload of three interviewers working 48 hours non-stop.

It is worth noting that many AI interview products, including many others, can intelligently answer some questions often asked by applicants, such as "enterprise treatment", "enterprise office location", "company benefits", etc, "These questions are customized by recruitment companies. Our virtual interviewers answer questions, so that candidates can complete their initial understanding of the enterprise at the AI interview stage, and can also play a role in saving HR interview time in the future." said many relevant persons in charge.

   At the same time, the benefit of AI interview is that it can have more dimensions and more detailed investigation on candidates

Tian Juanjuan introduced that, taking the AI interview product "Yi Mian" of Zhilian recruitment as an example, it can capture the post quality of the four modules of hard skills, soft skills, personality and motivation; In terms of personality, they can capture sophistication, tenacity, and agreeableness; Motivation includes internal drive, professional identity, etc. "

Previously, 36 Krypton reported that HireVue, the first American company to enter the AI interview track, could even get 500000 data points about candidates from 15000+different dimensions (including body language, language mode, eye movement, question speed, voice size, etc.) after a 30 minute interview.

Zeyu also mentioned that L'Oreal AI interview mainly focuses on whether candidates have the five major abilities of learning ability, empathy, judgment, resilience and career ambition. HRTPS can just evaluate talents from these five major potentials

   In addition, AI interview can also reduce the judgment bias of artificial recruitment to a certain extent

In addition to the single dimension bias factors such as appearance and region, the person in charge of multi-faceted AI introduced that multi-faceted AI would collect the diversified information of the interviewees and make a comprehensive evaluation on this basis, "For example, for a junior high school geometry exam, A candidate answered 70 points, Candidate B answered 80 points, but A is a grade 6 primary school student and B is a senior high school student. Then we will combine two factors to reach a fairer judgment. "

   Is AI interview reliable?

Because of its advantages in recruitment, AI interview has formed a certain market scale.

According to the official data of Haina, a domestic AI interview company, its intelligent recruitment solution integrating "AI interview+online exam+video interview+technical contest" has served nearly 5 million online interviews and exams for more than 5000 companies, including Wal Mart, Shunfeng, ZTE, and China Merchants Bank. Another group of data from Zhilian shows that as of July 2022, among the projects that have covered nine industries and more than 500 enterprises, the number of interviews has exceeded 150000.

However, according to incomplete statistics of Burning Time, at present, in addition to HireVue, Haina, Zhilian and Liepin, the representative enterprises involved in AI interviews also include Interviewer.AI, Beisen, Quanfang, Saima, Niuke, Moka, daydao, etc.

Although the interview efficiency is the most praiseworthy aspect of AI interview, the reliability of AI interview is also questioned by many people. For example, the lack of human judgment, the lack of targeted interview questions for enterprises and individuals, and the existence of unconscious or conscious bias.

In recent years, there have been many cases of AI interviewers "collapsing" due to lack of human judgment.

According to the quantum bit report, a blogger once tried to participate in an AI interview. The first interview used a white wall background, the second one used a bookshelf background, and the rest of the interview process had no variable factors. Surprisingly, with the addition of the bookshelf background, AI interviewers' preference for bloggers has suddenly increased by 15%. In addition, another woman just changed the variable of whether she wore glasses or not, and the score of the AI interview system showed a difference of up to 10 points.

"This is because AI algorithm has blind spots and lacks real insight. For example, interviewees like those mentioned above may not have high academic achievements, but they are good at finding other methods, so they may cheat AI. However, it is relatively difficult to cheat real interviewers through these methods." Liu Wei said to the analysis of combustion dimension.

   Another commonly mentioned disadvantage of AI interview lies in the routine and lack of pertinence of interview questions

Angela said, "If different enterprises use similar AI interview templates, it will be insufficient for the matching between the company and the position. Similarly, for applicants, when communicating with HR, the two sides trigger a dialogue based on the information exchanged, rather than conducting the interview in a one-way output based way, as in AI interviews."

 Figure/Liepin Multi dimensional 3D Digital Interviewer Tao Sirui Source/Liepin Figure/Liepin Multi dimensional 3D Digital Interviewer Tao Sirui Source/Liepin

   Conscious systematic bias is also widely questioned in AI interviews

Qiu Qiu and Lucy both said to Ran Ziyuan that not only they but also their friends would fill in their personal name, gender, age, education and other basic information before entering the question and answer system when attending the AI interview. "Although there is no relevant evidence, my friends and I suspect that AI interviews mainly use this information to operate and analyze, and then screen out people with low academic qualifications. The success of the interview is not related to the performance of the interview." Qiu Qiu said.

On the Little Red Book, a netizen wrote in a post related to the AI interview that Hisense "stuck" in the AI interview stage, which seemed to be the degree, as well as the proportion of job demand for men and women. Neither the boy nor I can speak English very well, but the other girl can speak English very well. But the final result of the interview was that the male student was admitted. "

 Figure/Some netizens think there is prejudice in AI interview Source/screenshot of Xiaohongshu Figure/Some netizens think there is prejudice in AI interview Source/screenshot of Xiaohongshu

"The reason why this question arises is that in the AI interview system, AI interview manufacturers and recruitment enterprises are deeply involved, and the candidates are in an information blind spot, so they don't know what the real screening criteria for AI interviews are." Liu Wei said to the dimensional analysis of fuel, "The way to solve this problem can refer to the relationship between electronic referees, athletes and coaches on the sports field: the referee and electronic referees jointly judge, but athletes can challenge the decision of the referee by looking back at the camera. In this referee system, electronic referees are transparent to the referee and athletes."

In addition, because AI's intelligent learning ability is still in its early stage, there is also unconscious bias in AI interviews.

According to Red Star News, many years ago, Amazon's self-developed "AI interviewers" did have an obvious tendency to "prefer boys to girls" during the resume screening stage. More paradoxically, this is not Amazon's original intention. In order to develop AI interviewers, Amazon's research team created 500 computer models, and then taught each model to recruit future candidates according to about 50000 words in the resumes of former candidates. However, with the learning of this model, this AI engine has automatically learned to discriminate against women.

Red Star News also mentioned that this is because Amazon's AI interviewers have gradually learned to ignore the common skills of job seekers, such as the ability to write code. Instead, they prefer self descriptions such as "execution" and "capture" in the resumes of male engineers. Later, Amazon found out that the AI recruitment system did not conduct gender neutral assessment for the recruiters, and it itself concluded that male candidates were more desirable. Of course, the AI interview system was abandoned later.

It is not difficult to see that at this stage, there may be many blind spots in the "self acquisition" of AI interview system in analysis and judgment. These blind spots are likely to let some potential candidates miss opportunities.

However, it is worth mentioning that AI interviews are also constantly trying to approach thousands of real interviewers.

For example, Liepin has systematically optimized many aspects and introduced intelligent questioning in the AI interview.

"Intelligent questioning will analyze the answers of candidates in real time, and then choose an appropriate direction to ask questions, so as to better tap the potential of candidates." Multiple principals introduced that candidates can also gain different experiences in AI interviews at the same post.

In addition, some AI interview systems and recruitment enterprises have also achieved information exchange with applicants. Some netizens said that they had participated in many AI interviews, and only one of them received an interview report on their soft ability matching, specifically involving learning ability, communication ability, team cooperation ability, problem solving ability, stress resistance ability, etc. This approach at least gives the candidate the right to know, and also makes the AI interview results more convincing to the candidate.

In Liu Wei's opinion, there is still much room for improvement in the AI interview system. "It is undeniable that HR has its irreplaceable nature, and AI also has its unique advantages. Therefore, human-computer interview may be a more ideal interview trend in the future."

Reference:

Global AI Recovery Market Share Is Expected To Grow At A CAGR Of 6.5% By 2028, source: Facts&Factories;

Software Accelerates to Swallow the World: The penetration rate of domestic AI interview continues to increase, source: Haina talent;

AI is self-taught, but the recruitment engine has a preference for boys! The reason makes people speechless. Source: Red Star News;

AI interviewers are too easy to cheat! Using bookcase pictures as video background can improve the popularity by 15%. Source: qubit;

Non contact Recruitment Season, Advanced AI Interview, source: 36 krypton.

*The title and some of the pictures are from Visual China.

*In the article, Qiu Qiu, Lucy and Angela are pseudonyms.

*Disclaimer: In any case, the information or opinions expressed in this article do not constitute investment advice to anyone.

(Statement: This article only represents the author's view, not Sina.com's position.)

Share to:
preservation   |   Print   |   close