Chapter 30 Fooling Artificial Intelligence

by Ersin Savaş

In this article, I will tell you how I was wronged before because of hardware with artificial intelligence software and how I defeated the machine by manipulating the same artificial intelligence hardware. 3 years ago, in January 2028, I applied for a job in the finance department of a large law firm. There were 15 candidates invited to interview with me. First of all, we had a joint meeting with all of us. We learned an important detail that makes this job interview different. CV details of all applicants were uploaded to an artificial intelligence-supported system, and artificial intelligence decided on 15 people who were called out of 36 applicants. We learned that support will be received from different artificial intelligence components in the continuation of the recruitment process. The company would decide on the candidates to be recruited at the end of a 3-day process, not with a single interview. In fact, the decision was completely left to artificial intelligence. In-depth interviews with us would be audio and video recorded and processed by artificial intelligence. Likewise, our answers to detailed survey questions would be recorded in artificial intelligence.

After the joint job interview on the first day, individual interviews were held the next day. On the 3rd day of the recruitment process, we saw that 5 out of 15 candidates were eliminated. The people interviewing us in front of us were asking us questions about our work aptitude and work experience in general. Sometimes they would talk about their non-work events and ask for advice, turning the interview into a conversational environment. It was obvious that this was done to provide detailed data about us and to measure our reactions to the artificial intelligence supported system that constantly records us. After the 3-day process, 3 candidates were recruited, and I was one of the 7 eliminated candidates.

This different hiring process felt extremely awkward. There were decisions made by a software that we did not know how it evaluated us. I knew that I was more equipped than other candidates, but how artificial intelligence evaluated me was important for my future job interviews. I did research on the use of artificial intelligence in job interviews, and evaluated the 5 candidates eliminated before me and the 6 candidates eliminated with me, as far as I remember from the joint interviews, with the data I obtained. I was trying to understand how AI makes decisions by making connections between candidates. I tried to find the common aspects of the 3 recruited candidates. As I thought about the 3-day interview process, there were interesting details that came to my mind.

While 6 or 7 of the 15 candidates invited for the interview were using glasses, 4 of the first 5 candidates who were eliminated did not have glasses. 7 of the last 10 candidates were wearing glasses. As far as I remember, the total work experience of all the remaining 7 candidates was over 10 years, while the 3 candidates who were recruited had 2 or 3 years of experience in the finance department. There were totally 15 candidates. The strange detail is that during the joint interviews, there were three people who gave long answers even the answers is very clear. There was not much resemblance to each other in the external appearance of the 3 recruited candidates. While the last 7 candidates who were eliminated, who were odd, wore clothes with intense black and white blue colors, the colors of the clothes worn by the last candidates were very different from each other. I have classified a wide variety of data, such as past work experience, schools attended, and how many years they have worked in a workplace, of all candidates recruited and eliminated through social networks such as Linkedin.

After 6 months, I applied to the job posting of the company’s purchasing department. This department was not directly related to me and my skills. And my friend who has been working in this department at a different company for many years applied for the same job posting.

I made changes in line with the data from first job interview I obtained in my CV file. Business had gone as planned, and my friend and I were called in for a job interview. The same 3-day job interview process started again. I changed my clothes, my answers to questions, similar to those of candidates previously hired for the other department. Although I have no direct interest in this department, I gave long answers to the questions asked and expressed ideas as much as possible. Finally got what I’ve been waiting for. Some candidates with more qualifications than me were eliminated for this department. My friend and I were among the last candidates. My friend, who had no intention of being hired, was eliminated in the final stage with a similar appearance, dress and question answering style to the candidates who were eliminated in previous job interviews. Although I am less knowledgeable and less experienced in this field than my friend, I was one of the 2 candidates who were offered a job.

I had two options before me. I could report all of these things and present them to the top management of the company. In such a situation, I could lose my job with the thought that I was manipulating the AI and cheating to get into this business. In addition, the artificial intelligence used for this recruitment had made its first learnings as a result of the directions made by the managers in that workplace. In other words, some managers might not want this misleading artificial intelligence project to emerge.

However, there was another possibility. The company could also have taken a positive approach towards me for achieving such a thing. In fact, it would be a good opportunity for me to be taken into the legal department, which is my main field.

As a result, I needed a job right now and I did not find it meaningful to take such a risk. It was actually an advantage for me to understand that artificial intelligence made a biased decision in this way. I started working in the purchasing department. Although the first month was extremely challenging, I gradually got used to the work. After working with the company for a while and interviewing new hires, I plan to report this matter to senior management. Perhaps artificial intelligence will improve its gend over time and will begin to make more accurate decisions in new hires.

These companies think that artificial intelligence reveals the most productive person for them in the future by making analyzes that are too detailed and complex for a human to do. He thinks it is right to evaluate the recruitment processes in a concrete way. However, the concrete understanding of a human and the concrete analysis of artificial intelligence are different from each other. People can make the right decisions with different emotions by using the same data according to changing situations. However, artificial intelligence is insufficient to analyze concrete data for different situations and individuals. Artificial intelligence can be manipulated by humans, as was my experience with the job interview. It seems that trusting the decisions made by artificial intelligence will continue to be a big mistake, unless it gains an emotional dimension at least as much as human understanding.