Over the years we have seen many different kinds of AI. Lately, attempts have been made to create robots, that appear more human-like and are thus capable of performing jobs once done by humans. Project Debater by IBM and China’s robot news anchor are a couple that come to mind. An AI robotics company in Sweden called Furhat Robotics has programmed an AI to conduct job interviews. It is the first of its kind and it will be interesting to see how it performs.

Her name is Tengai. Measuring 16in tall and weighing 7.7lbs she’s at eye level as she sits on top of a table directly across from the candidate she’s about to interview. They designed this robot to appear less creepy than a typical robot(however it is possible they went overboard). She can blink, smile, tilt her head to either side, and ask interview questions.

Other companies like Amazon have attempted to use AI to automate the job screening process. The algorithm they used proved to be biased against certain candidates and was therefore scrapped. The approach taken by Furhat may prove to be more effective.


Everyone knows that humans judge each other sometimes within mere seconds of meeting each other. These unconscious biases exist in everyone and they can reduce our ability to remain objective in a variety of situations. A job interview is no exception.

What if your next job interview was conducted by a robot that is incapable of judging you? Do you think you would get the job? Would you feel more comfortable and less anxious? More importantly, do you believe it would be a fair process? That is clearly the intention of this robotics company out of Sweden. However, like always, things are not always what they seem. More on that at the end. Next let’s break down how this robot works.


Tengai is similar enough to a chat bot but it is designed more so for information gathering than typical back and forth communication. This interview robot was trained to pick up certain cues and evaluate a candidates answer to her questions in order to figure out what question to ask next to keep the interview moving. There is no set pattern for these kinds of evaluations.

Tengai has to understand what a candidate says only in reference to her programmed competency framework which is set up to weed out people based on merit alone. This is much more difficult than programming a bot such as Siri for example. Siri simply searchers a database and outputs an answer but Tengai has to learn how to respond in addition to learning what to respond.

Despite the fact that the creators have not released much specific information regarding how Tengai was programmed, it is clear that a multitude of machine learning algorithms were used to train her how to deal with a person’s responses. This includes the way she nods her head and the other subtle mannerisms that are designed to make her appear less robotic.

Tengai was also supposed to have been taught by multiple recruiters to help prevent human bias from making its way into her programming.


As with everything in the world of AI it is important to keep things in perspective. Especially limitations or flaws associated with your product. Now, this robot is supposed to ensure that the interview process is completely unbiased meaning that candidates will be judged on merit alone and nothing else. First Tengai conducts an interview and takes in all the answers and sends a transcript to the recruiters behind the scenes who are supposed to make a decision based on the answers alone without meeting the person face to face to eliminate bias and discrimination.

The first problem that comes to mind is that when the human recruiters read the answers they absolutely will be judging the applicants based on the way in which they word their answers. The way we phrase our sentences in response to tough interview questions can reveal a lot about our character. Including but not limited to race, ethnic background, whether an applicant is bilingual or not, and a person’s level of fluid intelligence. So that right there guarantees this process is not completely unbiased right off the bat.

Another problem has to do with the fact that the human programmers believed they could keep their own biases out of Tengai’s programming. Amazon already attempted this a while ago and their recruiting algorithm turned out to be “biased”. The way they swapped recruiters over and over while training Tengai does not guarantee that even small and subtle bias did not intertwine with the inner workings of her code. We underestimate just how visible these biases are despite the fact that they are usually not all that visible to us.


We could go on and on about the many flaws with this idea since there are so many but the point here is not to bash the creators of this robot. All in all it is not a terrible idea to try and use AI to improve the interview process. However, people need to understand that when interviewing a candidate we must judge them on how well we believe they will do the job. The biases we hold in our subconscious are there for a reason after all. We humans have a very difficult time navigating the world without our various biases to help light our way.

That being said, the potential of this technology is huge. We may one day be able to hand over the entire interview and recruiting process to AI. Only then, could we be sure that it would be truly unbiased and objective.

Leave a Reply

Your email address will not be published. Required fields are marked *