The complexity of human behavior, how quickly we make decisions, and the fact that we have a very real fear of machines. That’s how Sinisa Strbac, Robot Trainer and Project Manager for Tengai Unbiased, sums up his top takeaways from the project so far. Meet the man who is leading TNG’s drive to train the world’s first-ever unbiased social recruiter robot.
We all have our preconceived ideas about robots. We unwittingly absorb these ideas from Hollywood films, the toys we played with as children, and our experiences with automated solutions that perform routine tasks at our workplaces. Still, making the leap from simpler machines to creating a robot that can flash you an acknowledging smile, follow your gaze, and display emotions—something that we humans have had a monopoly on until now—is revolutionary, to say the least. Not to mention a little surreal, too.
Since August 2018, however, this surreal idea has gradually become a reality as TNG’s Tengai Unbiased team gets better acquainted with Furhat Robotics’ social robot. The project’s day-to-day progress is guided by TNG’s Sinisa Strbac. In addition to heading up the Tengai project, Strbac is also a senior recruitment consultant and team leader for a talent acquisition team at TNG. With 12 years’ experience in the recruiting industry behind him, he is far from a newcomer to the field. Nevertheless, the Tengai project has presented Strbac with a whole new type of challenge:
“My colleagues and I find it super-exciting both to have the opportunity to play around with, and also work on developing the robot to be able to hold an in-depth conversation; a dialog in which the robot asks questions, the person being interviewed answers, and the robot then decides for itself which follow-up question is most relevant,” Strbac explains.
”Those of us who’ve worked in recruiting for decades and who think we know the job inside out have been given cause for reflection and reevaluation.”
It was clear from early on in the project that building the world’s first-ever unbiased social robot was going to pose a major challenge. Strbac draws a parallel between Tengai and Apple’s famous virtual assistant, Siri, to illustrate the difficulty. He says:
“In Siri’s case, the human user asks a question, and Siri searches out the relevant information and presents it to them in a fun and social way. At the moment, anyway, it feels like Siri must be easier to program than Tengai. Our task is considerably more complex. In our case, we need to transfer learned human behavior to a robot that has exactly zero previous experience of human social interaction. As a result, we’ve had to completely reset our own thinking.”
“Those of us who’ve worked in recruiting for decades and who think we know the job inside out have been given cause for reflection and reevaluation. Above all, we’ve had to make sure that we don’t bring our preconceived ideas and biases into the project with us.”
A road paved with challenges
So, just what is it that makes the Tengai project so complex?
The difficulty lies first and foremost in teaching the robot to evaluate a person’s answer and to understand which question should naturally follow. TNG asks questions about a candidate’s skills and competencies based on a verified competency framework from which our recruiters draw questions and answers. Still, a competency-based question always yields an answer with no set pattern, adding yet another level of complexity to the challenge. Strbac explains:
“A talents’s response can describe any imaginable situation, be recounted in a host of different ways and in their own individual choice of words. The robot needs to comprehend and analyze this information and then determine which follow-up question would be most relevant.”
In the long run, the greatest challenge will come in training the robot to evaluate and validate candidates’ answers against a competency framework. Since the ability to evaluate key competencies and skills is such a critical part of the project, TNG will enlist the expertise of data validity scientists working in the fields of psychometrics and behavioral science to make sure we get it right. Ultimately, the aim is for the robot to objectively be able to determine whether a candidate has a low, medium, or high level of a certain qualification that is important in the position for which they are being interviewed.
A surprise for many
While programming Tengai is a challenging task, the physical encounter between man and machine has also given rise to challenges of its own; an issue of which Strbac and his team are well aware:
“We’re asking our candidates to put themselves in a highly unusual situation: to sit in a room with only a robot for company and hold a conversation. To be asked a question, answer it to the best of their ability, and then move on,” says Strbac.
The reactions vary from person to person, Strbac explains:
“Some people are skeptical to begin with, but are surprised by how naturally Tengai handles itself in a conversation. Others maintain their skepticism even after their interview, since they have high expectations concerning a robot’s ability to mimic human mannerisms and speak naturally, which Tengai can’t do perfectly just yet, even if it’s well on its way. It’s social, but not in the same way as a human.”
“Then you have others who are positive towards the concept, but who get nervous prior to the actual interview. After all, they are being asked to ‘perform’ in front of a machine that follows their movements and records their answers.”
”There’s nothing strange about it, really; humans do the same thing. But their feeling is that the robot is much more than a human. ‘It literally records my responses!’ is the comment they make.”
In other words, although Tengai is social and acknowledges its human counterpart during interviews, some people can’t help but feel a little like they are under surveillance and so become overly self-conscious. At the end of the day, they are sitting there holding a complex conversation with a computer that will ultimately have real-life implications.
“On the other hand, most people are surprised by just how natural the discussion is; the longer the interview lasts, the less they think about the fact that they’re talking with a robot. It becomes just another part of the interview,” Strbac concludes.