In the fight against unconscious bias, Sweden experiments with AI – but will robotic interviewers really catch on?

Meet your interviewer, Tengai: she’s friendly, observant, and did we mention she’s a robot?

Tengai’s 16-inch glowing face sits on a table at eye level with the candidate. She smiles and blinks forming empathetic facial expressions as she asks “have you ever been interviewed by a robot?”

For most people today, the answer would be ‘no’, but that might not be the case in five years. For that, we have to thank Furhat Robotics, creators of Tengai. The AI and social robotics company that has spent the past four years building a human-esque computer interface that can replicate our speech and subtle facial expressions. The idea being, a human-like robot is much less scary than machine-like one.

A few companies, like the Swedish recruitment firm TNG currently trialing Tengai, are experimenting with the use of robots in their early interviews, and many more companies already use some form of AI in their selection process.

Many people have reservations about using AI to make hiring judgments. ‘How much should we trust these droids?’ they wonder. Indeed, the question has its merits.

Proponents of the technology point out humans aren’t great judges of character or ability because we are all affected by unconscious bias, which impedes our ability to assess fairly. And what’s more, this bias is hard to shake – numerous studies have shown that trainings which aim to minimize unconscious bias’ influence in the workplace has little to no effect on future behavior.  

On the other side, decriers of recruiting’s new reliance on AI-driven judgments point out that the underlying algorithms of these technologies can also be biased and given the homogeny of the tech and data-science field, we must be careful.

Let’s dive in…

Understanding Unconscious Bias

To grasp the argument for using robots in interviews, it’s first important to understand what unconscious bias is and how it occurs. As it has become a workplace buzzword, there’s often confusion around the actual defintion.  

Dolly Chugh is an award-winning psychologist and Professor at New York University. She specializes in the psychology of human bias. Speaking on the 10% Happier podcast, she offered a very helpful example which frames unconscious bias in a way that’s easy to understand:

“When I say ‘Twinkle, twinkle’ your mind probably automatically jumps to ‘…little star’. Somehow, ‘Twinkle, twinkle, little star’ became an association in your mind. And there are many other associations we’ve all internalized.

We may associate certain groups of people with certain attributes. But we probably don’t remember when that became part of how we think. This is because it’s all part of the flow of our unconscious mind. The unconscious mind represents the vast majority of our brain’s work. Unconscious mental processing is absolutely essential for us to function as human beings, but sometimes it can lead us away from being the person we’d like to be in certain situations.”  

Reining in biases, as a recruiter, can be tough. First impressions and gut feelings count for so much during interviews. But left unchecked, unconscious bias can result in unfair judgments, overlooked talent, and ultimately, discrimination.

Enter The Robots

Robots designed specifically for recruitment purposes are generally much better than humans at making unbiased decisions about a candidate. This has already proven to be true through a recruitment experiment conducted in Sweden.

Since late 2018, Swedish recruitment agency TNG have been using an AI-driven robot head called Tengai to conduct first stage interviews in place of human recruiters. The results have shown an encouraging example of AI eliminating discrimination rather than amplifying it.

New Technology, Same Process

Tengai was built by Furhat Robotics. At first glance, it’s appearance is a bit disconcerting. It’s a disembodied robotic head with a friendly-looking human face. It’s placed on a table where it sits about eye level to the candidate. All this to ensure the interview proceeds in a way which most closely candidates are accustomed.

Unlike a typical interview, Tengai doesn’t engage in small talk. This ensures all recorded responses are work-related. Each candidate is asked the same questions, in the same order, with the same tone of voice. Tengai standardizes the process to a degree that even the most methodical human recruiters can’t match.

Crucially, Tengai knows absolutely nothing about a candidate’s race, religion, gender, appearance, or other visual/auditory factors that commonly give rise to unconscious bias. Once the interview is complete, human recruiters then receive a transcript of the answers. It’s then up to them to decide whether or not to move that person forward.

By replacing human recruiters with Tengai, TNG and Furhat feel they are creating a fairer hiring process that still retains a ‘human’ touch. But it’s important to keep in mind that robot interviewers bring their own set of limitations and challenges.

Hiring Involves More than Interviews

Research shows unconscious bias often influences whether or not someone advances to the interview stage. Tengai and other robotic interviewers offer zero benefits when it comes to creating fairer resume selection processes. So, unless such robots are accompanied by other bias-tackling measures throughout all other stages of the hiring process, they are an incomplete solution to the problem.

Key Information may Go Unrecorded

Robots like Tengai create an extremely rigid interview process. And while standardization offers some benefits, it isn’t perfect. It prevents the natural ebb and flow of a typical human conversation. For example, a candidate may mention something in passing that grabs an interviewer’s attention. If they feel the information is highly relevant to the role, they can ask the candidate to elaborate or clarify. This helps prevent great candidates slipping through the net due to unintended omissions in their answers.

Candidate Experience And Employer Brand

Even when the rationale for using robots is clearly presented, some candidates are not going to like being interviewed by a robot. It’s easy to see how the process may be perceived as cold and clinical by some candidates. As a result, this may cause significant damage to the employer brand. Others will welcome the change, but introducing robot interviewers will undoubtedly polarise candidate opinion. So it’s important to weigh up the risk/benefit ratio, which will vary widely from business to business.

Not All Technology Is Similar

When selecting any new technology for a hiring process it’s important to consider what it actually does. Some robots are made to standardize the initial interview process to root out bias, while other tools use algorithms to evaluate the candidate’s facial expressions or speech. With the latter example, especially, organizations need to rigorously question the way the AI is making judgments and track if those judgments are leading to a homogeneous talent pool.

With so much focus on how machines will automate recruiting, one researcher argues we cannot overlook the importance of human psychology as a crucial machine learning tool.

Given the amount of sci-fi films released in the last three decades involving robot uprisings, it’s clear that humans often cower in the face of machines’ intellectual superiority. However, the crux of many of these movies is that humans often win against Artificial Intelligence (AI) thanks to our creative problem solving and empathy. We point to this distinction to reinforce the idea that human labor is more valuable than robotic automation, particularly for tasks that require emotional intelligence.

This logic applies to make the argument why recruiters are indispensable to the hiring process. Whereas machines are capable of carrying out defined tasks, humans can differentiate nuances between candidates, allowing them to determine the best fit for any given role. Nevertheless, this hasn’t stopped the rise of AI technology within talent acquisition.

Industrial psychologist Dr. Charles Handler believes that organizations will fully automate, and enhance, the applicant selection and screening process by combining deep learning AI with the science of psychology. Dr. Handler sees greater potential for machines and humans after combining forces. “While a computer can beat any human at chess 100 percent of the time,” he says, “a computer and a human playing together can beat even the most advanced chess computer.”

Experts like Dr. Handler believe that automated predictive hiring decisions will soon be able to achieve near 100 percent accuracy. This means that an AI assistant or chatbot will choose candidates with infinitesimal margins of error.

In theory, this could replace an entire recruiting staff and automate the majority of HR functions, leading to billions, if not trillions, in bottom-line savings. AI assessment software like SmartAssistant, Zoom.ai, Textio, and Ideal are already in use at many forward-thinking companies. What’s more, high-impact teams are 6x more likely to use AI, predictive analytics, and other tech solutions to make data-driven decisions over their lower-performing counterparts. Research indicates that top performers see 18 percent higher revenue and 30 percent greater profitability compared to those that don’t use these tools.

In some ways, current AI technology is reminiscent of the droids in Star Wars: Attack of the Clones. Much like the simple robots themselves, basic AI can complete programmatic tasks like keyword or image recognition, but fall short when facing certain, more complex operations. The droids have no problem volleying lasers in gigantic battles, but once an enemy gets up close, these droids struggle to hit the broadside of a barn.

Currently, recruiting AI can scan a resume to identify patterns and compare against pre-programmed “high-performing employee” profiles. The model then predicts if the candidate will be successful based on the amount of matching criteria. Today’s AI models are still leagues away from watching and evaluating candidates through video interviews or more traditional face-to-face interactions.

Dr. Handler believes that these shortcomings could be solved if engineers were able to develop deep learning AI with an advanced understanding of human psychology. He argues that machines currently follow guidelines much more efficiently than humans, but cannot make value judgements about their decisions. By somehow “teaching” machines elements of human psychology (i.e. empathy) and assigning meaning to their behaviors, automated processes like candidate selection would look, and feel, more human.

“For hire-bots to be able to do their job as well, or better, than humans, they are going to have to understand individual differences the same way that humans do,” says Dr. Handler. “In other words, to be truly game-changing, hiring assessment AIs are going to have to think like psychologists.”

In this sense, “thinking like a psychologist” means the bots must base their analysis on psychometrics, clinical measurements of knowledge, abilities, attitudes, and personality traits that determine an individual’s true self. Hypothetically, by combining this measurement criteria with programmed professional judgment, an AI model could spot individual differences between candidate profiles without ever looking at resumes and decide which candidate would be the best match for an open position.

Counter to these utopian ideas are ethical arguments criticizing recruitment AI tech. One argument warns how genetic data could factor into hiring decisions, introducing a previously unfathomable bias. Another fear from critics is the idea of AI “cyber-snooping” on social media to create potential candidate profiles for a position. Despite these concerns, Dr. Handler appears confident in the future of recruitment AI.

“There is a lot to be gained by using AI to help psychologists to better understand individual differences,” he argues. “So, when creating the hire bots of tomorrow, don’t forget to include good old-fashioned psychology into the mix.”

This ecommerce giant made its billions by automating shopping service, but when it came to recruiting it made three crucial mistakes that lead to bias.

According to sources close to the project, it was obvious from the first year that AMZN.O – Amazon’s Recruiting AI – did not like women… like, at all!

The classified project quietly started in 2014, the Seattle company sought to create in-house computer programs to review and score candidates, sources told Reuters. “Everyone wanted this Holy Grail,” one source shared. “They literally wanted it to be an engine where I’m going to give you 100 resumes, it will spit out the top five, and we’ll hire those.”

In much the same way that customers rate products, AMZN.O rated candidates from one to five stars. However, a year into the experiment a gender bias became apparent, especially for software developers and technical posts.

The problem was in the data – the algorithm was feeding on a decade of almost all male resumes and concluding that the ideal candidate is a man, or rather, not a woman. AMZN.O would dock points from graduates of all women’s colleges and downgrade resumes with the word woman/women like “women’s chess club captain.”

Sources say that the algorithm was edited to be neutral to these specific terms, but there was still the fear that the program would teach itself new ways to detect femme resumes and continue to grade them lower.

And though Amazon owes most of its success to its ability to automate everything from warehouse management to pricing, the project was scrapped last year as executives lost hope that AMZN.O could ever be functional. The sources, who only agreed to speak with Reuters a year after the project ended — and under complete anonymity — maintain that no hiring decisions were made using the bias AI.

For some, this story is proof that we aren’t ready for AI in recruiting, and indeed there is still much to learn. Computer scientists like Nihar Shah, who teaches machine learning at Carnegie Mellon University, warn that an algorithm is easier to make than to control.

“How to ensure that the algorithm is fair, how to make sure the algorithm is really interpretable and explainable – that’s still quite far off.”

However, according to 2017 CareerBuilder survey, 55 percent of US HR managers said that AI will be a regular part of their work within the next five years. So is the solution really to avoid artificial intelligence in recruiting all together, or are there some lessons we can glean from this AI debacle? For further insight, we talked to the director of product for SmartRecruiters, Hessam Lavi.

“Developers of this type of systems have an enormous responsibility to prevent negative biases to shape the artificial intelligence they want to produce,” says Lavi. “So, proper training needs to take place to learn not just the technical and process effects of artificial intelligence, but how AI will affect natural beings as well.”

Lavi, who recently headed the team in building SmartAssistant, the first recruiting AI native to an ATS, sees three crucial mistakes when it comes to AMZN.O

1. Thinking the bias is coming from the machine: Negative biases are unfortunately part of the recruiting trade, whether from humans or machines – only it’s much harder to detect in people. So, having a system that makes biases apparent is valuable in itself. The AI learns from the data you feed it, so it’s not the program that’s biased so much as the people who made the decisions that the computer is now analyzing. Eliminating the program is not tantamount to eliminating bias.
2. Limiting the data set: The dataset from one company, even one as big as Amazon, just isn’t enough. A singular company may be using bias paradigms unintentionally. The bottom line is, the more data the better.
3. Deriving future predictions from past events: Past-predicts-future AI can work great for domains such as medical imaging that have a very narrow focus, for example, forecasting the growth of a tumor where an AI can be trained to make clear-cut decisions and act as an expert. However, in hiring which involves a wide range of factors, this type of assumptive AI tends to emphasize biases of the past. If you only had men hired in the past, the algorithm may assume it’s because they are the best people for the job and will continue prioritizing them for future positions.

His best advice? Avoid the black box!

“When we built SmartAssistant we split up the decision processing into smaller, distinct components,” says Lavi. “For example, one component would analyze candidates’ industry experience, one would examine education, one would evaluate soft skills, and so on. Through creating these stand-alone units, we can trace negative outcomes back to their origin and understand why they are happening.”

“We believe the final decision in the recruiting process will be made by humans for the foreseeable future,” Lavi affirms. “But, AI has the ability to automate many of the repetitive tasks and winnow down the stacks of resumes that overwhelm recruiters and cause them to lean on their negative biases. AI technology is much more than just automating tasks and it can teach us about how we make decisions and point out shortcomings in our abilities.”

Fed up with human inefficiencies, this former recruiter used her math skills to automate a very human process, and save a ton of time in doing so.

Aida Fazylova used her background in mathematics and data science to sketch out a new approach to automate as much of the recruiting workflow as possible using artificial intelligence, automation and chatbot technology. The result, XOR.ai, is a customizable AI chatbot and workflow automation to engage, screen and hire candidates 33 percent faster than the normal hiring process, providing extended analytics about the candidates, including predicted employee lifetime value. With XOR, the recruitment process is completely automated, from the moment the candidate engages, right up to making a job offer.

Now, headquartered in Austin, Texas, with major customers in 15 countries around the world, XOR is the emerging leader in Human Resource/ Talent Acquisition process automation, and we sat down with Founder and CEO Aida Fayzlova (picture below explaining her amazing product at Hiring Success 18 Europe), to hear why this is so important.

Why is your product a necessary tool for any SmartRecruiters customer?

Our product combines the latest in Artificial Intelligence, machine learning, automation, and chatbot capabilities, eliminating the most time-intensive parts of the recruitment process. It’s available on your website 24/7, and can communicate in over 100 languages. XOR moves candidates through the funnel even while you sleep. Our AI will learn from your recruiters, hiring managers and candidates who come to your site, answering questions and gathering information, all to offer predictive analytics regarding future candidates and employees.

Briefly, on the back of a cocktail napkin, how does your product work?

Your candidate experience journey is elevated. XOR is with your candidates every step of the way, communicating conversations and data back to your hiring team in the ATS.

What does your product do that can’t be replicated?

Currently, XOR supports 103 languages and has global hiring capabilities. Our product is also known for its sophisticated scheduling functions.

How does your product help make the hiring process as easy as possible?

XOR automates 60 percent of repetitive administrative tasks that your recruiters are handling every day, which allows more free time for your team to focus on more imperative, strategic tasks.  

At what stage of growth are you, and where is that relative to how big you want to be?

We’re at a healthy stage of growth with a great headquarters in Austin, and major customers in over 15 countries around the world.

How does your product compliment the SmartRecruiters TAS – philosophically and technologically?

XOR is considered a one-stop solution for candidate experience. Providing self-scheduled meetings with recruiters, it’s available 24/7 to answer questions for candidates and walks applicants step-by-step through the entire application process. XOR will keep your candidates continuously engaged and send status updates until the position has been filled. Our software works the front end of the ATS for your candidates, automating a great deal of work for your recruiters and is proven to recruit 33 percent faster.

What led you to partner with SmartRecruiters?

SmartRecruiters has the same sweet spot we do. Most SmartRecruiters clients are enterprise customers and high-volume recruiters, which is really interesting for us. The simplicity and the beauty of the ATS is elegant and modern. Designing products and platforms in that same vein is why we chose to partner with SmartRecruiters.

How long did it take to become integrated into the SR marketplace, and what were the obstacles or adjustments in that process?

Our integration into the marketplace only took about a week, with one person, which was a pretty quick turnaround. We avoided any obstacles through preparation and research. There are some obstacles in getting the API to speak to each other but once we integrated it’s been working great.

How long would it take for the average SR user to implement and take full advantage of your product?

Set up will take about 3-6 weeks, most of this time will be spent with XOR’s customer success team, working closely with TA pros, learning the knowledge piece of the chatbot, going over screening scenarios and linking calendars to chatbots to self-schedule candidates. Our team wants to ensure your chatbot is customized to your recruiters’ needs and goals to be as successful as possible in their recruiting efforts.

What would you say to a company unsure about joining the SR marketplace?

SmartRecruiters marketplace is one of the best on the market. It’s full of incredible partners which gives you a great reach to SmartRecruiters customer base and vice versa. Great events like Hiring Success are organized to focus on bringing as much business to their partners as possible, and during those events, you are able to start building relationships with businesses and close a lot of deals.

What do you see as the future of Talent Acquisition and where do you fit into it?

We see it as a rise in conversational interface. Right now, most web interactions are form based. Forms are great, but there are better, more customer-friendly interfaces. Your friends meet you and they want to have a conversation, not fill forms, right? The same thing happens online. Users want to engage, ask questions, and have a personalized experience, not be stuck with one-size-fits-all forms.

Find out what un-conferences are all about, as SmartRecruiters’ head of DACH sales development navigates one for the first time.

First off, let me explain the concept. An “un-conference” is not quite the opposite of a conference, it’s more like a free-spirited sibling. In both instances you go to learn, it’s just two different ways of going about it. Un-conferences are typically intimate gatherings which ask a lot of the attendees in terms of engagement – one does not attend an un-conference, one participates.

I arrived at HR Bar Camp a bundle of nerves, as I knew the first activity of the day would be picking our schedule. And not just selecting sessions. Participants would be creating our agenda in a whirlwind pitching session.

Nine am and 200 eager participants are ready to get started. After a quick introduction from our hosts Jannis Tsalikis, VP of HR for VICE DACH, and Christoph Athanas, managing director and senior consultant at Meta HR management consultancy GmbH, the pitching begins and anyone who has something they want to share, or burning HR questions they want to discuss can stand up, give their spiel, and hope their suggestion gets picked as one of the day’s sessions.

Being my first unconference, I figure I should go all in. Why miss an opportunity for an intriguing conversation? I juggle a few ideas around CRM, job-matching, chatbots… but after a few minutes I know it has to be recruiting AI. We HR people are still trying to get our heads around this new reality that, until recently, seemed like science fiction, so I knew people would have a lot of questions: Will I be replaced by a machine? Will AI help root out unconscious bias?  

The timing couldn’t be better actually. See, SmartRecruiters recently announced the first native AI Recruiting Assistant, and this new capability helps discover candidates and checks bias while saving recruiters time. I teamed up with my colleague Camila, and 30 minutes later we were throwing our names into the ring as potential session leaders. Ours was chosen and we rushed to prepare for what turned out to be a full house.

There is a lot to cover and only an hour to cram in all this information, so we begin with one of the topic’s basic premises: candidate selection. How does the process actually work in most organizations?  Research shows that recruiters spend an average of six seconds looking at a resume and it makes you wonder – How many well-earned call-backs never happen? How many good resumes are lost among stacks of paper?

Most of our participants agree that six seconds is not enough time to make an informed decision on an applicant. Some even wonder if it’s worth it to look at resumes at all, as long as they can’t take the proper amount of time with each one. But that leads us back to the problem of volume. There’s no way to take five minutes with each application when you receive 300 for every opening.

Of course, there’s matching technology, “but didn’t we try that already?” a participant accurately points out, which is how the conversation proceeds to the difference between keyword matching and AI. In the former, if a resume lacks a certain vocabulary, or “keywords”, the system fails to identify a match. While the latter technology uses an algorithm based on clustering concepts to imply skills that aren’t mentioned explicitly.

Most of our group works in large companies and receive hundreds of applications daily so this solution really speaks to them, the general consensus being that an AI assistant surfacing relevant applications will enable the recruiter to spend more time on other parts of their job, like working with hiring managers or expanding their networks.

Our conversation about finding the best candidates incorporates two sources routinely overlooked, though they are perhaps the best resource for finding qualified people who are interested in your company already: internal candidates and former applicants. A native AI assistant on your applicant tracking system will assure those valuable prospects don’t get overlooked. Another great thing about recruiting AI is that it cuts way down on bias, inherent or unintentional, by providing a list of qualified candidates without the judgments that are likely to creep in with only a six-second screening.

Looking back, did we end up answering all the questions about recruiting AI? To tell the truth, we probably only created more. And that was my biggest takeaway about unconferences: it’s all about asking questions, and by asking questions, you start new conversations.