Scripted video interviews are all the rage these days at investment banks and other financial firms. If you want to get in front of a live person as a junior candidate, you’ll likely first need to put on a suit and tie and film yourself answering pre-recorded questions through an automated system. And if the firm is utilizing AI-driven platforms like HireVue – and many do – you’ll be assessed and scored by predictive analytics software before a human ever lays eyes on your performance.
While HireVue doesn’t disclose all its customers, it has been known to be utilized by banks like Goldman Sachs and J.P. Morgan as well as hedge funds and venture capital firms including Point72 and Sequoia, which will mix in technical and behavioral questions. The system is capable of assessing thousands of different traits, including word choice, voice inflection, eye movement and facial cues, among many other data points. We talked to the man charged with building those platforms to get a bit of a look behind the curtain.
While incredibly complex in terms of its technology, Hirevue’s objective is rather simple. It discovers the competencies, attributes and behaviors of the firm’s best employees who were hired for a similar role and builds the ideal profile for a given job description, according to Nathan Mondragon, Hireview’s chief industrial and organizational psychologist. Candidates are then judged against that profile. The algorithm is different for each company, and it can learn and adapt as it receives more data.
“Top performers answer questions in certain manners; they have specific thinking styles and behaviors,” he said. Different parameters are set for each industry and company. In healthcare, for example, displaying signs of empathy is important, Mondragon said. For banking clients, the use of analytics is a key metric. Showing you can problem-solve and make data-informed decisions during your responses is therefore critical.
How to succeed in a HireVue interview
On the surface, it’s difficult to provide advice for acing a HireVue interview because the system is designed so that it can’t be gamed. Its artificial intelligence technologies can pick up on false bravado, memorized answers and abnormal levels of eye contact, Mondragon said. “You can’t fake your way through it.” However, knowing all this going in can be extremely helpful.
In fact, the machine can actually penalize you for going overboard. Your score may improve by using a certain keyword, but eventually you will hit a cliff if you mention the word too often. “More is not always better,” Mondragon said. You can over-smile and be overly-friendly, he said, referencing a case study of an Indian printing company where the customer service was so “good” that employees were giving away printers to clients when they had paper jams.
On the other hand, you’ll also be dinged if you speak with no inflection in your voice and stare at your feet the entire time. And the attributes that HireVue looks for are different for every company and every job; a behavior is only included in the algorithm if it is shown to be statistically linked to job performance for that role. Being empathetic may matter at one firm while not mattering at all at another. The only real way to find the middle ground is by not trying to. “Be yourself and you’ll come across as genuine,” Mondragon said. If you try to beat the machine, it will always win.
How does HireVue’s scoring work?
When assessing candidates, HireVue uses buckets rather than specific letter or number grades. You’ll either land in the top-third, middle-third or bottom-third, so know that one tick here or there won’t make a difference, Mondragon said. While HireVue recommends clients only rely on its assessment as part of the recruiting process, Mondragon suggested that the analysis is fairly prescriptive. People who are hired from the bottom bucket will typically see a 70%-90% failure rate – getting fired, leaving within the first 90 days, receiving negative reviews etc. If the system is as accurate a predictor as Mondragon said, firms likely lean on it more than they'd like to admit.
The bias issue
One of the reasons HireVue exists is to try to eliminate the subconscious biases of human beings. While personal connections are important, a candidate sharing the love of a particular sport with a recruiter shouldn’t give them a leg up, Mondragon said. However, he did acknowledge that bias can exist within AI-based platforms because the algorithms are created using historical data that may already have subjectivity baked in. One such example comes from Amazon, which scrapped a similar hiring tool after it realized that male candidates were scoring higher than females because the technology sector is dominated by men. The system simply taught itself that male candidates were preferable because they kept getting hired.
Mondragon said that HireVue avoids these issues by crosschecking all datapoints across gender and race and removing them when needed. One such example: rate of speech. “Generally, females speak quicker than males,” Mondragon said. “We now know that rate of speech may be influenced by gender.” So there’s no need to speed up or slow down.