A.I. as Talent Scout: Unorthodox Hires, and Maybe Lower Pay

These algorithms are being based on pre-set measurements and not enough on a job candidates actual personality, creativity and work ethic. 30 - 40 years ago it was not unusual to hire someone who lacked a certain degree for a particular job. At some point HR started to take too large a role in the hiring process and managers forgot how to interview and figure out how to make a good hire.

Monster.com bought Trovix, a search engine that looks for skill fit, not keyword match, 10 years ago. A pity you couldn't take the time to survey the existing job search options and provide a more informative piece.

Algorithmic matching is on the rise, but the capabilities of these AI startups are mainly just skill-based matching and prediction. This is why the main application is for tech jobs, where hard skills factor into performance a bit more.The inherent issue is that skills, particularly self-reported skills, are only one very small part of what makes an individual thrive on a job. Other factors such as personality, preferences, team dynamics, environment fit, and industry can be equally or even more relevant, depending on the position. At SquarePeg we look at hundreds of indicators beyond skills, because what is important for a good match varies depending on the person and job, and it is a combination of factors. Also, there is skepticism around biases and inclusion, which is why it's more important than ever to use demographic data in matching to train algorithms to never prefer one group over another - even if the training data led to those results. There are ways to make algorithms more inclusive than humans, which should be a top priority for every tech startup (it is for us).

The better corporate interviewing courses have long taught that you want to examine whether the applicant shows evidence of having skills that can be used to handle the job. This includes translatable skills that are not necessarily an exact fit. For example, programmers are basically moving bits around. It should not be important that the candidate has checked off all the latest software buzzwords. The new techniques and languages are easy enough for a skilled developer to pick up.

Halfway into our second set, a couple of young women (college aged) started playing on the court next to us. Shortly thereafter, we played mixed doubles with the two girls. I took "Lori" as my teammate. She was blond, beautiful, tall, and smart. Afterwards I asked her out. She started laughing, hysterically. In a few moments, there was another woman who walked over; it was Lori's mom. She asked Lori, "Did this young man ask you out?" Well, it turned out this happened often. Why the laughter? Lori was 14! After some conversation, her mom invited me over for dinner a few days later. I learned that Lori's dad, "Dan" was a career officer at the CIA. We hit it off. Three years later, after forging a father-son like relationship with "Dan," and taking a battery of tests and psych evaluations, I was formally invited to join the agency. He had been in the clandestine operations directorate, served in Europe, then was brought home and served in a number of positions, including as a college recruiter, as he climbed the ranks. I submit that no AI program could have ferreted out my skill set or personality traits as "Dan" did. I eventually ran a counter-terrorism group for better than a decade. AI has great potential, but it will never replace the "Dans" of this world, and that's unfortunate for those companies that come to rely on AI. Want to be in the CIA? Apply on-line. Dan can no longer bring you in. There's no protocol for it.

AI as the perfect Public Defender...Think about it

I am beginning to wonder whether the term "artificial intelligence" is being misapplied to every time a computer is used to do something that people have traditionally done. These screens may or may not be intelligent.

If used wisely, there could be interesting applications addressing occupational segregation by race and gender. But there may also be danger if the algorithm is reinforcing certain employee profiles as successful. Humans are social beings - meaning that social networks and bias influence hiring and promotions. How might the algorithm be affected by (biased) social processes in the workplace?

"But established coders are sufficiently in demand that wages are likely to remain high even as it becomes easier to find alternatives." Well, nobody is beating on my door. Reality is that those in authority are happy to pay peanuts for whatever flavor of primate they can find to work at that rate, regardless of outcomes. If all else fails, they conjure up a massive outsourcing that saves less than it costs, and ride that zillion dollar boondoggle to their next promotion, leaving others to deal with the fallout.

Sounds as if MBAs from lower-tier schools, not to mention holders of a Bachelor's degree in business, are going to suffer a drop in demand... This is a detection machine for Emperors who Have No Clothes.

Once upon a time (before retirement) I worked as a programmer analyst. When I hired, I never looked for people with a specific skill set. What I wanted were people who could think and solve problems. And I wanted to do the interviewing, not HR. Likewise, when I looked for a job, I wouldn't go for an interview that started with HR. HR's job is to reduce the vast flood of resumes to a trickle that can be managed, not to find the best hire, or even a competent one.

As someone who has hired people both here and while working abroad during a long career, I could not be comfortable completely with this AI method. I prefer to review applications personally and to speak directly to those I would interview to gage human attributes and attitudes not included in a resume. In my experience it is a surer way of success.

Where we really need A.I. is in Congress. Just imagine, 435 robots in the House and 100 in the Senate sitting around and pontificating on the great issues of the day. And we wouldn't have to pay or lobby them, either.

A major issue with these techniques is learned bias in the algorithms. It is very possible for deep-learning algorithms to infer race or gender from subtle features of a resume. The algorithms are learning from historical data, where the race and gender balance in most fields is highly skewed. So if the field of software engineering is predominantly white and male, the algorithms can perpetuate that trend. One should not assume that the algorithms will be any less biased than human recruiters.

This is, MACHINE, evaluating and pricing human workers ....based on the " needs" of the corporation......no human or social considerations.....TO DETERMINE the potential "corporate perfect fit" OF employee recruits . Efficient(but meaningless for the worker) selection of "fodder" for factories of the GLOBAL corporotocracies.........Welcome to the dawn ..........of a BRAVE NEW WORLD......and prototype Alpha and Beta worker recruitment !!!

Is there an AI that can detect talent for working in an environment dominated by bean counters, backstabbers and other assorted bozos? Because that is the reality of most companies. If you want a productive application of AI, turn it into a tool for detecting and eliminating the aforementioned killer B's.

creepy. next up: AI matchmaking.

Taking the human out of human resources.

I had a professor who said that just because a dog can be trained to walk on its hind legs doesn't mean the dog does it well.

The argument or lower wages doesn't make sense. The pool of people is not changing. If you are now picking someone that otherwise wouldn't be picked by anyone, that person is getting a better job and higher pay. What this is doing is improving the situation for those with skill that lack credentials, and reducing the inflated cost of those with credentials but not the appropriate skills.

Finally...qualifying on the merits of true ability in hiring. What a concept.

There have been fake doctors whose patients were completely happy with them and their treatment. Then there have been real doctors with advanced degrees who get into lots of trouble with malpractice, poor care, etc. Of all the things education can teach, it cannot create talent. My life experience has shown me, that some who look good on paper, are less than some who don't look good on paper.

That's the plan all along, race to the bottom in term of wages but now in automatic fashion using a fancy name to deflect blame.

They didn't break any cycle. They're still hiring young people, and as cheap as they can't get them. How is that different than the old school bald white guy from 50 years ago?

The examples mentioned are things most humans would figure out - and have learned to act on - with some experience in the world. Athletes work hard, get along with teammates, and are used to the highs and low of competition. Such a background is likely good for sales. Chess players act individually, have focus and recognize patterns, which probably helps with coding. A computer algorithm that guesses that someone with a master's degree in statistics would be likely be good at data science is, well, simply not impressive. It is possible that this software makes more clever insights than the ones mentioned in the article, or at least reasonably aspires to do so. Once it moves below the "low-hanging fruit" mentioned in the article, there is a risk that the predictive value of these findings will taper off dramatically, or that they will too often reinforce some form of discrimination. Maybe chess is mostly popular with people of certain backgrounds, and those are correlated with previous professional success - but that success was biased by favoritism and/or exclusion of other groups?

I have personally used two "AI tools" for recruiting and neither of them was very effective. We beta tested each one for 6 months and walked away both times. AI does not do recruiting well - at least the tools I have used.

As if the soul-crushing, dystopian experience of trying to find a job couldn't get worse, now this. it's already virtually impossible for anyone over 50 to get a decent job thanks in large part to the use and misuse of algorithms. This will be the coup de grace for them and many others. Experience and proven ability will be rendered meaningless. Why hire someone who knows what they are doing when A.I. can pluck a supposedly superior, inexperienced, younger and cheaper candidate out of the blue? This turns the world on its head. All the things that are now assets in looking for a job -- experience, competence, proven ability, specialized education -- become liabilities. The result will be armies of intelligent, educated, experienced, highly competent workers, especially older ones, completely shut out of the job market. It will turn the already insane game of "beat/trick-the-algorithm" up to 11. What a nightmare. We need a new motto in this country: people, not algorithms. Even better, we should consider tightly regulating and perhaps even banning the use of A. I. and algorithms in hiring.

We've seen over and over again that deep-learning based AI has a very nasty habit of baking-in and amplifying the biases present in the data set used to train it. The biases often don't get noticed until you specifically go looking for them, which is difficult to do as deep-learning based AI systems are extremely opaque about what inputs are being used and how those inputs affect the result. These AI systems will discriminate, period. There is no realistic way to get a training data set that isn't tainted by discrimination based on protected classes, and that discrimination will metastasize in the resulting AI. Worst of all since these AIs are being implemented with no transparency there is no way to independently test for the biases. If Eightfold were to hypothetically find out that some CEO at one of the companies used for training data hated kids and their system was systematically rejecting people who had indicators of parenthood on their records (involvement in kid-related community organizations for example), would they make that public and open up every single business that has ever used the system to lawsuits? I think the ongoing cover-up scandals at Facebook gives us the answer to that question.

Since all the professional jobs go only to select connected elite minorities here comes another avenue to ensure that the inequity gap only becomes bigger. Watching President George Bush funeral you could count the number of minorities who were in the guess crowds. The people who work for Facebook my favorite company for stealing democracy and Google who spends all their day monitoring your behavior on the web. Amazon puts their employees in warehouses and the elite get to use the toys of the owners. Democracy is being stolen, health care is being under minded and now we have all the risks of A.I. with no benefits for the common man. Great future to look forward to for minorities who receive the worst education from public and charter schools. I guess by magical thinking we can get into those stem programs that are not even offered in our schools. Credentials won't even matter anymore, we see how well this administration in Washington is working for us all.

Has anyone ever really done a real study on how "well" these algorithms work? I believe that almost everyone assumes these algorithms work but can't be bothered to really find out.

So, average wages of highly paid traditional hires go down, but wages of nontraditional hires, perhaps including those lacking elite credentials, go up, maybe by a lot. Inequality decreases. Productivity goes up because of the better match between jobs and employees. Is this bad?

A I is just that, Artificial. There is nothing real about the jobs nor the skills needed to sit behind a keyboard. Any third-rate programmer can construct an algorithm to select the favored phenotype.

“Maybe lower pay”...that’s a certainty. Corporations don’t do anything that doesn’t disserve their workers and lower their pay.

The whole HR-driven concept of the "perfect fit" is flawed. Assuming that everything works perfectly (ha!) from the accurate resume through the accurate job description and candidate assessment, the search for a perfect fit surfaces people already performing a similar job. Peter-Principled burnouts get through while talented, entrepreneurial people ready to surge forward into new opportunities get turned away. No machine or human can get this right all the time, but it's hard to believe that machine learning can consistently do better than an experienced professional who is close to the job rather than close to the HR bureaucracy.

I have had a long career as an engineering VP of hardware and software development organizations with hundreds of engineers. I have always scanned resumes for hidden talent because it often results in the best hires. I have found it very useful when there is a shortage of talent in a specific specialty but there are many people with skills in other specialties who could adapt quickly. It is also useful when you want to push the state of the art in some technology area where the skills you need may not yet exist. I learned this from a brilliant recruiter who staffed the organization that designed the Internet. Since no one had ever designed an Internet he hired people with PhDs in mathematicians and physicists because they had the closely related skills. I have always been appalled when recruiters just look for key words in resumes to identify job candidates. It became even worse when recruiters began to use software to do key word searches. I can't even imagine how many excellent candidates were lost in the process. One problem with looking for hidden engineering talent is that the person screening resumes has to have a solid understanding of the technology being developed. I know a pattern matching algorithm used in AI can simulate understanding if there is enough data to train it. I don't think there is enough data to train an AI system to identify hidden talent for highly skilled professions.

Hilarious that you are writing a gushing article about a company which profiles people for jobs, and, a day after you branded Facebook evil for doing so, gush about Eightfold: "Eightfold, which is based in Silicon Valley, also makes a clever trade with clients: to improve its results, Eightfold asks clients to use its human-resources software, which imports employee data in anonymous form. This includes information on how workers with different backgrounds perform in different jobs, and how much they earn." Equally hilarious is that you came out with this odd article only a day after an article on affect recognition at The Intercept was branding profiling of job applicants by their emotional reactions as phrenology. You don't need deep learning to know that U Mass puts out people who do natural language processing. You need a hiring manager and an HR staff that pays attention. Speaking of which, what role does the hiring manager, who puts together the job script, play in this process? Your article doesn't say. Lastly, if Eightfold was looking for a C-class officer, would their algorithm make being Indian a criterion? You should check before boostering this software any further.

Moneyball comes to mind. The Oakland A's hired cheaper players after all.

It would appear that AI will have its own very real issues of built in bias, positive and negative. It will be interesting to see how many of the left leaning executives in the industry will walk the fine line in promoting its use.

Agree that all the hype about AI is sometimes misplaced. As someone who has hired software engineers for years, I learned long ago that someone with a music degree and some experience in coding was very often a great candidate. A person, like myself, with a computer science degree may be a good architect and terrible programmer. at the end of the day, you still need a smart person to review a resume and interview. It will be a sad day when machines perform interviews.

Great article but its already outdated. As an organizational psychologist, I can tell you there are already nascent companies developing AI algorithms which assess not just candidates’ job skills but also their personality types. Imagine everything you have ever posted on social media (written, pictures, and video) being used to assess your introversion-extroversion, emotional resiliency, or leadership potential. Imagine a job interview which uses camera based technology (e.g., FaceTime) and the camera pointed at you is reading the microexpressions on your face in real time and matching it to what you are saying (and have said via your previous posts) to make an assessment of your honesty and integrity. Dorothy, this ain’t Kansas anymore.

I was such a hire once. As a social worker with limited exposure to tech, I was hired in 1983 as a technical writer by a software start-up. it wasn't a great job, but it paid about a third more than direct-care social work, which wasn't covering day care. It also laid the foundation for my 30-year work life as a technical writer. But it didn't require AI for my first employer to find me, just a writing test!

It’s important to keep in mind that there has to be several “intelligences” to be used in the hiring process. The artificial kind needs to be another tool in finding an appropriate fit for a new employee. Companies seeing the use of AI as a way to lower payroll costs are taking the short view. They will train these unlikely employees for their next job. And those employees won’t need an AI agent to help determine how qualified they are.

Like the saying goes, “good things not cheap, cheap things not good” While someone with many years of experience can do the job at a fraction of time, one cannot say the same for a novice. So, it all depend on a company time horizon for a new product. For instance, this company, after 6 months of trying to get a model, built by a CS person with no prior knowledge of statistics, hired me to reimplement it. I did it in 10 days. Reason: knowledge of finance, statistics, and the tool. I understand why a physist will be hired as a data scientist. Applied physists know their statistics, and work with data (model physical world)