Ah yes, taking an IQ test for Amazon wasn't enough of a contrived hoop to jump through already. Let's take one example of a tech firm and broadly declare the resume dead, because silicon valley is the only industry that exists, right?
I'm all for innovating on hiring practices. Resumes and interviews aren't ideal. But, as with everything else, the details of how new ideas are implemented are what matters. Their examples of questions are certainly not applicable to all organizations, so it makes me wonder where they came from: If a hiring manager picked their own questions, it will codify the process with that manager's bias. But if it really did come from an working "AI" process that pulled common traits from the top talent in an organization, it sounds like a plausible process. But before I jumped into it, I'd need to know many more details. I also wonder about the claim that matching candidates to existing staff will promote diversity on anything but the surface levels of gender and ethnicity.<p>(EDIT: After a bit more thought, it is OK if only surface level diversity is achieved.. that is better than being unfair to some demographics... I just hope for more when hiring.)
If you think what you say is important, don't occlude it by a video of rant having nothing to do with it. And don't start the piece with a 20 years ago study. Inbound marketing.
<i>...distilled what used to be a four-hour, academic process of evaluating a person's cognitive and emotional capabilities into a 30-minute game-playing scenario.</i><p>uh. Next.
Is this for real? Engineering firms hire top talent with a 30 min personality test? No regard for previous accomplishments? Because “AI”. Excuse me while I rofl.
Another problem here starts with the definition of "top employee". In some cases it's clear cut. In many others, ask 10 different people and you'll get 10 different perceptions of who is a top employee. At many companies only the most competent nutjobs and/or brown-nosers make the list. Good luck.
Fuck employers. I don't want to be "hungry" for work. I want to think about your problems for a few hours per day, in exchange for some money. I don't want to think about your problems during dinner, during board games with kids, during sex, during my dreams, or during my weekends.<p>Edit: I used to be hungry for work before I burnt out and lost almost all of my social circle. Never again.
TLDR: potential candidates complete personality tests and logic puzzles and then "AI" compares candidate results to those of your current top performers.
I think it makes sense that smaller companies should be looking for the candidates who aren’t “obvious” hires for larger companies which can better compete for pay. On the other hand I’m not convinced that this A.I. test either works particularly well or is particularly fair, or even that “hunger” is the important trait to have.<p>The article doesn’t refer to any evidence that it works, and considering that it is basically marketing fluff, you would expect them to talk about such evidence if they had it. A second problem is that even if the test works then all it means is that you get more people who are like your “top engineers.” That adding more of these people must improve your productivity is a very reductionist view of the world. And what if the “top engineers” being optimised for aren’t actually the kind of people that are useful to the company. For example if a few dispersed jerks with huge egos produce a lot of code, that could be manageable. But if you hire 5 more of them and can’t keep them apart you could have a problem.<p>This is supposed to be something to help small companies find talent. Where is the training data coming from? A small company won’t have enough samples to get a useful model so perhaps an accumulated dataset from many companies is used. In this case one should ask if it is good or bad that different companies’ notions of top talent are being mixed. Another problem with the training data is that only people who are hired will be subsequently evaluated and put into the training data. So such an A.I. would only be able to evaluate people who are already like those who have already been hired. What happens if you aren’t like those people? Do you just get rejected because the system is uncertain about whether you are any good, or get a random result because the system just picks a few mostly irrelevant or wrong qualities that it’s seen before. How can this ever evolve to make predictions about a wider variety of people if they aren’t evaluated because they aren’t hired because they weren’t much like the training data.<p>This test is supposed to improve diversity but it is focused on optimising for a single quality which sounds like a way to hire people who may be diverse on the surface but are mostly the same. I expect extreme levels of “hunger” is also something that will correlate to being a young man with not much of a life, so not great for diversity. That was just a hunch, let’s look at the example questions:<p>>What did your parents do for work?<p>>What do you believe about the world that other people don't?<p>>Who paid for your college education?<p>>What has been your biggest failure in life?<p>>Why do you want to join a team where the hours are longer and the pay is lower than a big company's?<p>The first thing I wonder is how the answers are processed. Either they are typed into a box and some A.I. magic happens or they are judged and scored by a human based on how they fit certain criteria. In the first case, what happens if the sentence structure is weird or the vocabulary unusual or whoever types it in uses incorrect grammar or spelling? Would the system be incorrectly understanding what is written? In the second case it seems like the interviewer’s bias could be reflected in their scoring. This could be made more fair by typing responses into boxes and then having a second person score a set of answers from several candidates at once but this still would allow for some of that persons biases to slip in.<p>Looking at the questions themselves it’s unclear to me that these questions would be likely to improve diversity and it also is quite unclear what the “right” answers are in some cases, which makes the interview process more stressful for the candidate.<p>Asking about parents jobs seems a great way to discriminate: what if one only has one parent, or is an orphan, or has a parent in prison? What if the candidates is old and has parents who are retired or long dead, where their occupations wouldn’t seem to have much relevance to the candidate’s qualities. How would an answer of “I prefer not to say” be treated?<p>Asking about college education is a good way to exclude people who weren’t college educated yet surely such people are the kind of non-obvious hires one would want to look for. Perhaps such people wouldn’t be given this question but then what if the A.I. mainly focuses on the answer to the college question? This could move the “must go to college” requirement from the resume screen to the A.I. or training data.<p>The other questions seem like they could also have quite different answers between candidates from disadvantaged and advantaged backgrounds. Perhaps the A.I. would pick up on this through the answers and just transfer any bias in the training set into hiring decisions.<p>Finally, what happens if someone decides to sue a company using this test for discriminatory hiring practices? How could this test be defended in court?<p>From the candidate’s point of view, hiring practices can already seem bizarre and opaque but this test is even worse.