False positives and false negatives

Hiring for Tech
February 10, 2020

A city street with multiple "do not enter" signs.

What tech interviews feel like. Photo by Adrian Williams on Unsplash

I don’t want all my posts to be about high-level philosophy, but in these first few editions of the newsletter, I want to lay out some groundwork. I’ll come back next week with a deep-dive into an interview problem.

Some companies have obviously bad hiring practices. Those companies usually have an underlying culture problem. But there’s another type of company, the Silicon Valley tech giants, Google, Facebook, LinkedIn (and their offshoots), which are filled with smart people. Are these companies doing it wrong?

The problem is when a company is a desirable place to work, they need to reject a lot of candidates, and that puts companies on the offensive. In a recent blog post, Ned Rockson wrote about how the prevalent interview process is geared toward weeding out people:

The vast majority of interviews try to ensure the candidate is not a Secretly Terrible Engineer (STE). “Résumé be damned,” the interviewer thinks, “in this 1-hour block of time, I can sniff out the imposter.”)” The interviewers set up a mental obstacle course that the candidate must do well […] running through. They must also keep the correct interpersonal attitude: talkative, friendly, and professional.

This entire framework of sniffing out the STE propagated into nearly every interview process, partially because the interviewers want to ensure they don’t hire the dreaded STE. It’s also partially to indicate to candidates their potential coworkers could never be STEs.

This comes from a policy of minimizing false positives (under-qualified people passing the interview) instead of minimizing false negatives (qualified people not passing the interview).

And it seems to work. The people I’ve worked with at LinkedIn are smart, motivated, and if they haven’t yet learned to work on a team, do so soon thanks to a culture of investment in our employees. But they come from a subset of all the people who have those characteristics:

That means biases in the selection criteria of those prior institutions carry over to the later ones. If someone wasn’t encouraged to go into CS early, but later went through a bootcamp, they may not know anything about dynamic programming, even if they’ve been doing good work at a startup. This becomes apparent when looking at the gender and demographic breakdown of tech roles at Google, LinkedIn, Facebook, and plenty more.

(To its credit, LinkedIn has hired some great folks from non-traditional backgrounds, and the company has the REACH program to do more of the same.)

The problem has nothing to do with these companies hiring unqualified people. It has to do with hiring a non-representative sample of all the capable people out there. And no amount of allowing laptops, switching to non-algorithmic questions, etc., is enough if the culture is to filter out anyone who might be an STE.

So if you want to hire capable people and be representative in your team demographics, you have to do the following:

  1. You have to recruit from non-traditional sources, not just schools and other big companies. I think a lot of companies that say they care about diversity are doing this.

  2. Then (and this is the part this post is concerned about), you have to switch your interview strategy to actually pass along those from non-traditional backgrounds. Otherwise, your expanded recruitment was just for show.

  3. And of course, you have to invest in your employees. Again, I do think these companies are doing well on this front, but their investment get applied disproportionately to a subset of the people who would benefit.

Ultimately, it comes down to this: an interview is about finding an individual strengths, not attacking their weaknesses.

As usual, all opinions are my own and don’t reflect my employer’s. This post was inspired by the fact that I’ve worked with multiple people who would have a hard time passing a traditional big tech interview. I’d like to work with people like that again.

This post was sent out on the February 10, 2020 edition of the Hiring For Tech newsletter. Subscribe to get future editions sent to you by email, once a week.

powered by TinyLetter