On Algorithms and Interviewing

January 17, 2019 at 10:45 AM | categories: Personal | View Comments

As I write this, I'm hours away from starting to interview for full-time jobs in the software field. I've spoken with a number of recruiters and hiring managers and have received interview preparation materials from a handful of companies, many of which you've probably heard of.

I was hoping things would have changed since I last seriously underwent this endeavor ~7.5 years ago (I did interview periodically when I was at Mozilla in order to test waters, keep my interview skills sharp, etc). But it appears the industry is still generally fixated on algorithms and data structures in interviews. The way algorithms and similar coding tricks are emphasized in the preparation materials I've received, you'd think people in software spend a major part of their work days thinking about and implementing algorithms. But from my experience, this is very far from the case! So why are so many companies and interviewers fixated on algorithms. And is this a good thing?

When they matter, efficient algorithms, data structures, and other tricks are important and useful skills to have. But from my experience, they matter far less than you would think. If I were to make a list of important job skills and traits for software and programming, memorized knowledge of algorithms and data structures is so far down the list that I don't think I would even ask about algorithms fundamentals for most job candidates! (In fact I don't.) I think it is vastly more important to focus on behavioral qualities and potential to actually think and apply knowledge rather than regurgitate it. Algorithms and data structures, after all, are learned knowledge. All other things be equal, I'd rather have someone who knows when to ask for help with an algorithms issue or can pick up the skill than a curmudgeon algorithms genius who has an abrasive personality and clings to old habits.

In the spirit of full disclosure, I should to state that my algorithms skills are relatively weak. You can accuse me of writing this post to fulfill my own selfish interests. You wouldn't be wrong. But I know there are others like me who are good at programming yet struggle with algorithms and question the utility of algorithms in interviews. I'm attempting to write this post for all of us.

I have failed job interviews because the interviewer assessed my algorithms abilities as weak. I'm able to work through this deficiency with interviewers who care more about the behavioral traits I exhibit when in such a situation (I try to be quick about admitting my technical weaknesses and to ask for help when needed). But some interviewers aren't as interested in the behavioral traits or insist on a baseline level of memorized algorithms knowledge beyond my own. I feel like my relative algorithms weakness hasn't hurt me on the job, as I hardly find myself caring about algorithms in the work I do. In the majority of cases, the choice of an algorithm just doesn't matter for the size of the data set. Or a standard algorithm or data structure available in the standard library of the language I'm using is good enough. In the cases where I realize algorithms and data structures would matter, I run my technical questions past someone with more knowledge in the domain than me. Or if I don't do that, it often comes up during code review. Without strong algorithms and data structures knowledge, I'm able to maintain the Firefox build system, become a core contributor to a version control tool (something you think would require a lot of heavy algorithms knowledge), maintain various open source projects, diagnose and address low-level performance issues in complex software and systems. About the only impact that being weak in algorithms and data structures has had on my career is that some companies passed on hiring me because they perceived strength in this area to be important.

Albert Einstein once said, I never commit to memory anything that can easily be looked up in a book. A modern adaptation of that quote may go something like, never memorize how to implement an algorithm or data structure when you can just Google it or use a software library implementing it. If you have knowledge of how to implement various algorithms in your head, that's good for you, I suppose. But I think the bigger brain knowledge to possess is when algorithms matter and to a lesser extent, what types of algorithms are appropriate for particular problems. Answering these problems requires critical thinking. Actually implementing algorithms, by contrast, merely requires knowledge that can easily be looked up in a book (the algorithm or data structure itself) coupled with some programming knowledge for how to apply it. A capable programmer will be able to do both these things and pick up algorithms and data structure knowledge on the job, if necessary.

Some would say that algorithms are a good way to flush out coding ability. And coding ability is important to assess as part of interviewing a job candidate for a programming position! They aren't wrong. But there are much better ways to receive stronger signals about an interviewee's compatibility! On the coding front, there are infinite ways to assess programming capability without involving algorithms. So why involve algorithms as part of the interview?

One way I approach interviewing people is to imagine what the typical work day of that role will be like. How much time do they spend coding, investigating bugs, debugging, attending meetings, writing proposals, politicking with managers, etc. This produces a conceptual pie chart of that role's activities. I then try to structure the interview such that the topics covered in the interview correlate with and somewhat in proportion to activities in that job role. Is the role a heads down junior coder? A team lead or manager? When you start trying to map the time in various areas of the role to time spent in the interview, you realize that the common technical interview overly emphasizes some areas and often completely ignores others! One of the areas that is over-emphasizes is algorithms. Again, your typical programmer is going to be spending most of their typical day doing things unrelated to algorithms. So why are you spending precious interview time asking about algorithms when you could be probing an area that actually correlates to typical job activities? When viewed through this lens, the prevalence of algorithms in interviews just doesn't make much sense to me.

Perhaps knowledge of algorithms should be basic knowledge that every programmer should possess. If so, then asking about algorithms is fair game during an interview, I suppose. But I'm not comfortable with this line of thought.

I've always found it fascinating the ways that people with different backgrounds and degrees approach problems differently. From my experience, some of the best ideas and perspectives come from people with backgrounds and degrees which are minorities in the field. I've worked with programmers with degrees in philosophy and history who were some of the best programmers and overall minds in the room. One of the great things about software and programming is it is accessible to anyone, regardless of background. If you can code, you can land a (usually high-paying) job. Yes, the field is highly technical. But you don't need formal education or a degree to enter it like you do similar high-end professions, such as medicine or law. You can argue whether this is a good thing or not. But I think the accessibility of the software profession - the lack of formal gatekeeping - is something to marvel at, something that we as an industry should embrace and be proud of. Do arbitrary hurdles to joining the industry help or hinder it?

A problem with emphasizing algorithms in interviews is that algorithms are somewhat highly specialized and academic. There are entire areas of programming and software where detailed knowledge of algorithms just isn't that important. The bar for so much software is it works and it quite frankly doesn't matter if you have a quadratic algorithm instead of something better.

Most people I know are exposed to algorithms fundamentals during their university education as part of pursuing a degree in computer science or engineering. You almost certainly aren't going to have academic exposure to algorithms if you are say a liberal arts major - never mind someone who doesn't attend university at all (I also know plenty of terrific programmers who don't have degrees). From my own experience, my degree is in computer engineering. Not computer science or software engineering: computer engineering. I remember from my university days that my computer science friends seemed to have a much better grasp at algorithms and theory of software and programming than I did. When I was taking classes about how hardware and electronics work, they were learning all about the mathematical concepts underpinning the field, different approaches to programming language design, etc. I received very little of that. And on top of that, I struggled with my single algorithms course at university. So I entered the workforce without as good of a grasp on the computer science fundamentals as others I knew. (But I still probably knew more than someone in an unrelated field.) The point I'm trying to make is that because algorithms are somewhat highly specialized and academic in nature, requiring knowledge in algorithms will effectively bias your hiring towards people with strong computer science backgrounds. Stated another way, screening on algorithms knowledge undermines diversity and inclusion initiatives by excluding viable candidates who don't have strong backgrounds in computer science. Sure, if someone wants to enter the industry they can take the time to study up on algorithms. But why force them to do that? It feels like arbitrary gatekeeping given the relative non-importance of algorithms given the typical activities of the typical programmer. So why do it?

I suspect major contributing reasons to why algorithms are so prevalent in interviews are cargo culting, laziness, and lack of formal interview training / caring about diversity. As an industry, the software field is pretty bad at applying best practices and learning from our mistakes. I suspect this will change once the relatively young industry catches up to more-established industries and we're forced to cope with the realities of legal and monetary liabilities the way practically every other industry is. (We're starting to see this with monetary damages for security breaches.) Anyway, we as an industry are pretty bad at self-regulating and adopting practices with proven benefits. We like to settle for what is known. Laziness and the comfort associated with is easy. Seeking out and implementing change is harder. This is human nature. We see this with well-known people in industry rejecting the ideas of continuous testing (years ago) or fuzzing (more recently). We see it in C/C++ programmers who are delusional about their abilities to write secure code and decry e.g. Rust's safety guarantees as superfluous. The industry is disproportionately white and male (at least in the United States). And this brings with it certain personality tendencies. One is a macho attitude, which can manifest in interviews via the interviewer embarking on an ego trip proving they know some esoteric algorithm or data structure the candidate does not.

As a clear example of this, Google was known for asking brainteaser interview questions. (The practice may have been prevalent at Microsoft before Google was the darling of Silicon Valley, but that was before I entered industry.) This trend caught on and soon companies all over were asking brainteasers! The problem was that these questions didn't correlate to actual job performance! From a 2013 NYTimes interview with Google's VP of People Operations:

On the hiring side, we found that brainteasers are a complete waste of
time. How many golf balls can you fit into an airplane? How many gas
stations in Manhattan? A complete waste of time. They don’t predict
anything. They serve primarily to make the interviewer feel smart.

But the damage was done. I still heard these kinds of questions when interviewing in the wild long after Google realized they were bad questions and instructed interviewers not to ask them. I even believe I got a brainteaser when interviewing at Google after the supposed banning of these types of questions! And I won't be shocked if I'm asked a brainteaser in 2019 as part of the several interviews I'll be doing in the days ahead.

Asking questions with no correlation to job performance because a popular company asked that type of question for a while: that's textbook cargo culting. Failing to change your ways despite evidence saying you should: laziness. Insisting that your way is correct and others need to be like you: gatekeeping.

I'm not saying algorithms and data structures during interviews are intrinsically bad and that we should stop asking about them. What I am saying is that we as an industry need to examine how we interview. We need to invest in scientifically proven techniques. (Research shows that behavioral interview questions are better. Tell me about a time when, etc.) And after more than ten years in industry, my experience tells me that interviews place a disproportionate emphasis on algorithms and data structures compared to the daily activities of the typical programmer. And on top of that, due to their academic nature, I worry that screening for algorithms and data structures knowledge is undermining the diversity and inclusivity of our field by biasing towards people with strong computer science backgrounds. I think it is time we examine the role of algorithms and data structures in interviews and consider focusing on other areas instead.

blog comments powered by Disqus