Better hiring with less bias

When I worked at Google, all employees were encouraged to take an unconscious bias test to help us uncover and reflect upon our personal biases. The test covered areas such as gender, race and sexual orientation. The test I took is very similar to the “Sexuality IAT” test at this website (near the bottom of the page):

Project implicit social attitudes tests

The effects of uncovering my own unconscious biases were astounding to me. As a female engineer, I expected to have wonderful results. So I was shocked when the test suggested that I was moderately biased against other females.

After some introspection, I became aware that I tended to take a harsher stance on female interview candidates. I concluded that I held this bias because women are so underrepresented in my industry that I felt that every female engineer was a representative. Any female engineer who was less than exceptional risked being the one female engineer that her male coworkers knew, colouring their opinion of all female engineers in the industry. This may have been a terrible attitude, but remember that this was an unconscious bias, built on my own experiences of male coworkers saying disparaging things about the skills of female engineers over the course of my career. Once I was made consciously aware of this bias, I was able to notice it bubbling to the surface so that I could choose not to let it colour my perspective in future.

Unconscious bias in hiring is not just an ethical problem. Anyone who has borne the arduous responsibility of hiring employees knows how the costs of a bad hiring process can stack up terribly. Allowing poor candidates to progress through the system wastes valuable employee time, not to mention the financial and emotional costs associated with bringing the wrong person on board only to have to fire them within months of the hire. Losing a good candidate is equally costly, especially for positions that are difficult to fill.

As a manager at Google, I built a test and infrastructure engineering team in a department that had previously had only one software engineer supporting this function. As my team’s reputation grew, my client base rapidly grew from a small team of about 20 engineers to over 150. Hiring Test Engineers and Software Engineers in Tools and Infrastructure is extremely difficult – hiring Google-level programmers is difficult enough, and there are few engineers that have even heard of these roles, let alone specialise in them. Even with recruiters at my disposal, hiring took up a large percentage of my management time, while much of the rest of my time was spent acting as technical lead for every project in my team until I could find a suitable successor. Needless to say, reducing the costs of hiring was in my best interests.

After conducting one of my first interviews at Google, I left feeling moderately confident in the skills of the candidate. Simply put, I liked her. She seemed shy and not entirely confident and stumbled over some of her answers, but I overlooked that and saw her potential. Then I sat down to enter the details of the interview in the assessment form.

I quickly realised that the interview had not gone well at all and this candidate did not meet the bar. I liked this candidate as a person, and identified with her values, but her skills unfortunately did not meet the requirements. As I contemplated this, I realised that if even if I somehow managed to hire her, she would have a poor chance of success at her current level, resulting in a bad experience for her and a high cost for myself and my team. I realised that my desire to hire somebody for my team was influencing my desire for the candidate to succeed despite evidence that she lacked the necessary skills.

These experiences have influenced my future hiring processes in the following way:

  • I will always write down my minimum requirements for a candidate in detail and review the checklist after an interview.
  • I will make sure that my interview questions are directly relevant to my requirements for the candidate, and that I cover all of the requirements
  • I will always get a second opinion.
  • I will remain mindful of my unconscious biases
  • I will constantly evolve my hiring process

For more information about unconscious biases at work, see https://rework.withgoogle.com/subjects/unbiasing/

5 thoughts on “Better hiring with less bias

  1. This is very thoughtful. I’m currently working on an algo-trading system, and a significant part of that is removing subjective, emotional triggers from the process.

    Having a mechanical process to follow certainly eliminates some biases, but hiring is also qualitative, not only quantative, and I have every sympathy and empathy for the difficulty that imposes.

    Have you got any methods to eliminate biases across multiple team members? For example, how will you stop your teams becoming monocultures?

  2. Hi Andy,

    Thanks for the question, it’s an interesting one. The link I posted links to workshop resources designed for multiple team members, so that could help to eliminate unconscious biases for multiple team members.

    I think it’s important for any team to define what cultural values are core to that team. Having a feeling that someone isn’t the “right fit” culturally can be an obstacle to hiring, but recognising the reasons why can help to illuminate whether hiring that person is going to harm or enhance the team culture. For example, if a candidate doesn’t seem like the “right fit”, but they seem to align with the team’s values of candidness, integrity and humour, then perhaps biases are at work here. Perhaps the candidate doesn’t seem like a good cultural fit due to age, gender or race? Clearly defining which elements of the current culture are mandatory and which ones are flexible can help identify bias in the cultural aspect of the hiring process.

    I know that you mentioned that this sounds quite mechanical, but I think of it more as a list of reminders rather than a strict process. The human brain needs a bit of prodding to think from different perspectives and examine itself, and a checklist is a handy tool to help your brain work optimally. I’m a big believer in trusting your gut, and if there’s still something telling you “hire” or “don’t hire” despite what your process tells you then it’s a great signal to obtain more data and re-evaluate.

  3. We’ve been doing a bunch of work on how to design the optimum interview process, starting from the current behavioural science literature on bias, and you’re actually pretty close to being right on the money with the way you’ve handled interviews here.

    Instead of a checklist we’ve opted for a set of questions and assesment criteria, with each interviewer writing notes and assigning a score on each element… but it addresses the same problem as your checklist… i.e. how can you ensure that you’re judging candidates consistently, and not letting performance or biasing traits affect your evaluation.

    We’ve made heavy use of blinding during pre-interview assessment, which we’re finding works wonderfully. Google’s a tricky one in that respect because the entire packet is sent to the hiring committee so it’s difficult to blind. We’re also looking at confidence bias, which is a huge skew in tech interviewing.

    Anyway, great to hear more people thinking about this stuff… so mainly just wanted to say Hi :)

  4. Hi Rich,
    Your work sounds very interesting. Yes, “checklist” is an over-simplification – at Google I was using a set of questions and assessment criteria with scores and comprehensive notes like you have described. For smaller companies and teams without much process in hiring this can sound a bit heavy and daunting, so in terms of initial advice I thought I would start by recommending a checklist as a fast way to apply these concepts.

    We’ve made heavy use of blinding during pre-interview assessment, which we’re finding works wonderfully. Google’s a tricky one in that respect because the entire packet is sent to the hiring committee so it’s difficult to blind. We’re also looking at confidence bias, which is a huge skew in tech interviewing.

    It sounds like you are very familiar with Google’s process – have you been involved in its evolution? I would love to learn more about what you have said here. Can you point me towards any online resources you have published? Or can I pick your brain over Skype sometime? :)

  5. Great post Trish. In my history of conducting interviews, I also have a list of minimum requirements. However, there is something I tend to look for in candidates when interviewing. I try to talk about a subject that they know more about then myself. This gives me an opportunity to learn more about they way they think, the way they explain and in general the passion they have. This said, I can see how this may also lead to possible issues about letting my drive to find passionate people hide the requirements of the role at hand.

Comments are closed.