What is blind hiring?

Kate Glazebrook

27

November

2018

|

8

min read

X

We're rebuilding the hiring process.

Want to see more content like this? Subscribe to our fortnightly newsletter.

Give me more

Updated 27/01/2021 with additional research and steps for setting up a blind hiring process.

'Blind' or anonymised hiring refers to the technique of removing personal information from candidate applications during the assessment process. This is most often candidate names, but in some instances, like on Applied, it goes further: aiming to remove all socio-demographic details that might affect decision-making, such as age, home address, and even education.

Blind CVs have emerged in recent years as a popular solution to the fact that (even if we don’t realise it) we respond to candidates differently when we know details about them, and that can stop us from being fair. Blind CVs can help us overcome some, if not all, of the unconscious biases that often result in people from more diverse backgrounds being overlooked for the job.

Blind or anonymised hiring
Names have been shown to trigger a whole host of unconscious biases

The fact that my name is Kate may tell you a bit about my parents’ preferences for names (including my mother’s ability to steer my father away from some more questionable choices!), but it will tell you next to nothing about my ability to be a capable construction worker, astronaut, or postie.

But research tells us that you knowing my name is Kate when you’re assessing my fit for a job might influence you unconsciously: perhaps you had a bad experience with someone called Kate in the past, or maybe you’re just not used to women in those roles and unconsciously I don’t seem to fit the mould.  

Biases can work positively too, perhaps your daughter’s name is Kate, and you unwittingly feel a little more affinity for me as a result.  Either way, it’s not telling you anything about whether I’m the best candidate for the job, so we can save you the second guessing by removing the distraction.

Why do names (and other details) affect hiring decisions?

Our brains are designed for efficiency; they seek to find patterns in our experiences that help us to make faster decisions.  These can take the form of ‘heuristics’ that act like mental shortcuts or rules of thumb and which we take in our automatic and instinctive ‘System 1’ brain.  

Most of the time, these heuristics make our lives considerably better: the world is full of informational inputs, and heuristics can help us chart a faster path toward a decision.  

We’d never get anything done if we deliberated from first principles about every choice we had to make using our alternative, analytical and conscious System 2 brain.

System 1 vs System 2 Thinking

System 1 brain: Used for quick decision-making, relying on mental shortcuts and associations - e.g. waking to work

System 2 brain: Used for one-off, more considered decisions - e.g. managing work project

The trouble lies when our brains jump to conclusions too quickly or draw on misplaced patterns.

We have a habit of using our System 1 brain when we should be using System 2.

 Unconsciously, we form impressions of the world around us that sometimes means we fail to see it on its merits, and this can be particularly the case when we’re cognitively overloaded like when we’re tired, time-poor, or have lots of choices to make.  

As some researchers explain: “Decision-makers who are rushed, stressed, distracted, or pressured are more likely to apply stereotypes – recalling facts in ways biased by stereotypes and making more stereotypic judgments – than decision-makers whose cognitive abilities are not similarly constrained.”

How does this relate back to hiring?  Well rough estimates suggest that recruiters and hiring managers spend as little as 10 seconds reading your CV or resume before they decide whether they want to bring you in for interview.  

And we know that for many roles, there are upwards of 100 candidates for every job advertised.  

That’s a lot of quick thinking going on, which can open up opportunities for inadvertent bias to creep in. In fact, studies have consistently shown that these precious 10 seconds can be critical.

CV bias breakdown
Just a few of the biases that affect a CV screening

Researchers the world over have now studied how small, irrelevant personal details on a candidate’s application can alter their chances of being selected to go through to the next stage.  

The set up on these types of experiments - often referred to as correspondence studies - is quite simple: Researchers create a fictitious (but realistic) candidate CV and then send it out to thousands of real employers.  

But before they do so, they subtly alter one or two small details on the CV so they have created two or more nearly-identical CVs, with the only exception being a name or other detail.  

They then randomly assign different variants to different employers.  And then they sit back and wait for these prospective candidates to get calls to attend an interview and they measure the extent to which these small variations affect call-back rates.

One such study, done in Germany in 2016, looked at the extent to which ethnicity and religion - signalled through the name and the photo on the application - influenced a candidate’s likelihood of getting an interview.

They randomly assigned 1,500 real employers to receiving an (otherwise identical) application from Sandra Bauer, Meryem Ӧztürk, or Meryem Ӧztürk when she was wearing a headscarf.  

The names were specifically chosen to evoke either a person with a German heritage (Sandra Bauer) or Turkish heritage (Meryem Ӧztürk).  They found - astonishingly - that while Sandra was invited to interview 19% of the time, when the exact same application was received from Meryem without a headscarf, she received a call back just 14% of the time, and when Meryem was wearing a headscarf, she got a call back just 4% of the time.

Meryem Ozturk study
Average rate of call back for interview

Their study even compared different levels of discrimination based on job type.  They found that an equally-qualified Muslim Turkish woman would need to send 3.5x more applications to get the same rate of call back from a secretarial role, but a whopping 7.5x more for a chief accountant.

Studies like this one have now been done in almost every continent of the world, and have explored the impact of subtle cues of gender, race and ethnicity, caste, obesity, sexual orientation, age, and even home address (signalling wealth) can affect interview call back rates, and found rather depressingly that minority groups tend to be systematically overlooked in hiring processes.

(Side note: similar studies have also replicated findings in other markets such as rental housing, retail or even AirBnB).

A study conducted in the US found, for example, that a person with a ‘non-White’ sounding surname would need to send approximately 50% more applications out to get the same rate of call back for interview, or another way of thinking about it is they would need to have almost 8 more years of job experience to get the same response rate.

White-sounding vs Black-sounding name bias


And if we agree that these sociodemographic factors are rarely - if ever - relevant to whether you can do the job at hand, we can agree that this represents a huge potential loss to individuals and to the teams who miss out on vital diversity.

Another similar study from here in the UK had similar findings...

Adam vs Mohammed Bias
Researchers found that candidates with Muslim-sounding names are 3x more likely to be passed over for a job.

Does blinding hiring work as a recruitment strategy?

It’s fair to say there have been more studies proving the existence of discrimination than there have been studying how to remove it.  

But an obvious outcome of the correspondence studies is to remove access to any information that might inadvertently influence a decision.  

So far, the evidence suggests that removal of names and signifiers of ethnicity can help to level the playing field, particularly in teams that don’t positively discriminate or have a strong affirmative action policy in place (which is most employers).  

For example, a recent set of studies coordinated by the Victorian Government concluded that: “The result of the five pilot studies provide strong support for and useful insights into the broader implementation and further evaluation of CV de-identification in public and private organisations in Victoria”.

Interestingly, their studies showed that the impact of de-identification, or blinding, differed across job roles and organisations.  

Depending on the setting, it had the effect of increasing the success rates - and reducing or even sometimes reversing the success rate gap - for women, non-Australian-born candidates, and those from lower socioeconomic backgrounds amongst other groups.

These effects were not uniform and seemed to relate to different contextual challenges in each setting, with occasional evidence that the effect increased the success rates of non-minority groups in parallel.  

But even in studies where the effects have been more mixed - there is almost always evidence that blinding affects the shortlisting process.  

At a minimum, that’s an indicator that whether we realise it or not, we’re treating candidates differently on the basis of their name or some other factor, which ultimately corrodes the fairness we expect from job application processes.  

And given the wealth of studies pointing to the presence of discrimination against minority groups, it’s likely that the average effect of removing these potentially distracting details will be positive for equality of opportunity overall.

OK great, so I just blind applications and everything will be fine?

Sadly no, while the evidence tells us that blinding will help, it’s still only one piece in a more complex puzzle of how to truly rid hiring processes of potential bias.

For example, you can blind a process all you like but if you’re still selecting candidates on how many all-men’s weightlifting championships they’ve won then you won’t make many inroads into your diversity.  

However, it is a great first step to start off with and we've put together this blog on How to anonymise CVs to show you what to do.

Also if you're keen for more info check out our resource centre for a whole bunch of guides and best practices on de-biasing your recruitment further.

A simple 2 step process for true blind applications

Step 1: Forget CVs altogether in favour of 'work samples'

So we've already established that CVs allow for bias to affect decision-making.

But once you have a blind CV having removed names, addresses etc - what are you left with?

Education and experience.

Most hirers believe that education and experience are key factors in someone's ability to do the job.

But the evidence simply doesn't support this.

Below are the results of a landmark meta-study, looking at the predictive power of assessment methods.

Predictivity of hiring methods

As you can see, education and experience are not vert predictive of real-life ability.

It's no secret that the most privileged end up attending the best schools and universities.

And since we know that those from underrepresented groups tend to be overlooked - how can they be expected to attain the experience that we look for?

Education and experience do tell us something about candidates (hence their predictive validity is higher than 0), but they're just a proxy for skills.

Extensive experience alone doesn't make someone the best person for the job...

But skills learned through experience do.

At the top of the chart above you'll see 'work samples.'

Work samples take parts of the role and turn them into questions or tasks. They’re designed to simulate the role as closely as possible.

To create a work sample question...

  1. Start by defining the core skills required to do the job.
  2. Think of a real-life task or scenario that someone in the role would encounter that might test one of those skills. 
  3. Turn the scenarios into a work sample by posing them hypothetically (by asking 'what would you do?')
  4. Repeat to create 3-5 work samples that test each skill

Depending on the situation, you can either ask candidates how they’d approach the task, or simply ask them to perform it.

You might ask candidates how they'd prioritise a list of jobs or deal with an issue a customer is having.

Or you might simply ask them to perform a task - like drafting an email or blog post.

Here’s an example of a work sample we used for an Operations Manager role:

You have been helping the marketing team to organise a diversity event for 250 people at a venue in central London. Many of Applied's clients and partners will be there, as well as the press. 

One week before the event is due to take place, you get a voicemail and an email from the venue telling you that they have accidentally double-booked the room you had reserved. They offer you a slightly smaller room that will seat 200 in another related venue nearby. 

What actions do you take?

Work samples essentially get candidates to act as if they were already in the role.

Hirers don't have to work out whether not candidates can do the job based on their background.

They simply have candidates tackle small parts of it.

Using this process to hire our own team here at Applied, we found that we wouldn't have even met 60% of hires using a CV process.
Work sample cheatsheet


Step 2: Use a structured interview process

When meeting candidates face to face (or via Zoom), there's going to be some degree of interview bias - unless you're going in blindfolded.

That being said, there are measures you can take to keep interviewing as objective and predictive as possible.

A structured interview (the second most effective assessment on the predictive validity chart) is where all candidates are asked the same questions in the same order.

As for the interview questions themselves, the best practice for blind recruitment is to ask work sample-style questions, rather than delve into candidates' backgrounds.

Interviews offer an opportunity to test skills that can't be conveyed by just writing...

You can see how candidates would think through problems should they actually get the job.

You could ask candidates:

  • To give a presentation (you'd obviously need to give them the relevant details ahead of time)
  • To draw some quick insights from some fake data
  • To look at your real-life strategy and critique it (for senior roles)
  • To role-play a client meeting

Tasks like these enable you to see how candidates would perform in the role, without having to dig into their work history.

If someone can do the job to a high standard, does it matter what their experience is? If done right, your blind hiring process will leave you confident that the candidate you hire has all the necessary skills, having testing them objectively.

Even if you implement all of the above - your hiring will still be swayed by bias unless you give yourself scoring criteria to judge against.

Before you broadcast your open role to the world, make sure every work sample and interview question has its own scoring criteria.

Simply outline what a good, mediocre and bad answer might look like.

And then turn this into a 5 star scale like the one below...

Interview question criteria

In terms of your interview panel, you should have three reviewers.

This is to harness the power of ‘Crowd Wisdom’ - the rule of thumb that collective judgment is more accurate than that of an individual.

The more diverse your panel, the less biased the scores.

Why three? This number has been found to be the most optimal - adding any additional reviewers will bring diminishing returns.

Putting it all together

Using a blind recruitment process like the one above will enable you to make decisions based on data, not gut instinct (otherwise known as bias).

Adding up scores will enable you to build candidate leaderboard - so are grounded in actual performance, rather than preference.

You can set up a blind hiring process without spending a penny. You don't need technology like Applied.

However, we built the platform to make blind hiring as simply and de-biased as possible.

Applied automates feedback and uses behavioural science to remove ordering biases - as well as speeding up the process outlined in the two steps above.

scaling fast guide

If you'd like to see how the Applied Platform works, please feel free to request a demo and we'll give you a tour!