Written by Nathan Hewitt

Is it time to start trusting the machine?

Machine learning outcomes are testable and corrective measures remain consistent, unlike in humans.

“People are not your most important asset. The right people are.” – Jim Collins, author and lecturer on company sustainability and growth.

“Are the right people in the right roles?” [This is] the single most important factor for leadership success and for organisational success.” – Gail Kelly, former CEO of Westpac

How many research papers do we need to read or edicts from top-class CEOs before we get the message that in every organisation, it all comes down to the people?

Adam Bryant who pens the terrific weekly column, The Corner Office, for the NYT has interviewed a diverse pool of leaders, and a common theme from 99 % of his interviews with CEOs is that success correlates with hiring the best team.

My former boss Tracey Fellows, CEO of the REA Group, was also fond of saying that it is ‘people’ that keeps her up at night more than any other business challenge.

Most hiring in most organisations relies 100 % on people to make those most important decisions. Yet we do so with little objective data. Instead, we have layers upon layers of bias! And to give you an idea of how many there are, here is a whooping full Wikipedia list of cognitive biases for you to check out. This article lays out in great detail a plethora of cloudy, smeary and hazy biases I didn’t know could exist.

It concludes that they are mostly unalterable and fixed, regardless of how much unconscious bias training you attend in your lifetime.

There is no scalable, efficient and reliable way to train us out of our biases. Our biases are so embedded and invisible; mostly, we just can’t ‘check ourselves’ at the moment to manage them.

So, how is that diversity hiring program going?

Read: Why a Lack of Diversity is Costing Your Business

In some functions/ departments, your “Hiring for Diversity” may be going very well. However, diversity training and hiring isn’t repeatable, where humans are involved. And, if humans could be trained out of their biases, we may get more diversity in our new hires. But then, do we know that we are getting the ‘better’ hire from the applicant pool? How CAN you tell if you have no method of reliably testing for what matters for success?

You might say we rely on CVs to give us that ‘insight’ but did you know CVs are usually crafted, designed, worded and reworded to ‘best-light’ the applicant. Ever appointed an Excel whizz, who on hire doesn’t know a pivot from a concatenate? Or even worse, who cannot apply logic, reasoning and critical thought?

We have all done this – apply crude (biased) filters to screen applications:

  • Blue-chip companies on their CV – tick!
  • Stayed in their role for two years on average – tick!
  • Promoted at least once inside of a (good) organisation – tick!
  • Good school – tick!
  • Impressive referees – tick tick tick!

Because biases appear to be so hardwired and inalterable, it is more straightforward to remove bias from algorithms than from people.

This gives AI the potential to create a future where important insights underpinning decisions such as hiring, are made more fairly.

The machine can be trained to help you make repeatable and stable decisions.

Read: Why Machines make better decisions than humans (oh and why I hate Simon Sinek)

Algorithmic bias is not the elephant in the room. Some argue that algorithms themselves have bias. The reality is that machine learning, by its very definition, is aiming to find patterns in large volumes of data, mostly latent, to support decisive actions. Removing bias is driven by what bits of training data you use to feed the machine.

You can ensure there is no (or limited) bias in the machine learning and it is all about two things:

  1. What data is being used to build the model?
  2. What are you doing to that data to build the model?

If you build models from the profile of your talent and that talent is homogenous and monochromatic, then so will be the data model and you are back to self-reinforcing hiring.

If you are using data which looks at age, gender, ethnicity and all those visible markers of bias, then, sure enough, you will amplify that bias in your machine learning. Relying on internal performance data to make people decisions, that is like layering bias-upon-bias. Similar to building a sentencing algorithm with sentencing data from the US court system, which is already biased against black men.

So instead of lumping all AI and ML into one big bucket of ‘bias’, look beneath the surface to understand what’s going into the machine as that is where amplification risks loom large.

To ensure you are using machine learning wisely, only use objective data which has no biodata (that means a big NO to CV and social media scraping). Test rigorously and adjust to learn continuously. And, be certain to use multiple machine learning models to continuously triangulate the model versus relying on one version of the truth.

Machines are better at learning this stuff.

Unlike trying to solve human bias, machine learning is repeatable, stable, consistent and most importantly, testable. The value to the organisation is of course, immense.

  • Every applicant gets a fair go at the role;
  • Every applicant is assessed;
  • Hire the person who will succeed vs someone your gut tells you will succeed;
  • Use fewer resources to hire;
  • Reduce the cost of hire.

Now that is ticking all the right boxes. It makes the possibility of objective and valid decisions available at scale, a probability.

Machine learning outcomes are testable and corrective measures remain consistent, unlike in humans.

The ability to test both training data and outcome data, continuously, allows you to detect and correct the slightest bias if it ever occurs.

Soon (maybe already) you will be putting yours, and your loved ones live in the hands of algorithms when you ride in that self-driven car. Algorithms are extensions to our cognitive ability helping us make better decisions, faster and consistently based on data, even in hiring.

To keep up to date on all things “Hiring with Ai” and Machine Learning subscribe to our blog!

You can try out Sapia’s Chat Interview right now – HERE. Else leave us your details to get a personalised demo


This AI Model Can Predict If You Are A Job Hopper Or Not

Voluntary employee turnover can have a direct financial impact on organisations. And, at the time of this pandemic outbreak where the majority of the organisations are looking to cut down their employee costs, voluntary employee turnover can create a big concern for companies. And thus, the ability to predict this turnover rate of employees can not only help in making informed hiring decisions but can also help in saving a substantial financial crisis in this uncertain time.

What Drives Job-Hopping?

Acknowledging that, researchers and data scientists from Sapia, a AI recruiting startup, built a language model that can analyse the open-ended interview questions of the candidate to infer the likelihood of a candidate’s job-hopping. The study — led by Madhura JayaratneBuddhi Jayatilleke — was done on the responses of 45,000 job applicants, who used a chatbot to give an interview and also self-rated themselves on their possibility of hopping jobs.

The researchers evaluated five different methods of text representations — short for term frequency-inverse document frequency (TF-IDF), LDS, GloVe Vectors for word representations, Doc2Vec document embeddings, and Linguistic Inquiry and Word Count (LIWC). However, the GloVe embeddings provided the best results highlighting the positive correlation between sequences of words and the likelihood of employees leaving the job.

Researchers have also further noted that there is also a positive correlation of employee job-hopping with their “openness to experience.” With companies able to predict the same for freshers and the ones changing their career can provide significant financial benefits for the company.

Regression Model To Infer Job Hopping

Apart from a financial impact of on-boarding new employees, or outsourcing the work, increased employee turnover rate can also decrease productivity as well as can dampen employee morale. In fact, the trend of leaving jobs in order to search for a better one has gained massive traction amid this competitive landscape. And thus, it has become critical for companies to assess the likelihood of the candidate to hop jobs prior to selections.

Traditionally this assessment was done by surfing through candidates’ resume; however, the manual intervention makes the process tiring as well as inaccurate. Plus, this method only was eligible for professionals with work experience but was not fruitful for freshers and amateurs. And thus, researchers decided to leverage the interview answers to analyse the candidates’ personality traits as well as their chances of voluntary turnover.

To test the correlation of the interview answers and likelihood of hopping jobs, the researchers built a regression model that uses the textual answers given by the candidate to infer the result. The chosen candidates used the chatbot — Chat Interview by Sapia for responding to 5-7 open-ended interview questions on past experience, situational judgement and values, rated themselves on a 5-point scale on their motives of changing jobs. Further, the length of the textual response along with the distribution of job-hopping likelihood score among all participants formed the ground truth for building the predictive model.

Some examples of questions asked.

To initiate the process, the researchers leveraged the LDA-based topic modelling to understand the correlation between the words and phrases used by the candidate and the chances of them leaving the company. Post that, the researchers evaluated four open-vocabulary approaches that analyse all words for understanding the textual information.

Open vocabulary approaches are always going to be preferred over closed ones like LIWC, as it doesn’t rely on category judgement of words. These approaches are further used to build the regression model with the Random Forest algorithm using the scores of the participants. Researchers used 80% of the data to train the model, and the rest of the 20% was used to validate the accuracy of the model.

Additionally, researchers also experiment with various text response lengths, especially with the shorter ones, which becomes challenging as there is not much textual context to predict. However, they found a balance between the short text responses along with the data available and trained the model predicts for even those.

Model accuracy vs minimum text length in words

To test the accuracy, the models are evaluated based on the actual likelihood of the turnover with relation to the score produced by the model. To which, the GloVe word embedding approach with the minimum text of 150 words achieved the highest correlation. This result demonstrated that the language used in responding to typical open-ended interview questions could predict the chances of candidates’ turnover rate.

Wrapping Up

Leveraging data from over 45,000 individuals researchers built a regression model in order to infer the likelihood of the candidates leaving the job. It will not only remove the dependency of companies on candidate resumes and job histories but also enhances the process of hiring into a multi-measure assessment process that can be conducted digitally for recruiting.

By Sejuti Das, Analytics India Magazine, 02/08/2020

Join the movement

To keep up to date on all things “Hiring with Ai” subscribe to our blog!

You can try out Sapia’s Chat Interview right now, or leave us your details here to get a personalised demo.

Read Online

How the shift to remote work will change who we hire

Eighteen months after we were all forced abruptly to work from home, it seems as the world cautiously opens up and employers are looking to return workers to offices, having the flexibility to work from home is an increasing demand that people aren’t willing to give up.

Earlier this year, Amazon laid out plans for most of its 60,000 workers in the Seattle area to return to the office later in the year. But, it wasn’t good news to everyone with hundreds threatening to quit.  Microsoft, at Redmond in California, took a softer approach saying employees could work from home, the office or in a hybrid arrangement. Google, Hubspot and Intuit are some of the other companies that have opted for hybrid models going forward.

Others like Atlassian, Twitter, Shopify, Spotify and Slack have decided to become fully remote.  Recently, Slack CEO Stewart Butterfield declared that digital life has moved too far forward during the COVID-19 pandemic for companies to return to former ways of office-based working, even if they wanted to.

While these are some of the world’s most influential companies, it’s a conversation that almost every employer is having right now.

The reality is the demand for remote or hybrid work is fast becoming part of hiring negotiations and compensation packages. For many, work flexibility has become more important than pay. 

This has created a new dilemma for hiring managers that’s much deeper than offering strong commitments on flexibility as part of a job offer. 

If work flexibility is the future, how do you determine who’ll thrive in this setup? 

While it’s easy to guess some of the ideal attributes of a remote worker – that is they need to be autonomous, self-motivated, productive and able to collaborate online – there is another key characteristic that has proven vital to strong performance. 

What we’ve seen from companies that have prioritised remote working for a long time such as Automattic, GitLab, InVision and Buffer is the importance of strong written communication. This is because you are no longer relying on face-to-face interactions that occur naturally or through formal meetings in an office. For remote work to be viable communication needs to be predominantly textual and mostly asynchronous. 

When building a remote organization, Automattic CEO Matt  Mullenweg has said that at some point you realise how crucial written communication is for your success, and you start looking for great writers in your hiring. For this reason, Auttomatic job interviews are conducted via text only.

Mullenweg says the true asynchronous nature of a written interview reflects the remote work reality compared to real time video interviews, which are not scalable in an organisation. I think most of us found that out the hard way during the pandemic. 

So how can companies hire for remote capabilities?

In order to be effective remote and hybrid companies we need to rethink our hiring processes. To be frank, current hiring practices are just not going to cut it. CVs do not reveal the soft skills we need them to, and video is so inherently biased and stressful for candidates that many companies which opted for this early on in the pandemic are abandoning it as a top-of-the-funnel filter. We have several customers who have explicitly ditched video interviews.

The risk of making bad hires when you throw remote work into the equation is higher than if you’re bringing people into an office environment. You need to trust them from day one without any of the  ‘visibility’ you get from seeing someone everyday.

We need a new way of selecting candidates that can accurately identify soft skills like accountability,  autonomy, drive and writing skills. Can a text based interview reveal these qualities, while providing a great candidate experience and being highly relevant to the remote work context?

Mullenweg’s idea of a text only interview is not as radical as some might believe. We do thousands of them every day across the world, for a number of varied companies. We are able to reveal people’s character traits with over 90% accuracy (we know because we ask them).

It’s scientific, based on data and is the only accurate way you can identify both the written communication proficiency and soft skills required to work remotely.

How we do it

Our text interview includes open-ended questions on situational judgement and values, similar to a structured interview. When responses are analysed for skills that pertain to remote work it takes into account a multitude of features related to language fluency, proficiency, personality traits, behavioural traits, and semantic alignment. 

This allows a recruiter to quantify and compare a candidate’s written communication skills immediately as well as their suitability to the work environment. 

The revealing nature of text interviews is not just limited to the skill of writing, but also to the motivation behind expressing something in writing, which requires more effort and thinking than speaking it out. If someone is not motivated to express themselves in writing when a job is on the line, you can assume what it might be like once they are working in a role. 


While many companies are already scrambling to update their remote work policies and rethink their office space needs, if they are not reconsidering their hiring processes as part of this inevitable shift, then they are exposing their company to risk. 

Just because people want to work remotely, doesn’t always mean they can thrive in it. While you may be doing the right thing in offering flexibility for candidates, you also need to make sure that you are doing right by your company by understanding how well these candidates will thrive remotely.


Wondering how our Ai recruitment assessment tool can identify soft skills in candidates? Find out here.

Read Online

How to write good job ads, optimised for candidate experience

How to write a good job ad | Sapia Ai recruitment software

We agree with Katrina Collier: Recruitment isn’t broken, per se. It needs a bit of work, sure, but in the midst of the Great Resignation, dedicated talent acquisition managers all over the world are doing some of their best work. They’re finding top talent and helping businesses succeed.

Despite this, we can say that candidate experience is certifiably broken. Ghosting rates are up somewhere around 450% since the start of the pandemic. 65% people say they rarely receive notice of their application status (Lever), and 60% of people say they have bailed on a job application due to its length or complexity. 

Why candidate experience is important

Many mid-to-large sized companies spend in excess of $200,000 per year on sourcing and advertising (assuming a hiring rate of fifty people per year). Few invest in candidate experience. We tend to overlook the fact that the candidate journey from application to offer (or rejection) is just as important for the health of a recruitment funnel, over the long term, as good ads or recruitment strategies.

Good candidate experience, put simply, is your best chance at securing the talent you want. In the wake of the Great Reshuffle, employees have the power to choose when and where they work, and they know it. If you can’t reach them and woo them in a reasonable time frame, you’re at a supreme competitive disadvantage. They’re here today, gone tomorrow. That means that multi-round interview funnels and tedious psychometric games aren’t going to cut it anymore. Today’s candidate wants speed, perks, and flexibility. Your experience should be designed with this in mind.

There are a lot of ways candidate experience might be improved – this article offers some tips, including advice on a term we like to call the Gucci principle.

One easy place to start is with your job ads.

How to write a good job ad

Good job ads are concise and well-formatted. They put employee value proposition up front. They discuss the vision and purpose of a role, and not just day-to-day responsibilities. They avoid the term ‘competitive salary’ – in fact, they disclose salary ranges. They’re not necessarily short, either. Anyone who tells you that a job ad must be short to be good does not understand the anatomy of an advertisement.

Here are our top tips.

1. Make sure the spelling and grammar in your job ad is perfect, throughout

This seems like a minor point, but good spelling, grammar, and sentence structure is essential for your employer brand. It’s a matter of perception. Poor writing casts doubt on the legitimacy of your brand, and on your capabilities in general – after all, if you can’t write a clean job ad, how can the candidate be sure you can do other, more important things, correctly?

Have someone in your marketing team cast their eye over your ad before it goes out. Proof-reading should always be a part of your customer outreach. If you don’t have a marketer on which to rely, consider investing in editing software like Grammarly.

2. Keep the unique language of your brand

Funky company names are in vogue. Just look at ours. Because we’re called Sapia, we refer to our team (and even our customers) as Sapians. Therefore, we do the same with our job ads. It creates branding consistency, and works as an unconscious primer, suggesting to candidates that they’re joining a well-knit, stable, and purpose-oriented team. 

The same goes for language. If you’ve adopted or created certain words to make your brand stand out, they should also be used to make your job ad stand out. Look at this example from Gong: They tell the candidate that they’ll be creating edu-taining content. That’s a lot more interesting than “you’ll be writing content that is both educational and entertaining.” Had they chosen the latter sentence, you’d doubt their credibility, because that sentence is not remotely entertaining.

Gong job ad example

Or take this example from one of our own job ads. You might say that using a curse word (oh dear me!) in a job ad is inappropriate, but we don’t. We’re Sapians, and that makes us passionate humans. We understand that writing the way you speak is the quickest way to build rapport. Tell us that you don’t get that impression from this paragraph.

3. Clear categorisation and formatting of sections

A job ad doesn’t need to be short, but it should be formatted for scanning. Candidates should be able to easily read it, extract the main points, and make the call to apply, all within minutes. We like the following job ad section structure:

  • Perks and benefits
  • Responsibilities
  • Qualifications

Each section can be as long as you need it to be (within reason), but it should also be set out in dot points. Easier to read, easier to digest. Many are the job ads that set out position duties and benefits in great big walls of text. Go with dot points, like Gong has, and you’ll stand out.

4. Make it as easy as possible to apply

Depending on the platform you use, it can be difficult to control how candidates enter your funnel. Regardless, you can make it easier by clearly sign-posting the action you expect them to take. If it’s a LinkedIn EasyApply button, great – but don’t confuse candidates by asking them, at the bottom of the ad, to email their CVs to you. This happens a lot.

Make sure you have a single call-to-action, and make it clear. Add it to the top and bottom of your ad. 

You know what they say about first impressions? That’s why it’s so critical to get your job ads right. Check out this post on LinkedIn for more tips on writing the perfect job ad.

Read Online