Understanding Responsible AI in Recruitment.
Barb from sapia.ai here!
Welcome to my latest AI for Good newsletter. Despite my ambition to do this fortnightly, it ends up being more irregular. Thanks for tolerating the random release dates, which usually coincide with a long-haul flight.
Anything in particular you want to hear my thoughts on? Drop me an email and I might use it in my next newsletter
If you missed my previous newsletter, I shared a fable that is close to the truth still for the vast majority of candidates and recruitment teams.
A while back I dared to compare the experience pf applying for a job to the experience of applying for a home loan. You can read it here . People decisions, like most decisions are risk-laden decisions. Much like lending money. What can HR learn from how banks make lending decisions? A lot IMHO.
In this newsletter I am going back to basics, since we are all on a learning journey here to confidently navigate AI in the people space and do the right thing by our people (candidates, employees and people leaders) as well as the organisation. If you are looking for simple frameworks and tools to kick off that learning journey with our team, we have saved you the search time. Take a look at both of these papers which have simple frameworks to bring to the table.
?
What does AI in recruitment mean?
AI in hiring is the use of algorithms or other computerised processes to evaluate candidate data automatically or used to interact with candidates in service of progressing the candidate through the recruitment process without human involvement for some or the entire portion of that process.
Certain low level computerised processes such as filters have been used for many years to simply exclude candidates who did not meet certain base requirements.
These processes were generally decided by and deployed within a company for their own hiring purposes.
But the use of machine learning algorithms that combine types of data (behavioural, social, text, audio, video) that was not used before in new ways is where the innovation is happening.
AI compared to Psychometric testing
When it comes to making the best hiring decisions in customer facing roles, we all know that personality counts, but how do we efficiently and fairly test for that? And do we know what kind of personality really does drive better business performance?
What we know from specialising in this field is that combinations of certain personality traits have higher impact on desired outcomes, be those turnover, sales performance, or other business KPIs.
Predictive analytics goes further than the approach adopted by psychometrics testing. Post-hiring performance data is ingested, and using this objective data the machine identifies the patterns between the candidate’s personality profile and performance. Those patterns are always changing. Each time someone is hired, their performance profile is ingested into your data model teaching the machine to find the most recent relevant correlations. This makes predictive analytics an incredibly efficient way of dynamically updating and customising the ‘ideal fit’ recruitment profile.
领英推荐
Why machines help humans make better decisions
Cognitive bias exists in people. Wikipedia’s “List of cognitive biases” contains 185 entries. Decades of research have shown that traditional interviews are riddled with implicit & explicit bias, and are inconsistent.
Machine learning outcomes are testable and corrective measures remain consistent, unlike in humans.
It is more straightforward to remove bias from algorithms than from people - a machine has no feelings; therefore, it is going to be free of the bias’s humans bring to this critical decision. Machines are more malleable to learning and way faster at it. This is more critical when roles are changing dynamically and swiftly as industries are disrupted.
The datasets used to build our algorithms do not contain any information about race, age or gender. Nor do they contain educational or professional background. Bias is removed by focussing on performance.
The business benefits are:
● You use less resources to hire
● Every applicant gets a fair go at the role
● Every applicant is interviewed
● Hire the person who will succeed versus someone your gut tells you will succeed
Organisations typically have little data on personality traits from those who are hired and subsequently perform. Psychometric testing, whilst a statistical methodology for identifying whether someone is likely to be successful in a role, is inferior to predictive analytics because it is never validated by actual performance. Plus, it has the disadvantage of being based on global norms which may not be applicable to your business or region.
Predictive analytics enables companies to take a big step forward in building individual and role-based data DNA profiles that then equips HR to make the right investments in people to optimise for hiring and retaining the best.
?
Here’s a working example:
A sapia.ai customer operates outbound Contact Centres. Using more than 200,000 data points — spanning individuals, markets, and the operational performance of each call centre person — the company set out to find which variables corresponded most closely to retention and to sales optimisation. We used this data to build a series of logistic-regression and unsupervised-learning models that could help determine the relationship between drivers and desired outcomes (retention, call handling time, speed to productivity, customer satisfaction, and speed and sales growth by office).
The algorithm uniquely defined performance in this business to potential performance for those applying for roles. The model successfully identifies those applicants most likely to perform and stay longer and automatically recommends them for the next stage of the recruiting process. This released time for recruiters and now with a clearer field, they could focus on interviewing from the pre-selected pool of applicants to appoint.
Stay curious!
#aiforgood