Analysis: There has been tremendous interest in using artificial intelligence tools to streamline hiring decisions in organisations, but the blind use of them without understanding why the algorithm prefers some candidates to others is risky.

In a recent Irish Times column, Pilita Clark argues that it is bonkers to use an algorithm to hire a person. She provided an interesting anecdote to show how the unthinking use of applicants' Uber ratings as a screening tool could lead to bad hiring decisions. Her article was, as is often the case, eloquent in its analysis of what we shouldn’t do when hiring, and markedly less clear about what we should do.

There has been tremendous interest in using the tools of artificial intelligence (e.g., machine learning, predictive algorithms) to streamline hiring decisions in organisation, but also a growing appreciation that these tools can help to perpetuate sexist and racist biases in hiring.

This is particularly true if algorithms are 'trained' on the basis of past practices in an organisation. An organisation that has a long history of hiring white males from the 'best' colleges and universities is likely to find that algorithms based on their previous hiring practices will perpetuate these biases.

The blind use of machine learning algorithms without understanding why the algorithm prefers some candidates to others is risky, and potentially very costly to organisation.

On the other hand, a survey of actual practices in organisation, particularly small to medium-sized organisations, suggests that algorithms might not be such a bad alternative.

We need your consent to load this rte-player contentWe use rte-player to manage extra content that can set cookies on your device and collect data about your activity. Please review their details and accept them to load the content.Manage Preferences

From RTÉ Radio One's Today with Sean O'Rourke, advice On compiling a CV and doing job interviews with Sinead Engish, Career Consultant and Author of 'CV And Interview 101: How To Apply And Interview For Jobs'

Hiring practices vary considerably across organisation, end even more across countries, but there is one hiring practice that is nearly universal – the interview. When done well, interviews can be highly useful tools.

For example, interviews that use panels rather than a single interviewer and that impose structure (e.g., by creating a set of prepared questions that zero in on the skills and experiences that are necessary to perform a job well) can be valid and fair predictors of future success.

Most interviews, however, are not carefully structured, and they leave plenty of room for personal biases and idiosyncratic ideas about what makes a good candidate (e.g., a firm handshake, a particular style of dress), and the track record of unstructured interviews in selecting the best candidates is spotty at best.

An algorithm is nothing more than a simple set of rules, consistently applied. Psychologists have been studying job performance for over a century, and there are some well-established findings that can provide a good basis for valid and fair algorithms.

First, cognitive ability - the ability to reason, solve problems and obtain and use information to make judgments - is relevant in just about every job in the economy. It is more important in complex jobs than in simple ones, but no matter how simple the job, cognitive ability matters.

We need your consent to load this rte-player contentWe use rte-player to manage extra content that can set cookies on your device and collect data about your activity. Please review their details and accept them to load the content.Manage Preferences

From RTÉ Radio One's Drivetime Dr. Tomas Chamorro-Premuzic, Professor of Business Psychology at Columbia University in New York, talks about job interviews and what type of candidates tend to perform well.

Second, conscientiousness (dependability, responsibility, being organised and goal-directed) is similarly relevant across the board. There are few jobs, if any, in which it would be better to hire someone who is not very bright or very dependable than to hire someone who is able to make good judgments, learn how to do their job, and who can be depended upon.

There is a lively literature dealing with the best ways to measure these key traits, but very good and very affordable tests and measures exist that capture these attributes, and they are underused in business. One good algorithm might be 'applicants who are smart and dependable should be preferred to applicants who aren’t'.

It is always a good idea to monitor how algorithms are working and not to surrender total control to them, but I believe a good and simple algorithm will stand you in better stead that relying on the gut feelings of interviewers or hiring managers. Hiring involves making a prediction about the future, and if you are in the business of making a prediction, you will also be in the business of sometimes making an error.

The real question is what type of hiring system is least likely to make serious errors in the long run, and the smart money is on a systematic approach, even it this involves using algorithms.


The views expressed here are those of the author and do not represent or reflect the views of RTÉ