4 min read

Artificial Intelligence - Is it foolproof?

Natasha Harries
Natasha Harries January 14, 2021
Is artificial intelligence foolproof?

Artificial Intelligence - Is it foolproof?

Employers are increasingly reliant on AI, on algorithms, on the technology, to tell them whether or not a candidate should advance during an application process. There are huge benefits of using an AI tool to assist your recruitment function; eliminating unconscious bias is arguably the biggest. Yet, when the AI is being designed and built by a specific demographic, are their creations then inheriting bias?

There are increasing calls to protect employees rights in the ‘digital age.’ Using hiring algorithms that offer no flexibility, no insight or even explanation behind certain conclusions leave candidates vulnerable. They can be struck from an application process with no feedback given and no real understanding as to why.

Can an algorithm find the ‘best’?

In 2014, Amazon began developing an AI to identify the best CVs submitted for advertised positions. Amazon’s idea was to automate a process, increase efficiency and continue to automate workflow as much as possible - similarly as it has done with other aspects of its business, allowing it to be the operational ‘powerhouse’ we know it as.

However, in order for the AI to do this, it had to be given data to learn from. So it was given years worth of previous successful CVs to then learn what ‘worked’ for Amazon hires in the past. What seems obvious now came as a bit of a surprise. The computer taught itself that the best CVs came from men. It learned that CVs which included the word women, for example ‘womens’ club’, had not been selected in the past and therefore, should not be identified as a ‘best’ CV in the future. Amazon of course had to abandon this project.

Whether intentional or not, our ability to detect bias of an algorithm or an AI is extremely difficult because it can occur at any stage of the development.

Can a psychometric test be biased?

Numbers are universally recognised, in every language, in the same digit formation. It is therefore much simpler to avoid bias in numerical assessment designs. But what about verbal reasoning tests? Or situational judgement? Now more than ever before, relocating for work, deciding to emigrate because you are fed up of constant rain and having offices in various locations across the world is increasingly common. As a result, adding the requirement to speak multiple languages to an application has also seen a rise in demand. In the United States alone, the demand for bilingual employees has more than doubled year-on-year since 2010.

Are psychometric tests biased against candidates who have a different native language to that in which the test is written? Do verbal reasoning tests become nothing more than translation tests? Can it provide an accurate insight into a candidate’s ability to read, interpret and react to written information? Or does it become a test of that candidate’s ability to translate? These are all very valid questions.

It seems strange that a business that has built its entire success on the use of psychometric testing would be discussing the potential dangers of ingrained bias within AI. Well, the good news is that psychometric testing, when used properly, remains an immensely powerful tool when making hiring decisions.

Tool - A device or implement, especially one held in the hand, used to carry out a particular function.

Too often, we rely on the technology around us and seem to forget that although it is intelligent and provides drastic operational streamlining in both our professional and personal lives, it was ultimately built by us and is therefore vulnerable to human error.

The solution is simple - When it comes to using psychometric testing, designed and built by an algorithm or AI, they are to be used as just one tool in a recruitment belt. Psychometric testing produces useful data to assist in your hiring decision. You can measure cognitive abilities, lateral thinking skills, reasoning and even their stress tolerance levels. They can be a useful ‘filtering’ device, they can be used mid-way through your recruitment process, perhaps after a second stage interview. Or even post-hiring to assist with training or continuous professional development.

They are not a replacement to an interview or conversation. They are another level of assessment and a further layer of thorough understanding of a candidates’ competence, to be used alongside that ‘human’ interaction.

Some people call this artificial intelligence, but the reality is this technology will enhance us. So instead of artificial intelligence, I think we'll augment our intelligence.” Ginni Rometty, IBM.

Natasha Harries
Natasha Harries January 14, 2021

Enjoy what you’ve read? Let others know!