In 2014, Amazon conducted an experiment to automate its recruitment process. Amazon developed an algorithm and taught it to scan the web to find CVs of suitable candidates with specific skills, using CVs Amazon had received over the last 10 years as the basis of what to look for.
In 2015 though, Amazon identified a flaw in the algorithm. For technological roles, the algorithm learned to ignore what it had been taught and was no longer on the hunt for candidates with the specific skills. Instead it taught itself to pick candidates which used verbs favoured by men and ignored those which indicated the candidate was a woman, for example because the candidate went to a girls’ school.
However much the specialists tinkered with the algorithm, they could not guarantee the machine would devise other ways of picking male candidates over female ones. In other words, they couldn’t get the machine to stop itself from discriminating against women.
What’s the cost?
Technological advances have driven Amazon’s dominance in the market, from its warehouse function to its pricing methodologies and soon, its method of delivery. Its recent experiment with artificial intelligence demonstrates the risks those technological advances sometimes bring.
Amazon pulled the plug on its experiment soon after making its findings and has said that its recruiters never relied solely on the algorithm’s rankings. Had it not done so though, how many candidates may have been overlooked for a job opportunity? This would have given rise to a collective indirect sex discrimination claim as the algorithm’s code disadvantaged women when compared to men. The fact there was no intent on Amazon’s part would not have mattered.
Perhaps even more damaging in the long term would have been the talent Amazon could have inadvertently lost out on.
Do humans do a better job?
Artificial Intelligence is at a relatively infant stage in its development and is only so good as the information it is fed by its programmers. Failing to recognise its discriminatory behaviour without correction demonstrates a continuing need for human and regulatory intervention.
However, there is an argument that humans fair no better in the recruitment process. Sometimes our decision making is shaped by our background, cultural environment and our own personal experiences. This sometimes leads us to make decisions which we are unaware are tainted by prejudice. This unconscious bias can lead us to discriminate unknowingly, much like Amazon’s algorithm.
Rise of the machines
Technology will continue to play a bigger role in our day to day and working lives. Robots are already flipping burgers and driverless cabs and drone deliveries could be operating within the next five years.
Artificial Intelligence is the next step and will have the greatest impact on our way of life and the way we work. Before we embrace its benefits though, we need to explore its flaws and understand our own.
If you have any concerns regarding your recruitment processes or discrimination in the workplace, please contact Charlie Barnes.