Artificial intelligence is not new to biases. In fact, these machines typically can become more biased than humans themselves, and at a much faster rate. Many companies have learned this lesson the hard way.
It came to knowledge that most of Amazon’s hires were men. Amazon uses an artificial intelligence machine to scan resumes and searches for specific word types that the company is looking for in a candidate. When studied, it came to be known that Amazon’s AI had been scanning for more aggressive words that were typically used by men – and because of this, the machine then came to teach itself that men were the preferred candidate, and thus, explains the gap in men vs. women hired for Amazon.
On the other end of the spectrum, AI can increase a turnover rate of more engaged employees when engaged in the actual workplace versus in the hiring process (which comes down to the difference of AI being correctly and incorrectly implemented).
Humu Inc. is an AI machine that is being used to change employee’s behaviors without taking away their right to choose – it sends out ‘nudges’ as emails to employees suggesting small behavioral changes in a way that makes it seem like making it their idea, not a judgmental message from one’s boss. This behavior change, or nudge, is then supposed to increase the amount of engaged employees. Engaged employees in the workplace are happier, and then go on to perform better work because of this nudge given by the AI in the form of an automated email.
When implemented correctly in the workplace, AI can really be helpful – however, figuring out what is appropriate or inappropriate for AI to be involved in, especially in the workplace, is a very thin line.
Julia Riccio – jmr17e
LIS 4022 Spring 2020 