How can managers ensure their employees trust algorithms?

People are increasingly being asked to work alongside automated programmes – but for businesses to get the most from AI, staff must first understand how it works, says Karlien Vanderheyden

Algorithms are everywhere in our lives, both in and out of work – even if we don’t realise it – and they’re only going to get more common. When driving, many people will check Google maps for the quickest route for the journey. Then halfway through they’ll ignore the route the algorithm is telling them and take what they think is a shortcut, which often doesn’t end up being the case. The fact is, most people trust their own instincts, and that of others, over an algorithm. Yet all research points to algorithms outperforming humans most of the time, so why don’t we trust them?

The main reason could be down to the high expectations people have of algorithms – most think they should be flawless. In fact, research suggests errors are seen as human traits, and something that is not acceptable, or possible, in algorithms. If an algorithm does make an error, people tend to then completely distrust it – something you would not see in an interpersonal relationship. 

People are also less trusting of algorithms because there is not enough education around how they work. To many, algorithms are seen as a ‘black box’; we know the input and we know the output, but not exactly how it works. If we fail to understand the process behind algorithms, those responsible for inputting the information into them are not responsible for the outcomes, and are therefore not focused on the impact these algorithms have. 

Another reason people don’t trust algorithms is because they are able to pick flaws in how they work. For example, imagine a news channel used machine learning to find new anchors. In the process, it defines a successful anchor as someone who stays for five years and gets promoted twice. If women had been excluded from this channel in the past and the algorithm were to use this data, we would suspect women would be filtered out. Many see this apparent bias as a flaw but, according to research from Chicago Booth, algorithms are 10-15 per cent better at predicting human behaviour than humans themselves.

So how can we change people’s attitudes towards algorithms, and ensure we not only trust them but are open to working with them?

The first step should be education. Teaching employees about the algorithms they work with, how they operate within the business and the impact they could have is important. Employees must understand what lies behind these algorithms and how they work. 

Improving employees’ knowledge will allow them to not only better understand algorithms, but also help them to ask algorithms the right questions. Often people ask ‘will the algorithm help me achieve my goal?’ – which is often not the case, especially when it comes to challenging objectives. A better question is ‘is the algorithm better than me?’ In which case, we as humans would choose the algorithm more often. Once employees understand this, they will get the results they want from the algorithm. It is just as much about asking the right question as it is about having the right expectations.

Humans are also more likely to trust an algorithm if they have a certain amount of control over the outcome. This is the same in interpersonal interactions too; an employee is much more likely to follow a manager’s instructions if they’ve had a say in it, and it’s exactly the same with an algorithm. To ensure employees and algorithms work collaboratively, humans should be given the final say over decisions, giving algorithms a simpler advisory role. 

To address the issues, companies should be hiring people who can regularly audit and update algorithms to ensure there are few instances in which humans have greater knowledge to make a decision than algorithms.

Employees and algorithms collaborating together can hugely benefit businesses, as long as it is done effectively and efficiently – otherwise, it could have a negative impact. To make this collaboration as impactful as possible, employees have to learn the best ways of utilising these algorithms for their benefit, and really trust how they work. 

Karlien Vanderheyden is professor of people management and leadership at Vlerick Business School, Belgium