It is a fact that computational algorithms often outperform human judgement.
For instance, companies created algorithms to predict consumer behaviour. Organisations use algorithms to help predict when their employees will leave. Traders create algorithms to help them make the best automatic trades, based on myriad data points.
When we have a problem, we seek advice from other people. The rise of ‘big data’ has also given rise to a new source of advice: algorithms, which boast superior accuracy, capacity, and speed, relative to human judgment.
It explains the global shift in money pouring into artificial intelligence and machine learning companies—with a CAGR of 36.2% and set to grow to a USD$208.49bn industry by 2025.
Organisations and individuals are often using algorithms, whether consciously or unconsciously. Spotify recommends songs based on our playlist and frequently listened-to songs. We discover new shows and movies on Netflix via their recommendations. Facebook promotes content we may like as well as posts from our close friends.
Some other algorithms are backstage, with more profound yet less obvious effects, such as traffic lights, bus arrival timings and predicting crime rates in different city sectors for police patrol frequencies.
With such widespread reliance on algorithmic advice, it seems odd that studies have shown that we are skeptical of relying on them.
There is a wealth of literature, cited in many business press and publications, about the phenomenon of “algorithm aversion”.
For instance, we trust an algorithm considerably less than a human forecaster—when they both make the same prediction error, we are more likely to trust an inferior human forecaster over an algorithm, even though the algorithm is outperforming the human many more times over.
In one of the experiments in that study, participants were asked to look at MBA admissions data and guess how well the students had done during the program. Accurate guesses meant walking away with a small amount of money and they could either submit their own estimates or predictions from an algorithm.
Participants who had seen the algorithm’s results were less likely to bet on it, even if that was compared to their own dismal prediction performance. Many are much less likely to believe the algorithm would perform well in the future, and that finding held across several similar experiments in different settings, even after the algorithm has been made more accurate.
We must also understand that algorithms are placed at different levels of significance depending on context — we trust the algorithm behind our GPS more naturally than the algorithm behind a dating app’s recommendation.
Hence, it is worth discussing overcoming algorithm aversion, as seen in literature such as this University of Pennsylvania study in 2016.
Yet, with Harvard Business School’s (HBS) study suggesting that we generally trust an algorithm more than, the practical implications behind it run deeper for organisations who depend on algorithms for decision-making.
In contrast to the widespread conclusion of “algorithm aversion”, the HBS study suggests that people readily rely on algorithmic advice over advice from other people, no matter the context.
Be it geopolitical events, song popularity or romantic matches, the study revealed that people weighed algorithmic advice more heavily than human advice.
When given the choice, they chose to trust the algorithm’s judgment—they were also willing to choose it over their own.
Yet, that appreciate was not uniform, which is important for any organisation to consider.
Human Experts Trust Algorithms Less
The study showed that human experts trusted the algorithm a lot less as they were simply less open to taking any advice. Algorithm aversion was the most prominent here: due to their experience and knowledge, they betted on their own judgment more than the algorithm.
Human judgment can be unique amongst different human experts—coupled with experience and personality traits, even having the same amount of knowledge can result in varying decisions when you ask different individuals.
Algorithms can be one-sided: if you’re calculating risk, for that matter, the algorithm will easily tell you the amount of risk and quantify it to make sense. The decision you make will be based on the algorithm’s results, which therefore also transfers responsibility to you. Having a human expert can provide different perspectives to the algorithm, which brings greater benefits.
Ultimately, an algorithm can only go so far. Without making the right decisions and having that process being solely based on an algorithm—though that depends on the context—it creates a concentration risk.
By bringing in a human expert that takes the algorithm’s result with a pinch of salt, different points of view are brought together. Having a holistic view is key: you don’t want to be blindsided or to leave blind spots unaddressed, especially when there are high stakes involved in the decision.
Although an algorithm can outperform human judgment in the long run, an algorithm can only be as good as the data it is fed: without high-quality data and accurate formulae built into it, the algorithm cannot outperform the human. In fact, it might give inaccurate results, which may result in a lot of negative impacts that a business cannot afford.
This is where a human expert comes in to give their view, which allows a diversification: it is through a balance of algorithmic judgment and human judgment that an organisation can flourish with controlled risk.
A healthy amount of skepticism is great for any context, but you cannot control that amount in a human being. Generally, the unique blend of personality traits and experiences are what decides how we are going to be today and how we view certain things—which means some experts are going to be stubborn.
When one is stubborn, he loses the opportunity to grow. Having a fixed mindset and choosing to only believe themselves and their judgment is unhealthy, even if that person has a great track record. Generally, any organisation needs people with a growth mindset in order for the organisation to grow.
Having a human expert that one-sidedly trusts themselves more is going to be counterintuitive, and therefore defeat the purpose of an algorithm.
Algorithm aversion is still a real thing, despite the growing body of research that disproves it. There are professionals who are against making decisions based on an algorithm, and thus, it is important for any organisation to address that immediately.
That will take additional resources and time—and not everyone can be convinced, even if that means that they might lose their job.
Regardless of how an organisation makes a decision, the core philosophy in mind is to always consider multiple perspectives.
Just like an investment strategy, never put all your eggs in one basket—in the world of decision-making, that means the number of perspectives you can get from different sources.
For most organisations, an algorithm can definitely be of benefit to their business processes. Being based on multiple data points is one thing: making a decision based on all of them with speed is another.
No matter the algorithm, it is important to focus on the data points and the quality and quantity of data. For any organisation depending on algorithmic decisions, having inaccurate predictions will only be fatal.
To Motivate Your Employee, Ask Her For Advice
Research: Giving Advice Motivates Us To Solve Our Problems
Stop Asking For Feedback—Ask For Advice Instead
Studies Show That Advice Is a Better approach
This article was first published on the Human+Business, a publication aimed at leading conversations about how we can be more human in our businesses, management, and leadership in today’s context.
We also publish new and authoritative ideas in a newsletter, with the best articles and exclusive think pieces. Let us know your interest here.
The first edition of the H+B Newsletter will be coming to inboxes on 31st October.