What Should Leaders Know Before Using Decision-Making Algorithms?

Decision-support tools are becoming more common in places ranging from medical centers to boardrooms. Many of these products have advanced algorithms working in the background to process huge quantities of data and make relevant recommendations to users. Although these decision-making algorithms are often helpful, they’re not foolproof. Here are some things leaders should keep in mind when relying on them. 

People Can Become Overconfident in Advanced Tech

Many companies that develop decision-making algorithms fill their websites with compelling statistics and fascinating case studies about how their technologies have helped others. Such content can motivate people to try algorithm-based products, especially if they want to solve specific problems. 

However, a 2023 Aalto University study identified a placebo effect that can occur, making people too confident in technologies they believe will help them. Researchers confirmed that this can happen even when the high-tech tools people use aren’t actually providing benefits. Moreover, the findings suggest that when users strongly believe tech tools will provide improvement, that stance can alter decision-making processes. 

The researchers created an experiment where they told 27 participants that artificial intelligence (AI) would power a brain-computer interface to enhance their cognitive abilities. Those taking part used the brain-computer interface to play a game. Cards with hidden values would make them either gain or lose points. 

The researchers’ technology didn’t provide any genuine benefits. However, most players believed it had helped them improve their performance. Some also took bigger risks due to their belief in the tech. These findings highlight how the hype surrounding a new product could inflate people’s perceptions.

That’s not to say users should avoid all products that use decision-making algorithms or similar technologies. However, this study gives strong reminders that organizational leaders must not buy into the hype surrounding the products that capture people’s attention and newspaper headlines. 

One best practice is asking numerous organization members to try products and provide honest feedback. Such insights should give a more balanced view of whether a tool is likely to meet needs and give trustworthy results. 

Decision-Making Algorithms Can’t Always Explain Conclusions

People who create machine learning algorithms choose from various types depending on their specific applications. For example, some are better at detecting spam emails, while others excel at data classification. 

However, one of the downsides of many AI algorithms is that they can provide users with recommendations but not specifics about why or how they reached them. This reality results in a so-called black box issue. That’s a major problem that has led to people becoming increasingly interested in research about explainable AI. If users can make an algorithm work backward to show how it reached a decision, they’ll be likelier to go with whatever it suggested. 

That’s particularly true when people use algorithms to make choices that could change their lives or those of others. Suppose someone relied on AI to make a medical diagnosis, narrow the pool of candidates interested in a job or offer someone a loan. In that case, they’d want to know how the associated decision-making algorithms work before trusting them. That’s not always possible yet, but it’s becoming more common. 

In one recent case, University of Waterloo researchers built a new explainable AI model to reduce biased results and increase people’s trust. Experiments in the medical field showed this new Pattern Discovery and Disentanglement model could predict patients’ medical outcomes by examining their clinical records. It also found unique and rare data patterns, making it easier for practitioners to discover possible errors in the information. 

However, in general, today’s leaders should assume that the AI tools they use are not explainable. With that in mind, they may want to only rely on them for non-critical decisions where people will not demand further details about how a tool reached a specific conclusion.. 

Business Leaders Are Often Overwhelmed 

A 2023 study revealed that modern business leaders often face data deluges that are too great for them to handle. It’s no surprise, then, that 70% of respondents said they’d prefer to solve the problem by having robots make all their decisions. 

Another interesting takeaway was that 79% of those polled felt that using technology to make data-driven decisions makes an organization more trustworthy. The reliance on concrete information is particularly important since 78% of participants believed people make decisions first and then look for the data to justify them. 

Decision-making algorithms aren’t perfect, but they could help people make positive changes. Since well-trained artificial intelligence tools do particularly well with large amounts of information, they enable people to sift through the data faster than they otherwise could. 

Nearly a quarter of the people in the study believed most business decisions are irrational. Algorithms could bring improvements in that regard, too, as long as users pair the tools with their judgment. 

Since 85% of respondents said their inability to make decisions negatively affected their quality of life, it’s time for things to change. Additionally, many of the tools business leaders currently use don’t meet expectations. A study statistic indicated that 77% of respondents said the charts and charts they have do not always relate to the specific decisions they need to make. However, 93% felt that the right decision-making intelligence was a make-or-break aspect for organizations.

Today’s business leaders are ready for the support that decision-making algorithms could provide. However, people thinking about using them should do so along with other types of assistance, including their experience and feedback from others involved. Even the most advanced algorithms can fail, especially if exposed to poor-quality data. Using various sources to reach decisions can increase people’s confidence in the process. 

Use Decision-Making Algorithms With Caution

It’s easy to find a wide assortment of products that have decision-making algorithms working in the background. However, not all are equally useful or even relevant to many people who think they need to use them.

Anyone interested in decision-making algorithms should think carefully about their intended use cases and create processes with safeguards that prevent themselves or others from becoming overly reliant on the technology. Even though decision-making algorithms can support the choices humans make, they should not replace them.