Seeking Advice from Technology

An algorithm can be simply defined as a set of instructions for solving a problem or accomplishing a task. Within many business domains, algorithms are rapidly taking over the decision-making abilities of humans. However, full automation of business activity is only found in certain industries, such as manufacturing, where robots can work in a well-controlled environment to perform specific tasks.

Environments that are complicated to automate, such as retail industries operating in a real-world shopping site, require decisions regarding pricing and inventory control to be physically implemented. With partial automation, managers can give workers algorithmic advice, but ultimately, this relies on workers actually complying with the advice.

Previous studies have found that a strategy of ordering workers to follow an algorithm is not always effective. Interestingly, there is a phenomenon known as algorithm aversion, where people often obey inferior human decisions, even if they understand that the algorithmic decisions outperform them.

A recent study by Kohei Kawaguchi aimed to further investigate if and under what conditions workers would follow the advice of an algorithm. To test this, the researcher developed an algorithm to automate the task of experimenting with product assortment decisions for a beverage vending machine company operating in the Tokyo metropolitan area. “The algorithm chooses the product assortments each week for each vending machine in an area. At the end of the week, it observes the weekly data and updates the belief, continuing the process until the end of the season”, the researcher explained.

Within a simulation setting, it was found that the algorithm could increase the revenue of the vending machine business. As such, the potential impact of incorporating automation within this type of business was significant. The algorithm was also applied to a field experiment to identify the challenges faced by managers when putting an algorithm into practice. Specifically, the author explained that the aim of the field experiment was to “investigate whether integrating a worker's opinion into the algorithm can increase the worker's conformity to the algorithm.”

It was found that the field experiment failed to increase revenue, because the workers, on average, did not follow the algorithm. However, integrating workers’ forecasts into the algorithm was able to slightly encourage workers to follow it. Another interesting finding was that a worker was more likely to follow the algorithm if they had higher regret, faced lower sales volatility, or had a lower degree of delegation. Additionally, integration was more effective if the worker had lower regret, faced higher sales volatility, had more work experience, and a higher degree of delegation.

These findings offer valuable insights into the impact of automation for other industries. This is because product assortment decisions are common for consumer businesses, which is particularly the case for industries that operate in real-world shopping sites, such as supermarkets and convenience stores.


Assistant Professor