Shota Ichihashi, Assistant Professor of Economics, Queen’s University
- Alex Smolin, Assistant Professor of Economics, Toulouse School of Economics
This blog article is derived from the authors’ paper entitled Buyer-Optimal Algorithmic Consumption, a project of the Economics of Digital Services (EODS) initiative led by Penn’s Center for Technology, Innovation & Competition (CTIC) and Warren Center for Network & Data Services. CTIC and the Warren Center are grateful to the John S. and James L. Knight Foundation for its support of the initiative.
In the fast-paced world of the digital age, algorithms are becoming an essential part of our daily lives. They’re changing the way we travel, shop, and manage our finances, and it’s all thanks to the impressive advancements in information technology and artificial intelligence. As the technology develops, these changes are only expected to grow. From chatbots like Expedia’s travel assistant planning your next vacation to price-tracking apps such as Honey that find the best deals, algorithms are shaping our consumption in unprecedented ways.
While the productivity benefits of algorithmic decisions such as Amazon’s use of algorithms to optimize warehouse efficiency are clear, the bigger picture that must necessarily include strategic interactions in the economy isn’t as widely understood. That’s what our research aims to shed light on, especially in the areas of trade, pricing, and the well-being of consumers.
Finding the Best Algorithms for Consumers
In our study, we want to understand what makes an algorithm truly beneficial for consumers when faced with strategic sellers. We focus on two fundamental aspects of algorithmic decisions:
– Finding Valuable Information: This is something most of us are familiar with. Algorithms help us discover information that leads to smarter decisions. They help connect consumers to the right products and uncover new possibilities.
– Commitment Power of Algorithms: This is a more overlooked aspect. Algorithms can follow a set of predetermined rules without human intervention. This consistency allows for a shift in bargaining power, affecting how products are priced and how active the market is.
For concreteness, think of a situation where a buyer is unsure about the true value of a product sold by a monopolist. A recommendation algorithm comes into play, because it can figure out the value and suggest the product based on the value and the price. Any given algorithm affects the product demand that in turn influences the seller’s pricing decisions.
We show that designing a buyer-optimal algorithm is not as simple as recommending the product whenever its value estimate is greater than the price. This approach overlooks strategic effects and might miss the chance to influence the seller to offer a better deal by holding back if the price is too high. On the other hand, recommending the product only at extremely low prices might cut off trade opportunities when the product is costly to produce.
So, how do we strike the right balance? We applied game theory, an analytical approach widely used in auction design, to find the optimal algorithm. Our findings revealed an algorithm that uses a price-dependent threshold to decide when to recommend a product. If the value is high enough, the product gets the green light. The optimal algorithm isn’t perfect, however. Sometimes it may suggest a product even if the buyer values it less than the price. Other times, it might hold back a recommendation even when the value is above the price. This isn’t a flaw; it’s a feature. By making these “mistakes,” the algorithm encourages the seller to set lower prices overall.
This algorithmic consumption leads to a market where sellers with different costs offer various prices. Those with higher costs might charge more and make fewer sales, while others might charge less but secure a steady stream of buyers.
Data in the Algorithmic Economy
Our analysis allows us to assess various data interventions. We looked at two key cases:
- The Impact of Data Leakage to Sellers: In the online world, sellers might get their hands on information about buyers to set prices differently for different buyers. One may think this could tip the scales in favor of the seller, but our research found something surprising. Under the buyer-optimal algorithms, both the buyer and seller’s gains stay the same, even if the seller knows more about the buyer’s preferences. How? The algorithm can adjust its strategy to counterbalance this leakage of information. This discovery shows how smart consumption algorithms can act as a shield to protect consumers.
- The Effect of Algorithms Having More Data: What happens when the algorithm itself gets more information? Our findings here are equally intriguing. More precise data can lead to both higher product prices and better outcomes for consumers. It sounds contradictory, but here’s why it works: More accurate algorithms can pinpoint buyers willing to pay more, allowing sellers with higher costs be active in the market with higher prices. This price increase, however, is balanced by recommendations that are more on target, making sure consumers get what they really want.
Algorithmic decision-making is transforming our world, affecting everything from pricing to consumption choices. Our research unveils the complex relationship between algorithms, data, and market behavior.
We’ve discovered that algorithms can be more than just efficiency tools; they can protect consumer interests and reshape commerce. These insights are not just interesting academically but have real implications for policymakers, businesses, and consumers.
As we navigate an increasingly digital world, understanding how algorithms work is vital. Our study provides insights into this complex ecosystem, underscoring the significance of strategic economic reasoning.