The effects of an “algorithmic monoculture” depend on the details

The impact of "algorithmic monoculture" hinges on specifics. Resume screening algorithms, for instance, streamline hiring by improving efficiency and consistency. However, some scholars fear that using a single algorithm for all decisions could lead to "algorithmic monoculture," which may result in systematic exclusion. This notion suggests that a rejected job candidate would be rejected everywhere.

Yet, MIT researchers now propose that this may not always be the case. They examined major criticisms of algorithmic monoculture, such as systematic exclusion, and concluded that many arguments either fail or don't decisively oppose all types of monoculture. According to their mathematical proof, monoculture often creates informational echo chambers that hinder exploration.

In hiring, this might make it less likely for the best candidates to secure jobs. However, combining various hiring algorithms into a single "ensemble" could potentially overcome this drawback, possibly even making monoculture as effective, if not better, than a diverse approach where each firm uses different algorithms. Dr. Brian Hedden, a professor at MIT and co-author of the study, emphasizes that a shift towards algorithmic monoculture is plausible, but its implications depend on specific details, like the industry and algorithm accuracy.

Algorithmic monoculture, where decisions in a specific domain are made using the same algorithm, is not new. For example, lending decisions were previously made by individual banks but now rely on a standardized credit score algorithm. Similarly, many Fortune 500 companies use the same resume screening algorithms. The concern is that as more people leverage AI and algorithms for decision-making, the possibility of correlation increases.

To understand this better, Dr. Manish Raghavan and Dr. Hedden assessed the advantages and disadvantages of algorithmic monoculture, focusing on hiring. One common objection is that relying on the same algorithm could systematically exclude certain individuals. While it might happen in hiring because a rejected resume would likely be rejected by every firm, the researchers argue that the overall number of hired individuals remains unchanged.

Instead, algorithmic monoculture could enhance job candidates' bargaining power, potentially driving up wages as firms compete for the same pool of candidates. Another criticism is related to agency. If a job candidate applies for a position and their resume is forwarded to all firms using the hiring algorithm, they won't have a chance to refine their resume to improve their chances.

However, if job candidates can revise and resubmit their materials, this objection may not hold. On the other hand, algorithmic monoculture might enable individuals to manipulate the system by reformating resumes to improve outcomes. The researchers also considered the possibility that monoculture could lead to the homogenization of information, which contradicts the "wisdom of crowds" theory.

This theory suggests that a diverse group of independent decision-makers could outperform a single person. In hiring, diverse hiring algorithms could lead to a higher-quality pool of new hires, while algorithmic monoculture could cause consistently hired candidates with similar characteristics and credentials, potentially limiting diversity.