The plumber has just left after patching that stubborn leak in the basement. Moments later, your phone pings with a request to rate their service. It’s a familiar scene: if the job wasn’t terrible, do you rate it three, four, or five stars? According to new research from the University of Toronto’s Rotman School of Management, this seemingly straightforward decision can carry unintended consequences. Multi-point rating systems, like the familiar five-star scale, are not just prone to subtle, often unconscious racial bias but can have significant financial impacts on non-white workers.

Drawing on data from a real-world home maintenance matching service, researchers at Rotman found that white workers received higher average ratings than their non-white counterparts. The numbers reveal a troubling pattern: white workers earned top ratings 86.9% of the time, while non-white workers only achieved the same score 83.4% of the time. Though this gap may seem minor, it significantly affects earnings and career prospects over time. Even a slight difference in ratings can lead to substantial financial losses when pay is directly tied to average ratings, as is often the case in gig economies where workers rely on customer reviews for their livelihoods.

In the pay structure of the service studied, workers’ earnings were directly tied to their average ratings, meaning a slight but consistent rating disadvantage translated into significant financial disparities. The researchers estimated that under this system, non-white workers ended up earning just 91 cents for every dollar made by white workers. This pay gap arises not from explicit discrimination but from the cumulative effects of subtly biased evaluations, which can profoundly impact over time. As Katherine DeCelles, a professor of organisational behaviour at the Rotman School, explains, “While the objective difference, on average, between white and non-white worker ratings is very small, it matters because of the impact it has on income, highlighting the importance of structure and organisational design for racial equality at work.”

Unlike overt racism, which might involve a customer refusing to hire a non-white worker outright, these subtle biases are more problematic to identify and address. They often manifest even when individuals do not consciously intend to discriminate. Embedded within seemingly neutral rating systems, this bias can be especially insidious because it influences financial outcomes without being immediately apparent, making it challenging for platforms to detect and correct. However, the research suggests that simple changes in rating design can help counteract this problem.

For instance, when the platform being studied switched from a multi-point scale to a simpler, two-choice system — asking customers only if they would use the contractor again (essentially a thumbs-up or thumbs-down) — the racial gap in top ratings virtually disappeared. New workers who joined the platform after this switch saw no significant racial differences in earnings for similar jobs, indicating that simplifying rating structures can effectively mitigate these biases. Further experiments, using real-world data and controlled online simulations, reinforced this finding, showing that participants with subtle biases produced more equitable ratings when given only two options, rather than a range of subjective choices.

Given the growing importance of digital rating systems in the gig economy, the researchers recommend that platforms adopt simpler, two-choice rating structures and regularly audit their systems for bias. They also suggest alternative methods for customers to offer nuanced feedback that does not directly impact worker pay, ensuring fairer treatment for all workers. As Prof. DeCelles notes, focusing on whether a job was good or bad, rather than asking customers to quantify quality on a scale, can help strip out the subtle, often unrecognised biases that can affect ratings — and, by extension, livelihoods.

More information: Katherine DeCelles et al, Scale dichotomization reduces customer racial discrimination and income inequality, Nature. DOI: 10.1038/s41586-025-08599-7

Journal information: Nature Provided by University of Toronto, Rotman School of Management