Airbnb hosts less likely to accept bookings from Black people than Whites
Ah c'mon, we've gotta be better than this
Airbnb guests perceived to be Black by hosts are slightly less successful at booking properties than White guests, according to a report from Airbnb itself.
The report comes from Project Lighthouse, an effort Airbnb launched in 2020 to measure and tackle racial discrimination and bias on the platform. The initiative focused on the rate at which folks in the US from different racial groups would have their reservations confirmed after they tried to reserve a place, something the company referred to as the "booking success rate."
The results from the project revealed "the widest disparity exists between guests perceived to be Black and guests perceived to be White," the internet biz said in its report [PDF].
"Our 2021 data shows that guests perceived to be Black were able to successfully book the stay of their choice 91.4 percent of the time, versus 94.1 percent for guests perceived to be White," the document adds.
Folks perceived to be Asian or Latino had booking success rates of 93.4 per cent.
Project Lighthouse only considered racial biases, and did not consider other personal attributes like age or gender. In 2018, Airbnb removed the ability to see a guest's name and photo when they reserved properties or rooms, to prevent racial discrimination. Now, the information is only revealed after a host has accepted and confirmed a booking.
Still, hosts can guess what a person might look like from other types of data, such as their name, for example. When profile pictures were hidden until a reservation had been booked, the gap between the booking success rates of White and Black users only decreased by half a percentage point, and did not change much and was not "statistically significant" for other racial groups. Cancellations after the host saw the people's pictures did not go up.
- Turns out humans are leading AI systems astray because we can't agree on labeling
- There's nothing AI and automation can't solve – except bias and inequality in the workplace, says report
- Q. If machine learning is so smart, how come AI models are such racist, sexist homophobes? A. Humans really suck
- Once again, racial biases show up in AI image databases, this time turning Barack Obama white
Other factors that lead to lower booking success rates include Black and Latino folks being more likely to be first time users, and therefore less likely to have been reviewed by hosts. That means they're less likely to be offered Airbnb's "Instant Book" feature which doesn't require hosts to review reservation requests.
Airbnb said it has since removed the requirement for guests to be recommended by hosts and users will only need to have "a good track record" to make Instant Book more accessible to different racial groups.
"We are working to better understand how to increase trust between Hosts and guests throughout the reservation process. In this work, we are exploring changes to Host and guest profiles to highlight information that fosters more connection, such as interests and the kinds of activities guests enjoy while traveling. At the same time, we will further explore the impact that other features may have on the opportunity to create bias, for example assessing the impact of initials rather than full names," it concluded.
Laura Murphy, President of Laura Murphy & Associates, and senior advisor for Airbnb's anti-discrimination efforts, wrote in a statement: "I am excited that Airbnb is embracing transparency by using this report to share this data with the public. Too often companies find discrimination problems and want to bury them in secrecy, but since 2016, Airbnb has been committed to taking action and being straightforward about both its progress and its challenges"
"By sharing the key insights it has learned through Project Lighthouse and how the company is putting them into action, Airbnb is once again demonstrating its genuine commitment to fighting discrimination." ®