Spotify looks to staff up AI ethics team

If you search up “algorithmic responsibility” on Spotify, you’ll find zero songs and a handful of podcasts. But, behind the scenes, the company is seemingly working to expand its AI ethics efforts.

In the last month, the company has posted three open roles for research and data scientists on its algorithmic impact team. The new hires would help evaluate the impact of Spotify’s algorithms on audio culture, plus investigate and help avoid potential harms related to the company’s AI.

Spotify is known as the world’s leading audio-streaming service, with 400+ million active users across its paid and free tiers. Due to the company’s reach—including users in more than 180 markets—its recommendation algorithms have significant impact on the music and podcasting worlds, and close scrutiny is required. A 2021 study showed music recommendation algorithms were more likely to favor male artists.

Currently, Spotify has at least two individuals working full-time on its algorithmic impact team, per LinkedIn, meaning the new hires would give the team at least five full-time workers. The company did not return multiple requests for comment on the official size of its algorithmic impact team.

Who they’re hiring

Spotify’s incoming research scientists have a long to-do list: They’ll help create concrete company policy and assessment processes around responsible AI—including working with Spotify’s policy, product, and research teams on methods to address potential “data, model and, metric inequities.” The job would also include publishing research, as well as reviewing “sensitive content” that might come up as a result of algorithmic harms.

• For reference, this doesn’t sound too different from Twitter’s goals for its relatively new responsible AI team.

For their part, the data scientists would help set parameters around recommendations for podcasts, music, and more. They’d also design studies and experiments centered on algorithmic trust and safety, plus develop metrics to study “algorithmic amplification” and help with content moderation.

In January, Spotify published an algorithmic impact study on how lesser-streamed podcasts could increase their audiences—and, last March, a study on how reinforcement learning could drive Spotify users toward more diverse content.

Big picture: Spotify’s job postings come at a time when companies across the board are trying to shore up their AI ethics efforts, after a year of it being a headline issue—and ahead of potential federal regulation. In February, a group of senators re-introduced the Algorithmic Accountability Act, which would require companies using automated systems to make “critical decisions” to conduct impact assessments.