In his interview, Neal Mohan disingenuously fails to mention that YouTube gives users a heavily weighted single recommendation after each video via autoplay
Brian Feldman / New York Magazine : Tweets: @dtunkelang Tweets: Daniel Tunkelang / @dtunkelang : YouTube chose to make autoplay a default. So it is disingenuous for YouTube to deny responsibility for actively pushing users towards its recommendations. http://nymag.com/...
Context & Ripple Effects
The piece lands alongside Neal Mohan's interview on YouTube's recommendation algorithm, where he addressed radicalization and the rabbit-hole effect while framing recommendations as something users encounter rather than something YouTube imposes. Brian Feldman, drawing on Daniel Tunkelang's critique, points to the one design choice that undercuts that framing: autoplay on by default, which delivers a single heavily weighted recommendation after every video.
This is not a new tension. YouTube's own engine was already reported to drive over 70% of viewing time while surfacing divisive and misleading content, and Mozilla's later research found 71%+ of videos users flagged as objectionable were recommended by YouTube itself. The autoplay point matters because it converts an abstract algorithm debate into a concrete product decision YouTube made.
First-order effects
- Mohan's responsibility argument in the interview is directly weakened: with autoplay as a default, the post-video recommendation is an active push by YouTube, so denying responsibility for steering users toward recommendations becomes hard to sustain.
- The immediate pressure shifts onto YouTube's product leadership — Mohan, then chief product officer — to defend or change the autoplay default rather than discuss the algorithm in the abstract.
Second-order effects
- Subsequent evidence compounds the problem for YouTube: Mozilla's later study found the 'dislike' and 'not interested' buttons have little impact, meaning users have few effective counters to a default they never chose.
- Rivals and regulators gain a concrete talking point: the question moves from 'is the algorithm biased?' to 'why is the pushiest recommendation the one users get automatically?' — a framing any competing platform or policymaker can adopt.
Third-order effects
- If the pattern holds, accountability frameworks for recommendation systems shift from auditing what algorithms output to scrutinizing interface defaults — treating autoplay-style design choices as deliberate distribution decisions rather than neutral UX.
- That reframing points toward distribution-layer liability: platforms whose defaults actively route attention may face pressure to justify those defaults the way publishers justify editorial choices, a structural shift away from the neutral-host defense.
The trend: Platform accountability debates are shifting from what recommendation algorithms produce to how interface defaults like autoplay make recommendation an active act of distribution.