Down with the down-vote

Netflix announced it’s getting rid of its more granular five-star ratings, and replacing it with a simple thumbs up/down solution. I am sad to see this news; not because I like the granular listings that much (although I do take great care in picking my ratings), but because I was hoping we’d see a big platform realize that down-votes do more harm than good.

The simple up/down voting system is, of course, old hat for online platforms: YouTube announced their switch to it eight years ago, and popular sites like Reddit and Hacker News have had such a system since their inception, as far as I can tell.

Made most famous in modern times by Gene Siskel and Roger Ebert in their At The Movies film review program, the “thumbs up/thumbs down” concept seems intrinsically benign: you either like something, or you don’t—simple as that. Where could be the harm in that, right? However, the advent of technology has the potential to change the nature and psychology of simple concepts, and this one is no exception.

What we should have learned from years of vote brigading occurring on YouTube, on Reddit, on Hacker News and other places around the web, is that down-votes are a feature that can be easily abused by people who are motivated enough to do so. And in the age of an ever-connected world where millions have the ability to participate in online mob assembly against someone or something they don’t like, features that are easily exploited for abuse become vectors of abuse themselves.

Another thing that we can see, and data from YouTube could easily confirm this if they looked at it, is that certain groups of people abuse these features a lot more than others. Put simply: kinder people don’t abuse or vote brigade, whereas petty and malicious people do. (The behavior is intrinsically a petty and malicious one, so it makes sense it would appeal most to people who share those traits.)

I wrote in a lot more detail about this in my article for A List Apart, A Dao of Product Design. In it, I outline five principles for more humane design and technology. These are principles for a 21st-century understanding of products and technology, and the responsibility we as product designers bear to consider the sociological and societal impacts of our decisions. I even explicitly use the up/down-vote system as an example.

From the second principle, titled “Foster positivity and civility”:

Features like upvotes and downvotes may seem like a balanced solution for people to express opinions, but the downvote’s only purpose is to feed and perpetuate negativity; it can be avoided or removed entirely without harmful consequences.

Don’t give angry people shortcuts to wield negative power; make them either articulate their anger or deal with it in more constructive ways. Social media platforms never benefit from angry, biased groups suppressing messages (often positive and constructive) from people they despise. In those scenarios, everyone loses—so why design the option into your product?

Any feature that petty, time-rich people can abuse to game your product’s ranking or discovery algorithms is a feature that eventually serves up toxic behaviors (regardless of the person’s politics) and is best left out.

It’s worth noting that Netflix’s ratings model, concerns and audience are not the same as, say, YouTube’s or Reddit’s. It’s much harder, and much less rewarding by comparison, for people to abuse Netflix’s down-vote feature as a means to harass certain people or suppress certain views. Netflix does not host user-generated content, so the concern here is not that other Netflix users are prone to become targets the way that, say, feminist voices on YouTube frequently become targets of harassment.

But what if Netflix were to allow third-party content in the future? It would need to either remove the down-vote then, or create a split ratings model, if it wanted to adhere to modern design principles of positivity and civility. Additionally there is the example it’s setting to other companies, who may not make the self-aware realization that they are hosting user-generated content, and implementing a thumbs up/down feature would enable an easy means of suppression against people.

So what would it look like if Netflix followed the principle from my Dao of Product Design? It would feature only a thumbs-up choice at the end, and not using it would, by itself, not indicate a dislike. If they really need to track when people dislike a movie, show or episode, they could look at the people who regularly upvote other content but didn’t upvote this particular item, despite watching it to the end.

(Aside: note that media content is inherently different than, say, utility products. Most media is the expression of ideas, whereas a product like a curling iron might work very poorly and deserve a 1-star rating or, yes, an explicit down-vote. “I disapprove of this person and/or their ideas” is very different than “This product is broken.” As always, context matters.)

A reporting feature (i.e. “flag content”) would still be needed, but those features generally require the user to justify their criticism in more detail, and with a lot more effort. That’s okay; most content is good, and making people justify their (serious!) dislike of something isn’t harmful the way enabling a simple vector for abuse is.

Amusingly, Todd Yellin, Netflix’s vice president of product, said in their announcement that “Five stars feels very yesterday now.” I would argue that thumbs up/down feels very five years ago, before we learned that down-vote features were little more than an abuse vector for targeted harassment and the suppression of progressive voices and ideas.

I say it’s time to ditch the down-vote.