About a year ago, I gathered some really smart people who don’t work in news together in a room. I set out some index cards and some beer and told them we were going to think about new ways to design a news homepage.

Over the next few hours, the group came up with dozens of new ways to think about a homepage, many of which were radically different than anything that exists today. There were homepages for people with limited time and homepages for people who wanted to read the news from different locations at the same time. One homepage offered news from the perspective of someone living elsewhere in the world; another only offered articles if no one else in a particular social circle had read them.

The experiment went really well, in part because it made me think in radically different ways about how news can be presented and distributed. A news homepage doesn’t have to look or feel like a print newspaper. It can be something entirely different, depending on the audience and the medium.

That also holds true for the comments section of a website. I’ve rounded up quite a few ways to think about comment sections and reader engagement, and I’m hoping you’ll share more in the comments or on social media.

But first, we’ll start with the basics. What makes a good website comment? I reached out to Nick Diakopoulous, an assistant professor who studies comments and commenting platforms at the University of Maryland, for his thoughts.

“Just like a good quote, a good comment illustrates and adds to your understanding of a story, drawing on experience or expertise to crystallize an important perspective,” he said. “I think that the best comments are thoughtful and reflective, respectful and fair, all while being accessible and easy to read and follow. A great comment is one that can be quoted in the story itself without the reader batting an eyelash.”

Diakopoulous recently conducted an experiment examining whether algorithms could be used to help editors highlight high-quality comments on their news sites. The algorithm assigned ratings to comments based on their relevance, readability and conversational tone and compared them to comments evaluated by humans. The comments picked by the computer correlated with what people selected as “good comments.”

“Of course the scores are not perfect — you always lose something when quantifying something as subjective and nuanced as human language,” says Diakopoulous. “For instance, our ‘readability’ score reflects the grade level reading difficulty of the text, but you might also consider other dimensions like formatting and punctuation in a readability score. There’s still room to iterate on our algorithms for scoring these factors, as well as for developing algorithms to measure other dimensions like thoughtfulness, argument quality, fairness, and novelty.”

Diakopoulous' tools look at the comments themselves; a commenting product by The Coral Project looks at commentators and their future likelihood for leaving a good comment, based on their previous commenting history. Both projects have the same goal: to make it easier to highlight good contributions without having humans sift through every entry.

The possibility of computers moderating comment sections is an intriguing one. Unmoderated comment sections are the bane of many newsrooms, and the Chicago Sun-Times, Popular Science, Recode, Mic, The Week, and Reuters have all recently pulled the plug on their comment sections. Comments can shape perceptions about news organizations and devolve quickly into bizarre commentaries that have little to do with the articles themselves.

Diakopoulous says that hostile or irrelevant comments can lead to other negative comments, but good comments can also lead to more nuanced, civil discussions.

“Comment sections are no different than our neighborhoods. If they look littered, with incivility or irrelevance, people will think it’s okay to make those kinds of comments,” he says. “Research suggests that positively cueing people within the comments can signal to the community what 'good behavior' looks like. I believe that showcasing the best comments sets the tone and expectations and can lead to better discourse over time.”

How are best comments selected? There are a few different ways.

You can have a moderated comments section. Think NYT Picks, which has 13 part-time journalists making sure there’s a range of well-written views.

Or you can have a community with moderators but also self-regulation. Think Reddit, where up and down votes determine what you see, or Metafilter, one of the most robust conversation sites on the Internet, where moderators routinely delete comments and members pay $5 to be able to post and comment. And then there’s Civil Comments, a platform that makes people peer review a comment before self-rating their own — which then helps the platform decide which comments to publish.

You can also have a scenario where a community can upvote for good stuff but not downvote for bad stuff. Think the way material rises to the top of sites like Product Hunt or Hacker News or Slashdot, which also has comment filters for insightful, informative, interesting, and funny material.

Or you can think of commenters or readers as additional writers themselves. For example, Ask Metafilter, Ask Reddit, and Captain Awkward all ask readers to give and receive advice. And El Pais’ Eskup platform asks readers to “write short articles, opinion pieces or even running commentaries” on soccer games, Iker Seisdedos points out in The Guardian.

But comments are not always limited to a response in a text box below the article. Medium and Genius let readers offer in-line highlighted comments or in-line annotations. This lets the reader add context or insight, while also providing each platform with valuable metadata and insight into what and how readers consume content.

There are commenting platforms that also function as spaces to have a back-and-forth and conduct live interviews with readers, like Digg’s Dialogue Platform, Quora’s Writing Sessions, or Reddits AMA’s. News commenting platforms can even exist independently of a news site. Think original Internet forums like The Well or more recently, Parlio.

Of course, these are all Web-based and assume that interactions are free-text written responses and will take place on a website. That’s not always the case. Comments don’t always have to be written down.

One of my favorite examples is Consider.It, created by Travis Kriplean, a computer scientist in Seattle. Consider.It asks users to identify how they think about an issue on a continuum, which helps identify patterns of thought across a group of people.

“Consider.it and the other discussion systems I've created are built on the observation that text-based communication, like email and online comment boards, make it easy to speak into, but hard to listen through,” says Travis. “When we're face to face, we can use body language, like nods and furrowed brows, to communicate whether the speaker's message is getting through. We lose much of that when we move to text-based mediums, which creates all kinds of problems, like people shouting louder and more aggressively because they feel like they are not being heard. I've been trying to create systems that makes it easier for people to actively listen to each other.”

The city of Seattle is currently using Consider.It to collect feedback on a visioning document for the city’s future growth. The format makes it easy to see where people agree and disagree, and does so in a way that’s engaging (and not off-putting.)
Video comments abound on sites like YouTube, where the customary response to a video is to make a video adding commentary or reflection, and on Twitch, where people talk to each other while watching other people play video games or paint pictures. There are reader polling sites like Pol.is, which collect opinions and groups similar ones together. And arguman, a “argument analysis platform, which maps arguments using logic. Auditor allows people to respond to blog posts or websites using audio. Both Kinja and Product Hunt allow their audience to express themselves using animated gifs. Rabble.tv lets people add their own live commentary to live sporting events — essentially replacing paid commentators with a streaming, live comment section.

There’s also hidden commenting — I’m thinking of things like private Facebook groups or Slack rooms, which might only be available to those who know they exist. And emoji-based commenting. Think Periscope, which allows people to add comments in the form of stars and hearts. (Emoji as metadata is also something Facebook does and Buzzfeed experiments with at the end of articles.) (Related: Can stickers be a type of commenting platform?)

And then there’s Motherboard, which is experimenting with asking for reader’s phone numbers so that staff members can talk to them on the phone, which reminds me of McSweeney’s Kickstarter campaign, which offered personalized book recommendations from its staff to contributors.

There are also design considerations. Should comments look like boxy comments? Could they look completely different? How do we design the comments in a way so that threads are easily followed? What does a conversation look like on Web versus a phone versus a push notification? (What if push notifications were comments?)

I return back to the homepage experiment I ran with my friends, none of whom worked in news. We came up with lots of ideas — some good, some not so good. It was our way to expand what a homepage could be.

And now I ask you: Where do we imagine comments going? Does it contain text? Audio? Video? Emoji? None of the above? Does it ask the writer to add to the discussion or ask other readers to get offline? Does it further our understanding of whatever it is we’re reading about?

I’d love to hear your thoughts — or should I say, your comments.

(Thank you to everyone who responded to this tweet, as well as Andrew Losowsky and Jacob Harris — both of whom added many suggestions for great commenting platforms.)