Visualizing Comments by Gender at the NYT

I recently read Emma Pierson’s study about commenters and gender at the NYT. I thought it was a great piece with compelling data, some of which I tried to pull out in the following infographic.

A few challenges: the program I used didn’t allow me a lot of flexibility in terms of editing the charts, so I had to be creative about which points I chose to pull out of her findings. This visualization also uses word clouds, which some folks find terribly unsophisticated, but I really liked the visual comparison of the types of words that men and women use in comments on the same articles side by side.

Without further ado, here’s the visualization…(unfortunately, I had to paste in a screenshot because the original png file wouldn’t copy into this, so the quality on this version is a little lower than I would have hoped)

comments and gender snip

Writing A Code of Conduct For Your Nerd Space


Note: No fancy tools for this one. A write-up.

Earlier this week, female gamer and Tumblr user Latining published a harrowing post titled “Tabletop Gaming Has a White Male Terrorism Problem.”The post describes in graphic detail the harassment and outright sexual assault the author experienced in gaming stores and conventions, and the general apathy and hostility other members of the game community and law enforcement showed when she tried to report it. Sexual harassment and assault, she writes, are tabletop gaming’s “white male terrorism problem.”

The post exploded within the gaming community. It was retweeted by feminist gamer and developer Brianna Wu, debated on BoardGameGeek, and picked up on reddit.

It prompted soul-searching, rebuttals, and support.

While the original author’s claims are being debated in the wider world, she is right about one thing: this isn’t the first time misogyny has reared its head in nerd culture (yep, each of those words goes to a different link) or in nerd spaces.

So the question is: why is this still happening, and how have people taken action?

Jessamyn West, formerly a head moderator at the comment community MetaFilter, suggested a solution: “Some initial work at creating practical and enforceable ground rules can keep every contentious discussion from turning into a first-principles slugfest.”

Jessamyn has years of experience in community management, and knows that rules matter. They send a signal and they set a standard. Many groups – conventions, Meetups, stores, movements and forums – basically any space where people gather and interact – have adopted codes of conduct. And you can adopt one too, even if your space is just six people who get together to play games every Friday night.

Bad Codes of Conduct are fusty bits of bureaucratic language that no one ever reads, handed down from one generation of bosses to another with little feedback from a community. Kind of like Terms of Service, if TOS were even less cool. The worst codes of conduct are those are those that are never enforced, because lack of enforcement “sends the message that the values in the code of conduct aren’t actually important or respected in your community.

But at their best, Codes of Conduct can provide a space for a community to gather, share experiences, and discuss how they want to govern themselves. They can turn good intentions into a system, and open space for dialogue.

If the word “code” sounds too authoritarian, you can opt for something like “Community guidelines” or “Our Principles” or even something like “What We Can and Can’t Do Here.”

Here are some guidelines to get you started:

According to the Geek Feminism Wiki, a good code of conduct includes the following:

  • Specific descriptions of common but unacceptable behavior (sexist jokes, etc.)
  • Reporting instructions with contact information
  • Information about how it may be enforced
  • A clear demarcation between unacceptable behaviour (which may be reported per the reporting instructions and may have severe consequences for the perpetrator) and community guidelines such as general disagreement resolution.

Geek Feminism also includes a list of links to organizations that have adopted codes of conduct, and an evaluation of these codes.

I’ll add my own two cents:

A code of conduct should send a strong signal about how you want people in your group to treat each other. It sends that signal to those who might abuse the space, but it especially sends a message to members of marginalized groups who might otherwise be wary because they are used to being unwelcome or unprotected. The folks at Cosplay is Not Consent link to a photo of the Jekyll Comic Con’s “Respect Thy Fellow Cosplayer” policy. Is the policy perfect? No. But that’s not the point. The point is that it’s there, and the organizers put thought and energy into crafting it, and its existence signals that they’re (hopefully!) not going to ignore complaints or reports of inappropriate behavior.

But a good code of conduct goes a step further. It answers practical, thorny questions. How does your group feel about anonymity? And if you allow it, how do you signal it? What about confidentiality? If someone faces a threat, what phone number should they call? What person – in a position of authority – can they talk to? It forces your group to construct a pipeline of responsibility, rather than just trusting to people’s hazy good intentions. Any space, no matter how small or informal, can benefit from having policies around these questions.

It provides an opportunity for your community to debate and discuss how you feel about norms. If you have an old code of conduct that was written years ago, maybe it’s time to pull it out and have a discussion about it. A good code should evolve with its community, and be open to feedback. It should be something that the people in your space buy into.

Finally, at least according to the team at the Ada Initiative, it should get specific: “The major weapon of harassers is arguing whether something is actually harassing. It is difficult to enforce a CoC if you have to have a month long nasty argument about whether it was violated. It burns out people like you.” A good code of conduct saves you – a person who wants to have a good time in your space – from tedious discussions. You can just point at it and say, ‘this is how we work.’

If you need guidance on how to word your code, here’s some free language courtesy of Geek Feminism.

In reading the responses to the original Tumblr post, I was struck by one from gamer Samwise Seven RPG, who said “In all honesty, I tend to stay away from gender and race topics when they concern the role playing game community. My internal dialogue tends to be: can’t we just play elf games and get along while we forget the real world and all of its miseries.”

If only everyone could.

Posted in All

The Government Versus Apple

I’ve come across timeline.js before, and decided to  give it a go for this assignment. I chose to pick up the threads of the battle between the FBI, the US DOJ, Apple, and others in the tech industry over the unlocking of the San Bernardino shooting’s suspects’ iPhone. There have been innumerable explainers penned on this subject, some better than others, but considering the mountain of legal paper that’s built up, I thought it might be interesting and worthwhile to go for a chronological layout of what’s been said and done. Also, the government announced yesterday that it decrypted the iPhone, so this seemed like a great and timely subject

General Observations

This was not the most straightforward application. The spreadsheet-based interface takes a bit of getting used to and isn’t remotely intuitive, although a glance at the documentation and some tinkering makes it clearer how to use it. This is what the spreadsheet looked like while I was filling it out:


When I finally published the story, I noticed one significant challenge: I don’t know why, but the timeline.js interface could neither capture nor card any of the news organization websites/stories I linked to (although it did a beautiful job with YouTube). I finally opted for a strange workaround: I either took screen captures of the news stories and used those as static images (misleading from a UX perspective, since the user will almost certainly expect these to be clickable links) or I saved the stories as PDFs, hosted them on my own website, and then made them available as PDFs within timeline (less attractive, and still loses some of the richness of the original document).

Another problem: their interface doesn’t explain this, but in order for timeline.js to interpret your data and render it, you need to “publish your spreadsheet”. It becomes public, thereby exposing your data to the world. I don’t know if there’s a way to publish the spreadsheet privately and share the link with timeline.js, but I feel like this is something that could possibly be better outlined/addressed.

That said, one thing I did like was that it was easily possible to update the spreadsheet (once I published it). Live changes made to the document fed immediately into the rendering, making editing fairly breezy (I seemed to remember this from before.)


In retrospect, although this does function as an explainer, it’s a bit extensive and requires a pretty serious reader commitment in order to fully grasp. I’m glad I worked with the tool, and I think the chronological order helped, but there’s probably an even better way to build interactive tools for laypeople to read legal documents, and this project made me aware of the need for that.

Click here to check out the end result.

Posted in All

Thanksgiving Dinner: A Role-Playing Game for Handling Controversy

Creating the Game

Christina Houle and I decided to team up for this assignment. We tossed around a few ideas for creating something interactive, and she mentioned that she’d previously used Muzzy Lane to create an online interactive game (Muzzy Lane builds software that in turn allows teachers and educators to create games for learning). Christina and I wanted to capture the ways that discussions unfold in real time, while at the same time offering people feedback on argumentation strategies. We thought it would be interesting to allow people to role play a difficult conversation online. By offering players multiple response options (as well as feedback on those responses), we thought the exercise could become more interesting and demonstrate practically how to lead arguments with values.

We decided that our role play scenario was going to be Thanksgiving dinner with a friend’s family. Why Thanksgiving dinner? When we started talking about our own experiences with controversial conversations, we found that these tough conversations often happened with family members. What makes disagreements in this context so difficult is that we care about the people involved, and can’t just walk away even when disagreements can be profound.

The topic we wanted to explore: paid family leave. This is exactly the kind of subject on which members of a family might have very different views. We wanted to bring out the family dynamic, as well as allow different family members to share their experiences.

Our scenario:

“You’re visiting your friend Rita’s family for Thanksgiving Dinner. You’ve never met any other member of the family, and don’t know what people’s political beliefs are. After a warm welcome, you all sit down to dinner. The topic turns to paid family leave – a discussion that has been much in the news. As you navigate the conversation, your goal is to learn what other people’s values are, and use what you’ve learned to guide your responses to what other people say. Hopefully, you’ll learn something new while still advocating for your own position – which is that paid family leave in the United States should be expanded.”


Rather than grading responses as right or wrong, we allowed players to earn points for “judgment” or “values.” When players choose to lead with values – which means understanding another character’s point of view – they get a point for values. If, however, they opt to go straight for fact-based confrontation, they earn a point for judgment. At the end of the game, they get a total score and some general feedback on strategy.

Link to the beta version of our game:

Our process:

Muzzy Lane’s interface is fantastic! Here’s how we created our exercise, followed by a few screenshots from the actual gameplay.

So You Want to Read: Zimbabwean Fiction

This was a really challenging assignment, because it forced me to think hard about what qualified someone as an expert. I didn’t want to take the easy route and interview Fungai about his work. Although it’s a great and interesting topic, it’s also something that he talks about all the time. I did a little Facebook stalking and noticed that he posted a lot about literature. I thought it would be great to put together a contextual primer on literature by Zimbabwean authors. I was interested in the subject, and I knew very little about it. I liked the idea of a “contextual reading list.” One of the things I find frustrating about reviews on sites like Amazon’s is that book reviews are provided independent of wider context. For ex: who is the person doing the review? What expectations, history or loyalties do they bring to the review process? I thought Fungai could serve as a great expert. In addition to being a media maker, he has deep emotional and personal connections to the material that make for interesting reading on their own.

My second goal for this assignment was to try and find a good use for FOLD. I think material like reviews makes for great FOLD stories because the contextual embedding format makes great sense. So I thought I’d use FOLD to juxtapose narratives. On the one side, in the main story blocks, I’d have Fungai talking about his experiences with some of these key novels and texts by Zimbabwean writers. In the right-side blocks, I posted links, reviews and other material related to those texts so that readers could go off and explore more on their own (which was the whole goal). The goal of a review is to communicate the essence of a book. But by using FOLD’s capabilities in this way, I was able to focus on the interview sections on Fungai’s experience, and to package the summaries/etc in the links I presented on the side. It was a great and interesting experience, and I’d be very interested if anyone has feedback on what it was like to read in this format.

Story here.

A Little Too Much Screen Time

This post is about two things: 1) The arbitrariness of the ways in RescueTime (and I!) define media and 2) the realization that I really need to spend far, far less time looking at screens. (Far, far less.)

I used RescueTime to track my media consumption, but then threw that data into buckets. A few attempts to wrestle through some of the questions I found interesting, below.

News, Entertainment & Social

NEShdThis first rough slice looks at three categories: the time I spent on news, social and entertainment. I used these categories because RescueTime uses them, but I already see ample space for confusion and critique. We’re already playing in somewhat arbitrary territory, in the sense that the time I spent on BuzzFeed got classified as news consumption, but could possibly have gone into entertainment (was it a listicle? Was the listicle newsworthy?). Meanwhile, BBC America went into news, but I’m pretty sure I spent those 11 minutes looking at Doctor Who action figures. On a more substantive note: there is genuine slippage between categories, here. For example, the “Entertainment” category includes all of YouTube (I didn’t track each video separately, but I almost wish I had). Part of that YouTube time was definitely informational videos about gravitational waves, and part of it was “Last Week Tonight,” a product that easily belongs in more than one category. Meanwhile, most obviously, “Social” is both news and entertainment. In an era when I’ll see Tweets about an earthquake before it appears in the news, I’m genuinely getting breaking news off social media (never mind the more hyperlocal ‘news’ about my friends’ lives.) Finally: let’s stop pretending the news and entertainment are separate. I don’t think that helps anyone.

News, a Breakdown by Source

News Breakdown I found this breakdown interesting. Here I’ve broken down “news” by point of origin, excluding social (ie, FB). The big green slice towards the bottom is blogs. This includes Medium and WordPress sites, and they dwarf any other individual news site. Also, I don’t know what killed news brand loyalty (search? social?) but it’s clearly dead.



Users Versus Big Media 

UGC vs Big Mediahd In the news and entertainment categories, I thought it would be interesting to look at media based upon who produced it. This is where it all got interesting. The clunky term “user-generated” (which still roams at large in many newsrooms) is both off-putting and, it turns out, not very useful. The “User-generated” slice here includes everything I watched on YouTube. When I think about the stuff I watch you YT, it includes a ton of Youtube cover bands, makeup tutorials (deal with it), and comedy groups. My point: I spend a lot of time consuming media on YouTube that is produced by neither Big Media nor casual users, but by some in-between professional class of producers whose only distribution mechanism is YouTube. Some of these folks are quite famous, some of them make a comfortable living off what they do on YouTube. But they’re not part of the group of producers we consider Big Media. Meanwhile, User-gen also includes Reddit, which is some strange and uncomfortable hybrid of user-generated and big media.

User-gen, breakdown

User GeneratedhdThis chart demonstrates the weirdness of “user-generated” as a term, still further. I’ve got content from Reddit, Blogs (Medium, Medium’s Matter publication, etc) and YouTube all jumbled together in here. None of this is produced by ‘users’ in isolation – even platforms like Reddit and YouTube enjoy massive monopolies over particular types of interaction, and they are hardly small players. They provide development, editing and interaction tools on their platforms, thereby participating in the process of content creation. Therefore, the final product, even on a site like Reddit, is partnership-generated. This content isn’t generated by professional journalists, but we (professional journalists) might be the only people who care about that distinction anymore.

News vs. Telegram

All news vs One Social siteThis chart looks depressing, but it’s actually possibly vaguely heartening.  Remember up top, when I said I need to spend less time with screens? That’s true, certainly, but it isn’t true that screen time =/= relationship time. In this chart, I juxtapose all the time I spent on news this week against the time I spent on Telegram, the social networking site I use the most. Here’s the interesting part: I only use Telegram to talk to a handful of friends. That means that I spent 6 hours and 52 minutes talking to approximately four friends, online. In a way, this is an homage to relationship-building, isn’t it? Even when I’m online, the #1 thing I’m doing is maintaining relationships with a core group of friends. The most interesting part of all this is that these are the four friends I probably hang out with the most in daily life, too. Interesting? Weird? Cult-ish? Maybe a bit of all three, but at least I feel a little less like an antisocial nutjob.

Print vs. Online

Online vs. PrintUnfortunately, I can’t find the exact quote, but a few months ago a fairly senior official at a big news publisher made headlines (Tweetlines?) by suggesting that young people will one day get tired of consuming news online and will go back to print newspapers. Whoever he is, wherever he is, this chart is for him.



Video vs. the World

Video vs Rest This is another one of those charts I kinda struggled with, but it gets at a distinction I wanted to address. Because I’ve been using RescueTime for a while, very little of what I’ve discovered this week surprises me. That said, I was reminded just how much time I spend watching video content online. (And it’s interesting, because despite overwhelming evidence to the contrary, I still think of myself as a ‘reader’.) What’s interesting is that this statistic might actually underestimate the amount of time I spend watching video, because it doesn’t count the time I spent on, say, a Vox media page watching an embedded video on that page. (‘Video’, here, only counts time on video-specific sites like YouTube or Netflix.) This also raises really interesting questions about video distribution. BuzzFeed makes a ton of money off video. Here’s a quote from an article about their financial statements:

“You have to remember,” Dempsey added, “that BuzzFeed doesn’t operate on any sort of subscription model, is growing at a significantly higher rate versus traditional media companies, and also is doing a lot in their original video content, which is largely viewed outside of, almost making this part of the business more comparable to an original content company vs. digital media publication.”

Part of winning the video play, at least if you’re taking a page out of the BuzzFeed playbook (and who isn’t, these days?), is being seen on aggregators like YouTube and Netflix. How many news publishers are serious about building audience on YouTube (I know what you’re thinking: does John Oliver count?) Meanwhile, if you search Netflix for news, what do you get? Nothing at all.

Note: if the image quality on these charts is lame, I’ve got higher resolution pdfs I can share.

Civil Comments: A New Way to Create Civil Online Conversation?

The tool I’m interested in is Civil Comments.

How it Works

Their video offers a great overview of the system.

Civil Comments – How It Works from Civil Co. on Vimeo.

Why It’s Interesting

The Civil Comments tool provides an interesting intervention in the comments space, because it offers a unique but also intuitive answer to some of the key problems facing online content publishers when it comes to their comments.

  1. The Expense of Moderating, Especially at Scale – many content publishers currently face a huge problem when it comes to moderating comments. Most moderation requires a human editor. Although many tools will automatically filter out abusive terms, it still takes a human moderator with judgment (and maturity!) to read user-flagged comments. For publishers who deal with heavy content volume, human moderation can be very expensive. By leveraging the power of the audience, Civil offers to make comment moderation free, and scalable. This is huge. The Civil interface is also pretty simple – it looks a lot like TripAdvisor’s or Amazon’s ratings, which are proven interfaces that people like to use.
  2. Limiting incentives to abuse – the most thought-provoking claim that Civil makes is that they’re able to single out abusive commenters through this crowd-sourced system. I’m not entirely sure if that’s the case, but their initial run with Willamette week appears to have garnered some positive reviews. Although untrained folks may NOT always be able to filter out abusive comments, this crowd-moderated system raises some interesting questions about the incentives to post uncivil commentary. If comments are social, as many people (including Joseph Reagle of Northeastern, in his book on the topic) have suggested, then that begs the question: if only 1 or 2 people at most are going to see an abusive comment before it gets buried, will trolls even want to post abuse in this kind of system? Is the incentive to abuse lessened when there’s no audience? When it comes to mass troll attacks, Civil claims they have a system that will detect them.
  3. Hierarchy? Values? By enabling a form of peer moderation, it’s possible that publishers who use the Civil system will send a positive message about the role that their community plays in setting the site’s values. It also marks the comment section as an independent space, one where both readers and journalists get to set priorities. At the same time, because readers get random comments to review, this peer moderation system might offer ways to avoid some of the bias (towards highly ranked commenters, towards familiar commenters, towards early comments) that other peer moderation systems are prone to (Lampe et al).

What it Won’t (Necessarily) Do

  1. Eliminate issues of site-wide bias: Many moderators of peer moderated sites whom I’ve spoken to have mentioned that their sites have a particular political bias. I don’t see that Civil will address site-wide bias very effectively, especially considering that people tend to moderate comments more favorably when those comments reflect their own views.
  2. Invite minority views/communities into conversations. One of the moderators whom I spoke with offered a compelling case study of how their site had drawn flak from trans members about transphobic language. The moderators made an executive decision to change community norms, and enforce those changes, even though the majority of site users weren’t as affected by the issue. Sometimes moderators might want to enforce values/etc that the community does not. How are these more subtle social norms introduced? How are they maintained and shown to new members who are visiting for the first time? It seems like the initial judgment made by the crowd might be a large-grained filter at best, and exclusionary at worst.
  3. Protect identities and data. Conspicuously absent from the Civil Comments’ webpage: any mention of what happens to users’ comment data. Civil says that they offer analytics, which means that they must collect data or offer a data collection option. But publishers run their own ‘instances’ of Civil. How are those data stored and anonymized? Who has access? Will Civil turn around and sell that information? Particularly relevant in conjunction with point #2, but problems of online harassment in general.

Anika Gupta


I’m Anika Gupta, a researcher and journalist (and currently a CMS master’s student). I’ve written about science, technology and entrepreneurship for various publications around the world (links to articles here, in case you’re interested in that kind of thing) but these days I’m more interested in the links/interfaces between news organizations and users. I’m doing a thesis that looks at issues related to online comments in context of the evolving relationship between news organizations and audiences.

I am also very interested in the globalization of news. I spent five years working as a journalist and product manager in New Delhi, and started Hacks/Hackers New Delhi (now Hacks/Hackers India). I’m really looking forward to meeting and learning about the other people in this class! My skills include: user research, product management, mutilingual/multicultural audience development and traditional narrative journalism!

My Twitter.

2015-11-09 16.53.09