August 10, 2017 (Saving Seafood) — WASHINGTON — Many news sites’ comment sections have devolved into miserable places for conversation on the internet. Online insults and internet trolling has become endemic, and even the target of an ongoing storyline on South Park. Because we don’t have the staff to monitor comments, until now, Saving Seafood has not included comment sections in our posts. But in our current regulatory and political climate, the fishing community should have a public place to discuss the issues we confront, and one that is relatively safe from trolls and bots.
So beginning today, we are adding commenting on Saving Seafood posts, using a service called Civil Comments. The system is unique and one we think will create a lively and engaging community that we hope you’ll join. With Civil Comments, you’ll be asked to rate the civility of several other comments from the site — and then your own comment — before you’re allowed to submit it. This kind of crowdsourced moderation is a unique way to try to ensure higher quality comments without the heavy hands of moderators deleting offending posts. Once you start commenting, and rating others’ posts, the machine-learning algorithms will get to know our commenters, and its ability to weed out trolls and spam will gradually improve. The system is programmed to quickly detect attempts to game the system by voting disingenuously on others’ posts.
Here are some questions and answers about the new system, from the Civil Comments FAQ page:
Why do I have to rate other people’s posts?
Internet comments can be a place for spirited, meaningful debate about all kinds of topics. All too often, though, they turn into a festering garbage fire of harassment, abuse, and spam. Computer algorithms aren’t yet good enough to accurately determe meaning from free text, but human beings are great at it! To help keep your community civil and respectful (and free of annoying spam), we ask that commenters pitch in by helping with moderation. Over time, with proven civility and trust, this rating system becomes opt-in, as a reward for qualified commenters.
How do you keep people from gaming the system? Won’t users just vote down posts they disagree with? What if the majority votes down minority opinions, even if they’re civil?
We were worried about this, too! It’s why we put tons of thought, care, and effort into designing the backend algorithms that analyze the reviews, to prevent bias. Having a “civil” comments section doesn’t mean a groupthink echo chamber. Each comment is reviewed by multiple people, and we pool and analyze those reviews to check for patterns of abuse. We also use that data to help decide who sees which comments for review, and we anonymize veteran commenters to prevent bias against specific users. So far, it’s been working well! One great finding has been how rare the attempts at cheating/gaming the system actually are. We’ve found that the overwhelming majority of people actually vote in good faith, so +1 for humanity. With Civil Comments®, people are attacking ideas, instead of each other.
What do you mean by “good” comments vs. “civil” comments?
For each comment, we first ask the reviewer, “is this a good comment?” Here, “good” means whatever you, personally, think is “good.” Does the comment contribute to the conversation in a meaningful way? Maybe it’s funny or insightful, or you just agree with it. If you have no opinion on a comment’s quality, you can choose the middle option of “Sorta” or “Somewhat”. If you really think the comment is bad—off-topic, poorly-reasoned, or you just don’t agree with it—you can choose the “No” option. Answers to this first question don’t affect whether or not a comment is published, but they do contribute to the way comments are ranked when sorted by “Highest Rated” (along with “likes” received).
The second question our system asks is, “Is it civil?” A civil comment can still be angry and passionate, but cannot include threats, personal attacks, name-calling, racial slurs, doxxing/de-anonymizing (revealing personal information about another commenter, including their real name, home address, phone number, etc.) and obvious spam.
Think about what would be considered “rude” in real life; it’s a good guideline for what we mean by “uncivil.”
Isn’t this censorship? What about free speech?
The first amendment guarantees that no one can be thrown in jail for what they say. We’re strong believers in freedom of speech and freedom of the press, but our stance can be summarized by the old adage, “my right to extend my fist ends where your face begins.” Spam, personal attacks, harassment, abuse, and even death threats are so common in comment sections that many sites are just turning comments off entirely, which effectively silences everyone. Sites that do keep comment sections open need to be constantly vigilant, with staff policing the rampant bad behavior and deleting spam. Since comment sections clearly do need moderation, we think it’s far more fair for your comment to be judged by several of your peers—the people with whom you share the community—than by a single moderator, who is separated from the community and typically employed by the site.
Why don’t sites just hire moderators?
Most publications just don’t have the time and/or resources to hire moderators to monitor the comments 24/7. A few very large sites have full-time staffs dedicated to this, but some spend over $1 million per year on moderation, and still must limit the number of comments they publish, due to staffing constraints. With traditional moderation, each comment is only be seen by one person, where it receives one pass/fail grade. With Civil Comments, each comment is judged by several people. Our patent-pending Behavior Engine then analyzes those judgments and attempts to account and correct for bias and people trying to cheat the system.
Despite all of your safeguards, I saw some spam/harassment/abuse in the comments!
Oh, no! Our system works really well so far, but it’s still new and improving. Sometimes uncivil things will slip through the cracks—that’s why the we have a flag button! If you see anything that qualifies as abuse, harassment, or spam, you can flag it to notify publication staff. Keep in mind that flagging without valid reason will have a negative impact on a user’s trust rating, though, and their flags will cease to “count.” So, please only flag if the comment is truly problematic!