An algorithm for online civility
First came the inter-worldwide web-net, and soon after came the rudeness. There’s something about the anonymity and speed of online conversations that triggers hasty replies and the it-seemed-clever-at-the-time snapbacks. One researcher named this the Online Disinhibition Effect, which explains how online rudeness seems OK even to basically polite people.
While there have been tons of studies on why this happens, no one has yet figured out how to stop it from happening in the first place. When a conversation degrades and the insults start flying faster than an Olympic-level table tennis game, it’s usually hard to reestablish civility.
A recent paper from researchers at Cornell University might hold the answer. They developed a predictive model to discover if an online conversation is heading downhill before it happens—what they call the “intriguing phenomenon of conversational derailment.”
By using machine learning and crowdsourced vetting to label conversations and by identifying warning signs—“linguistic cues”—that indicate a good convo is about to go bad, the researchers were able to predict if an exchange would get mean. One interesting finding: asking direct questions or starting with the word you are good indicators that some nastiness is on the horizon.
There’s more work to do: subtle and casual inferences, for example, might not be caught by this model in its current form. But the good news is that anyone can work to advance the civility of comments—the Cornell Conversational Analysis Toolkit is on GitHub. Will this lead to a future of crowdsourced civility? Here’s hoping.
The startup as public menace
San Franciscans called it “Scootergeddon.” In the spring, seemingly overnight, an influx of stand-up electric scooters appeared on the streets in cities across the United States. The complaints began soon after.
Scooters are environmentally friendly, inexpensive to rent, and easy to ride. But they were quickly decried as yet another arrogant startup enabling bad urban behavior. Scooter companies have taken the approach favored by Uber and other startups of “launch first, deal with regulations later.” But they’ve already scooted into some Uber barriers.
San Francisco was not amused. The city kicked the scooters, which it called a “public nuisance,” off the streets, while the companies applied for permits. Other cities, including Denver and Nashville, followed its lead. (Indianapolis asked one operator to “suspend operations for 30 days”; the company said no.)
City lawmakers are quicker to regulate now, probably because of previous experiences trying to reign in “disruptions” after the fact (or launch). Is this the beginning of the end of cities as publicly funded test labs for the latest startup idea?
Part of the problem is that the streets aren’t a sandbox where you can observe what users will do—and what can go wrong. Scooter riders, for example, should be wearing helmets, but they’re not; they’re riding on sidewalks, which is illegal (although the scooter startup Bird, founded by former Uber and Lyft exec Travis VanderZanden, is proposing to make it otherwise); and they’re leaving the dockless scooters all over the place.
Scooter-mania could prove to be the test for cities going forward. The devices encountered similar objections when they launched in Paris recently, although Bird competitor Lime has reached agreement with the city to operate there, provided the scooters stay off the sidewalks. There’s always another idea right around the bend. Coming soon: moped sharing. Let’s see how cities react to that one.
Influencer bot Lil Miquela
Credit: Lil Miquela
Influencer bot Shudu Gram
Credit: Shudu Gram
I post, therefore I am?
Klout is no more. The platform that made a big splash when it started in 2008 (and that you immediately forgot about) aimed to assess social media reach and score users’ subject expertise. It came under fire, however, for its approach to privacy and also for some nonsensical “expertise” ranking results.
In some ways, Klout now seems quaint, maybe because we’ve since become so embedded in social media and influencer life. Thanks in no small part to Instagram (which came along in 2010—eons ago in Social Media Time), we have professional influencers, talent agencies especially for them, and brands paying to benefit. (Fraud, too, and a growing backlash.)
But Klout, which parent company Lithium shuttered in May 2018, must have left some type of void, because an app has launched to fill it. Skorr purports to measure social media influence across platforms by using machine learning and artificial intelligence. Does tracking your posts’ performance in real time sound obsessive and slightly unhealthy? We’ll leave that to the psychologists. The app’s ultimate target market is pro posters whose livelihoods are tied to their social media presence.
Maybe not for much longer. Influencer bots are coming for them. Lil Miquela and Shudu Gram are both avatars with thousands of followers—Lil Miquela has over a million. Interestingly, both bots post in first person. Which begs these questions: Do the U.S. Federal Trade Commission’s sponsorship rules, which enforce making business relationships between poster and brands explicit, apply to CGI? Will consumers be convinced that a bot can vouch for products based on “personal” experience? What’s personal, anyway? I post, therefore I am? This is getting way too metaphysical. D!