Can machines change human behavior? OpenWeb, using Jigsaw's Perspective API, releases case study measuring the effects of real-time feedback and "nudges" in decreasing toxicity in online discussions
The study measured the impact of deploying a "nudge" to potentially offensive or profane comments, encouraging commenters to reevaluate their message. The goal of the test was to measure if such a nudge could reduce toxicity within conversations. The study involved more than 400,000 comments across a select number of OpenWeb's partner publishers, including AOL, Salon, RT, and Newsweek.
In addition to testing the effect of the "nudge," OpenWeb created a variety of messages to measure how users react to different statements. The study then assessed the effect that responses to the nudge had on overall community behavior and toxicity.
Highlights from the study:
- 400,000 comments and 50k users were analyzed over 3 months, from May through
- 34% of commenters who received a nudge edited their comments
- Of those who edited, 54% changed it to be immediately permissible
- 45% removed or replaced the toxic element
- 8% reshaped their entire comment
- 12.5% lift in civil and thoughtful comments being published overall
"The results of this case study are further evidence that technology can have an effect on human behavior," said
"We're thrilled to see more platforms using Perspective to help host better conversations online, especially platforms with such diverse and comprehensive audiences as OpenWeb. Machine learning has the potential to open up more spaces online for people to express themselves and interact with the community, but the technology is only as useful as the partners who implement it and bring it to users. This case study shows how Perspective can deliver measurable improvements in reducing toxicity. We're proud to be working with OpenWeb to bring better conversations to millions of users across the web," said
"Ensuring that every member of the Salon community has an engaging experience with our features, videos, and our comments has always been a priority for us," said
OpenWeb aims to improve the quality of online discourse through extensive, layered, moderation technology and proprietary AI, and runs Perspective ML models on millions of comments to catch abuse and toxicity in real-time. This Real-Time Feedback feature gives users an opportunity to change their message if it is suspected to break the Community Guidelines, using the nudge concept, a known theory in behavioral sciences that proposes positive reinforcement and indirect suggestions as ways to influence the behavior and decision making of groups or individuals.
"We believe encouraging good behavior is equally as important as removing the bad actors to support a thriving community," said
While initially deployed as a test, due to the findings of the study, OpenWeb has begun rolling out the feature to its entire network of publishers.
To see the complete case study, and learn about OpenWeb's multilayered moderation, visit OpenWeb's blog.
OpenWeb (formerly Spot.IM) is on a mission to democratize and improve conversations online. The platform utilizes AI and machine learning to incentivize healthy dialogue, decrease toxicity, and create thriving, engaged communities. OpenWeb works with more than 700 top-tier publishers and hosts 100 million active users each month.
Founded in 2012, OpenWeb has 100 employees in
View original content to download multimedia:http://www.prnewswire.com/news-releases/can-machines-change-human-behavior-openweb-using-jigsaws-perspective-api-releases-case-study-measuring-the-effects-of-real-time-feedback-and-nudges-in-decreasing-toxicity-in-online-discussions-301134841.html