A new Cornell University study identifies a framework for determining when online discussions will turn ugly. By analyzing conversations between Wikipedia editors, the research team developed a computer model to predict, with 65% accuracy, when interactions would become "toxic."
People can identify these discussions with 72% accuracy. You can test whether you can identify which conversations will turn into personal attacks using an online tool.
The model analyzed politeness and other rhetorical strategies, for example, editors' use of "please," expressions of gratitude, and fact-checking. When editors used more direct questions and "you," conversations were more likely to go awry.
- How did you do on the online quiz? What did you learn from the experience?
- What is the value of this research? Consider social media sites such as Twitter. How could the model be used, and what are the potential positive and negative consequences?