AI Could Improve Quality and Civility of Online Discussions, Say Researchers
Online discussions and comment sections have become a breeding ground for arguments, threats, and name-calling, especially when it comes to divisive topics. However, researchers from Brigham Young University (BYU) and Duke University believe that AI could be the solution to improving the quality and civility of these conversations.
The researchers conducted a field experiment, involving 1,574 participants, to test their AI-powered chat assistance. The participants were asked to engage in an online discussion about gun regulation in the US, a topic known for its political controversy. Each participant was paired with someone who held an opposing view on gun policies.
In the experiment, conversation pairs were randomly assigned to either the treatment or control condition. In the treated conversations, one participant received three suggestions from GPT-3, an AI language model, for rephrasing their message before sending it. The participants had the option to choose one of the AI-suggested alternatives, stick to their original message, or make edits.
Throughout the experiment, a total of 2,742 AI-generated rephrasings were suggested, and participants accepted these suggestions two-thirds of the time. Chat partners of individuals who implemented AI rephrasing suggestions reported significantly higher conversation quality and were more open to listening to their political opponents’ perspectives.
According to David Wingate, one of the study’s co-authors, participants felt that the conversation became less divisive and that they were heard and understood when the AI rephrasings were used more frequently. This indicates that AI intervention can play a crucial role in combatting the toxic online culture that has plagued the internet.
The researchers believe that their findings demonstrate the potential for scalable AI solutions to enhance online discussions. Compared to traditional methods like professional training sessions, AI intervention can be broadly implemented across various digital channels, making it more accessible and effective.
Wingate also expressed his hope that more students from BYU and other institutions would develop pro-social AI applications. He envisions a future where machine learning is used ethically and positively to navigate the world’s information.
Ultimately, this study highlights the positive impact that AI can have on creating a more respectful and empathetic online landscape. By providing users with suggestions for rephrasing their comments, AI offers the opportunity for conversations to be more civil and productive.
The study, which was published in the journal PNAS, serves as an important step towards leveraging AI to foster better online discussions. As the digital landscape continues to evolve, it is crucial to explore innovative solutions that promote kindness and safety in our online interactions.