Two of college students’ favorite pastimes — social media and arguing—were topics of a recent Cornell study. The paper titled “Winning Arguments: Interaction Dynamics and Persuasion Strategies in Good-faith Online Discussions” was published on arXiv — an online e-print service owned by Cornell.
By using the ChangeMyView debate platform on Reddit, the research team had unique access to a sample of people dedicated to reasoned debate and the exchange of ideas. Grad Vlad Niculae, one of the paper’s authors explained why CMV was a great platform to study.
“CMV offers a combination of conditions that are very fortunate for our research purposes,” Niculae said. “It [CMV] is a well-moderated community where interactions consist of attempts at persuasion and open-minded discussion, all in one place.”
Social media platforms like Facebook, on the other hand, are notorious for unreasoned, inflammatory comments and as such are not ideal for studying how people’s minds are changed during debate, Niculae said.
“An attempt to find successful acts of persuasion in the wild Internet is like looking for a needle in a haystack; finding multiple comparable arguments against the same view is near impossible,” Niculae said.
CMV, then, was a godsend.
The crew found that certain linguistic strategies correlated with higher success at persuasion. For example, word count is positively associated with persuasive success — the longer your response, the greater your chances at convincing someone. On the flip side, bulleted lists seem to do little in terms of one’s ability to make a convincing argument.
Niculae said that the team found some surprising results — most of which related to word usage.
“One surprising finding to us is that within two comparable arguments, the effective one turns out to be less similar to the original opinion in terms of word usage,” Niculae said. “This may indicate that offering a new perspective using different words makes one less defensive.”
Furthermore, exciting words were found at the end of a sentence.
“We were intrigued by the trajectories that certain factors take over the course of an argumentative Reddit message,” Niculae said. “For instance, all messages used more arousing, exciting words towards the end than at the beginning. But the successful ones start off much calmer than those that failed.”
Of course, not all arguments prove successful in persuading others, so the researchers examined how arguments fail in addition to how they succeed. Thanks to a CMV requirement that participants state their beliefs and reasons for holding them clearly, Niculae and Crew were able to glean differences in the sorts of opinions subject to change and those resistant to it.
They found, for example, that some beliefs are simply less susceptible to change than others.
“The language used can reveal interesting aspects about the relationship between a person and their opinion. For instance, do they identify it deeply with their individual world view eg. ‘I, for one, believe that’ or do they diffuse responsibility within a larger group eg. ‘we believe that’?”
In that vein, the research team found that personal pronoun usage is indicative of a malleable belief, while the use of first person plural pronouns is linked with ideas resistant to change.
The paper pinpoints a variety of manners in which individuals can succeed at persuading argument partners. These include the use of hyperlinks to substantiate statements which correlates with persuasion success. Interestingly, the researchers found that .com links seem to be more convincing than PDFs or .edu domains.
Beyond that, employing examples –– signified with introductory phrases like “for example” or “e.g.” –– is another lucrative method for formulating a powerful argument.
The paper’s authors also discovered that some much-used techniques are not as convincing as one may believe. The use of quotations to quote an argument partner, for one, was not found to be a persuasive approach. Nor, according to the data collected, are question marks advantageous to persuasion. While questions can be helpful in clarification, question marks are at best ineffective in terms of correlation with persuasive success.
Nevertheless, questions remain. In the paper, the authors recommend more research into the effects a debate experience can have on an individual’s persuasive success, better semantic models for studying online content and how sequences of connecting words (but, now, then, etc) add or detract from persuasiveness.
The paper may indeed give insight into composing victorious arguments, but an equally large takeaway is the ability to have calm, reasoned, and convincing arguments over the Internet.