Previously a Visiting Scholar at Kellogg
Sandy & Morton Goldman Professor of Entrepreneurial Studies in Marketing; Professor of Marketing; Co-chair of Faculty Research
Professor of Management & Organizations
Let’s say you are writing an online review of a pair of socks. You might describe the socks as “warm” and “well-made.” But what if you had more at stake in the review? What if, say, the manufacturer offered a free pair of socks if you persuaded someone else to buy a pair? Would you write that your socks were “awesome” and “amazing”?
Surprisingly, academics know relatively little about how the language people use changes when they are trying to be persuasive.
There is a “theoretical hole in the literature,” says Loran Nordgren, an associate professor of management and organizations at Kellogg. “We know very little about how people naturally try to persuade others.”
So Nordgren and two Kellogg colleagues, marketing professor Derek Rucker and postdoctoral researcher Matthew Rocklage, tackled the question. They found that participants’ language tends to become more emotional when they are motivated to persuade others to buy a product.
This pattern continues even when people are told that their audience is very rational and thus unlikely to find emotional arguments persuasive.
“The gravity of emotional language is hard to escape, even when you know the audience wants to hear a more cognitive appeal,” Rucker says.
The results suggest a potentially interesting quandary. When given incentives to write positive, persuasive reviews of products, people may use more emotional appeals even when emotional appeals are not optimal. For some categories of products, such as toasters, users may mostly want to hear rational arguments—after all, when is the last time you got giddy about heating up bread.
If a review calls a toaster “amazing,” as opposed to “perfect,” “you might actually be less persuaded by the message,” Rucker says. “The emotion could get in the way.”
The researchers used a tool that Rocklage developed in previous research. The Evaluative Lexicon (EL) rates words on three features: valence, or whether the word is positive or negative; extremity; and emotionality. For example, “perfect” is more extreme than “wonderful,” but “wonderful” is more emotional than “perfect.”
“We observe that people express more emotion even in categories where that might be odd and potentially less effective.”—Derek Rucker
To study how people use language in persuasion, Rocklage, Rucker, and Nordgren first recruited 1,285 people on Amazon Mechanical Turk to participate in an online experiment. The researchers chose 20 products that had received many positive reviews on Amazon, in categories such as appliances, toys, and electronics.
Each person read a neutral description of one product and was asked to write a 5-star review. But one group was told to describe the product’s positive traits, while another group was instructed to persuade other consumers to buy it. The researchers then analyzed the words in the reviews, as well as those in 840 real Amazon reviews of those products.
People who had been told to simply write a positive review used words with an average emotionality rating of 5.38 out of 9 on the EL scale, the team found. Levels were similar in the real Amazon reviews. But participants who were told to persuade others exhibited a slight shift in their language such that they expressed greater emotion, with an average EL rating of 5.48 out of 9. For example, one person in the first group wrote that a printer was “useful,” while someone in the persuasive group called it “delightful.”
“We observe that people express more emotion even in categories where that might be odd and potentially less effective,” Rucker says.
Rucker acknowledges that the difference is small, but notes that the goal of the study was to see if the relationship exists. “In early research we often anticipate effect sizes will be small,” he said. “However, now that there is some evidence for the link, future research can explore if and when those small effects translate into meaningful differences in the real world.”
The team also explored whether this tendency was a result of conscious effort or whether it happened more automatically. How could they provide insight into this issue?
They reasoned that if the use of emotional arguments was due to deliberate reflection, it should decrease when people’s minds were more taxed.
To test this, they asked 288 online participants to write several positive words to describe a novel they liked. Some participants were told that they would earn $1 every time a participant in a future study picked that book based on their description; others were not given such an incentive. This was used to influence people’s motivation to be persuasive.
Everyone also had to memorize a code and recall it later. Some got a two-digit code, others an eight-digit code. The researchers posited that if the link between persuasion and emotion was more automatic, then the emotional levels in the language shouldn’t differ much between the two groups.
Indeed, the length of the code did not make a difference, suggesting that people were changing their language without much conscious effort.
“That was our first piece of evidence that this is a more automatic association,” Rucker says.
Another experiment offered additional evidence of an automatic association. The researchers showed people various words and asked them to press a key if that word was linked to persuasion. When the word was more emotional, participants tended to press the key more quickly, which suggests a natural association between emotion and persuasion.
But what if people knew their audience was highly rational and unlikely to be swayed by emotion? Could they turn off their automatic tendency to use emotion and employ more cognitive arguments instead?
“We know very little about how people naturally try to persuade others.” —Loran Nordgren
The researchers asked 781 people online to write down positive descriptions of a restaurant where they had recently dined. To incentivize people to persuade, one group learned they would earn $1 for each future study participant who picked that restaurant.
Then some participants were told more about whom they were trying to convince. In a not-so-subtle hint, one group’s audience was described as people in the arts who called themselves “The Emotionalists.” For another group, the audience was “The Society for Applied Rationality and Mathematics,” which included scientists and analysts. A third group was not told whom they were persuading.
Participants who were incentivized to be persuasive, but did not receive details about the audience, used words with an average emotionality of 4.17.
In contrast, people who were targeting The Emotionalists boosted their emotional language to an average of 4.26. They might have felt that, given the nature of their audience, they had license to shift even more toward emotion, Rocklage speculates.
But participants addressing the rationality society did not tone down their emotion; it remained around 4.18. The results suggest that people stick with emotional arguments even when it should be apparent that they are less likely to be optimal.
“I’m now telling you that the secret sauce to persuade them is to use rational arguments,” Rucker says. “You should skew your appeals to be more cognitive, but that does not appear to come naturally to people. It goes against these hard-wired associations.”
Many firms now use sentiment analysis to glean information from social-media messages. These findings indicate that rather than just looking for positive or negative words, assessing the text’s emotional level could be useful as well, Rucker says.
For example, let’s say a company is planning to offer new products to bloggers to try for free and review. A blogger who generally uses more emotional language on social media might be more intent on persuading an audience. So it might be better for the firm to pick that blogger instead of an equally positive but less emotional writer, Rucker says.
On the other hand, companies need to understand whether their customers are more likely to be persuaded by emotion or reason. If the latter, then giving customers incentives to write positive reviews could be counterproductive if it means they start boosting their level of emotion.
“They become less effective agents of persuasion, even if they genuinely enjoyed the product,” Rucker says.
More research is needed to determine whether the patterns seen in the online experiments translate to the real world. The researchers would like to work with a company to investigate how incentives for persuasion affect language and, in turn, sales, Rucker says.
He suspects that emotion starts to become linked to persuasion in childhood. After all, kids frequently use emotion to make their case.
When it comes to influence, “emotion is arguably the earliest form of communication we have,” he says.