Emoji characters may seem trivial, but they are fast becoming a foundation for online communication. Ten billion were sent on Twitter alone last year, and roughly half of all Instagram posts include them.
Researchers are seeing signs of an early emoji “language” emerging as the symbols (there are now 1,282 and counting) grow more complex and nuanced, eclipsing even Egyptian hieroglyphics.
But as digital interactions start to displace more face-to-face communications, academics are now confirming what many social media enthusiasts have complained about for years: Emotional communication online is a mess.
In a new study, researchers at the University of Minnesota uncovered major discrepancies in the way people interpret everyday symbols to communicate basic emotions. They found that people often disagreed as to whether the same emoji represented a positive, neutral, or negative sentiment—and the confusion increased when versions of the same emojis rendered by different companies were presented.
For example, the “grinning face with smiling eyes” emoji, a symbol that appears differently on social media and smartphones by companies such as Apple, Samsung, and Twitter, has no less than 17 different appearances. Here’s a sampling:
The difference in meaning construed by the subjects was widest for Apple’s teeth-gritting version, which actually elicited strong negative reactions by most people. One Twitter sentiment study (pdf) by Stanford University even grouped it under the emojis for “scared.”
It’s a small sample size, yes, but the fact that there was a sample at all speaks to the growing academic interest in the power of emoji. Although the field was almost non-existent 10 years ago, more than 400 papers dealing with emoji appeared last year, compared to just one English publication in 2005, according to Google Scholar. The University of Minnesota study (pdf) is scheduled to be presented in May at the Association for the Advancement of Artificial Intelligence 2016 conference on web and social media.
Emoji are so popular in part because humans are terrible at interpreting emotions in written text. A 2005 University of Illinois study found that people interpreting written text could accurately identify emotions like seriousness or sarcasm just 56% of the time, hardly better than chance. The ratio rose to 73% when the same words were delivered as voice recordings.
Compounding the challenge of gleaning meaning from text, according to researchers, is that people tend to chronically overestimate their ability to interpret others’ intent (pdf, p.8), and are prone to needlessly negative interpretations of written text, especially in the absence of explicit emotional cues. This makes sending an emotionally complex text message a perilous exercise at best, as most couples can attest.
Typographers have long experimented with punctuation marks to represent emotions, without much popular success. In the 1900s, Morse code used numbers to transmit emotions such as “love and kisses” (88), “best regards” (73) and “lots of success” (55). But the real start of online emoticons, a combination of the word emotion and icon, started with a 1982 message board proposal to use ” :-)” to denote jokes.
Their use quickly spread across ARPAnet, the precursor to the internet, and then to Japanese mobile phone carriers, who coined the word emoji. It all led up to the creation of the nonprofit Unicode Consortium, which sets the common standard we use today and aims to add roughly 60 new emojis per year.