Researchers are now trying to solve that, and to do it they're using the internet's sarcasm honeypot - Twitter.
A research paper titled 'Contextualized Sarcasm Detection on Twitter' details how researched pulled together tweets that included a #sarcasm hashtag. The system then analysed those tweets, using words like "clearly", "gasp", "I'm shocked" as indicators as to whether the tweets were indeed sarcastic.
They also included wider contextual cues including the identity of the author and the topic they were tweeting about. 85% of the time, the system correctly identified if a post was sarcastic.
Oh yeah, sure, totally
The results showed that including wider context about the author and conversation was more effective in detecting sarcasm than using the tweet alone. Further, the #sarcasm hashtag was actually found to not be a natural indicator of sarcasm being expressed.
"This has important consequences for the study of sarcasm and other speech acts on social media with complex audiences," the paper concludes.
The findings could help artificial intelligence become better at differentiating between statements that are true and false, among other applications.