Humor is an inherently social phenomenon, with humorous utterances shaped by what is socially and culturally accepted. Understanding humor is an important NLP challenge, with many applications to human-computer interactions. In this work we explore humor in the context of Cards Against Humanity -- a party game where players complete fill-in-the-blank statements using cards that can be offensive or politically incorrect. We introduce a novel dataset of 300,000 online games of Cards Against Humanity, including 785K unique jokes, analyze it and provide insights. We trained machine learning models to predict the winning joke per game, achieving performance twice as good (20\%) as random, even without any user information. On the more difficult task of judging novel cards, we see the models' ability to generalize is moderate. Interestingly, we find that our models are primarily focused on punchline card, with the context having little impact. Analyzing feature importance, we observe that short, crude, juvenile punchlines tend to win.

The creators of Cards Against Humanity are back for their annual Black Friday stunt, and this one is delightfully dystopian. Starting at 11AM ET today and lasting for the next 16 hours, the human writers on the CAH team are facing off against an artificial intelligence to see who can create the most popular new pack of cards, based on how many people pay for more $5 packs. You can upvote or downvote your favorite cards for each side on CAH's website before buying, and you can also watch the humans struggle to come up with new iterations in real time over live stream. On the line are $5,000 bonuses for every employee if team human comes up victorious, or heartless termination in the event the AI takes the top spot. We don't think CAH actually plans to fire their writers if they lose, but it is a clever stunt nonetheless to drum up the human vs. machine narrative at a time when automation may pose a very real threat to millions of jobs in the coming decade, writing included.