by Dan Cox — Old Dominion University
March 23, 2015 – 12:43
If I was forced to describe many of the projects I’ve created over the years, “silly” might be the first word that came to mind. Often, when I start to chase an idea, it’s the premise that amuses me the most and pushes me toward trying to post it somewhere to share with others. If it makes me laugh is the first and sometimes only requirement for me to invest time into figuring out some algorithmic way to create more examples of it.
When I started work on my first Twitter bot, I was chasing that exact same pattern. At the time, I had become fixated on the seemingly formulaic descriptions of the #1155CC">Boxcar Children series of books. Joking with friends that the plots in the books often settled into a summary of a set of the characters performing some verbs on some others, I set about trying to emulate that style myself. Reviewing the descriptions of the 137 books as of 2014, I compiled a list of all the verbs, nouns, and common phrases that occurred in the descriptions to create an algorithm for writing them. However, after a few iterations of designing this new automated process, I ran into a problem. They weren’t funny to me anymore.
In trying to make the generated descriptions as realistic as possible, I bypassed what made them silly to me in the first place. The new versions no longer held the spark that made me laugh in considering if, for example, the dog Watch could go to the store or the youngest sibling, Benny, could drive a car. It was the simple juxtaposition of some character doing some random verb to another that delighted me in the first place. Thus, boxcar_ebook was created as a simple experiment in following the basic rules of Some Character Name + Some Singular Verb + Some Character Name repeated three times. It’s a working formula that produces tweets every 10 minutes that range anywhere from the plausible to the bizarre.
It was also the lesson I needed to learn about creating Twitter bots: simple is best. While trying to carry out complex tasks to create tweets can certainly make for more interesting content, often the best way to make a message intriguing is to simply combine a few words in an interesting way. Leveraging the power of the platform to create short, temporal messages is the key to unlocking simple, silly content. Something as basic as a handful of words can be equally funny and thought provoking at the same time.
In revisiting the simple is best lesson recently, I came across a new formula for creating tweets. By choosing random words on the ends of a simile, it was possible to interrupt the normal interpretation process, I discovered. Through having a program choose words at random using the Wordnik API, the very literal comparison between two truly dissimilar things could produce some weirdly profound tweets. All that needed was to follow the pattern named after the account itself: X is like a Y.
Just like for boxcar_ebook, success was found in building off the premise and maintaining honesty to what I found funny in the first place. Instead of trying to integrate complex functionality or methods, using only three words bookended by random choices could create something serious and silly at the same time. It really was as simple as that.