Like Modulus, I was fascinated with the question of what Twitterbots are for. I also had trouble making my bot functional (although I did have it posting snippets from a body of text by the end of the workshop)! I blame that on my own failure to grasp everything in Codecademy (…or finish it, to be honest), not Matt’s awesome teaching. That said, I’ve learned enough about code to perceive that making a Twitterbot is a relatively easy task for those with basic coding knowledge (or just a lot of patience). Constrained by our fluency with Python, our Twitterbots were pretty simple—we were tweeting from preexisting text files or responding to hashtags, primarily. We were spammers, albeit ones with a sense of humor.

I’d like to think more about how bot-generated spam text drives and circumscribes online communication. I recently read a pretty good book by Sarah Jeong called The Internet of Garbage. She explained that spam text is as old as the Internet; people would troll the early IRC chat rooms by pasting in endless lines from Monty Python’s Spam sketch. Since screens could only show a few lines of the chat at once, this would effectively shut down conversation. Since then, meaningless “garbage text” has proliferated all over the Internet, and any web company has to devote considerable resources to getting rid of it so they can provide a functional service.

I’m not sure exactly how Twitter polices bots, but it’s easy to see how a determined team of programmers could completely shut down conversation around a particular term by overwhelming Twitter with bot-generated responses. Spam also has a chilling effect on human conversations—e.g., I avoid mentioning Viagra or Xanax in emails because those words automatically get it relegated to the recipient’s spam folder. Right now, our anti-spambots are just as unsophisticated as our spambots. But I can completely understand why! Essentially, they fight spam by using very rudimentary strategies to conduct a reverse Turing test. This makes me wonder what a perfect anti-spambot might look like. Is it possible to use computation to tell whether text comes from a bot? For that matter, is it always possible for humans to tell? What do you guys think?