During December and January, Geek Feminism is republishing some of our 2012 posts for the benefit of new and existing readers. This post originally appeared on March 19, 2012.
I work at a startup, and most of the time, I enjoy it. Compared to most tech companies, and certainly most startups, we have quite a few people who are relatively clueful. There are relatively few moments of “brogrammer” culture. There is, however, one thing that has been bugging me for months, ever since it was introduced.
I took it for granted that everyone was familiar with the “That’s what she said,” joke, but a recent conversation with a consultant friend made me realize some industries don’t feature it on a daily basis. For those who haven’t heard it a million times, the idea is that when somebody says something that could remotely be turned into a sexual joke, e.g. “I’m trying to solve this problem but it’s really hard!” you say “That’s what SHE said,” in a lascivious tone.
Now, I admit to having made this joke myself, at times. Once in a while, I even find it funny. What I don’t find funny is a bot we have in our general IRC channel at work, that has some basic AI devoted to determining when to interject TWSS into the conversation.
I asked a number of times to have that bot function turned off, but got the usual combination of being ignored, being told it’s funny, and being told I should lighten up. I tried explaining once why it was objectionable, and managed to get the guys saying variations, e.g. “That’s what your DAD said,” for an evening, but that was about it.
Last Friday, the bot went a bit crazy and started throwing TWSS into the conversation with no apparent rhyme or reason. Finally, I had had enough. And then it came to me: I would write my OWN bot, that responded to TWSS with a quotation from a notable woman. If they are so keen on what she said, why don’t we get educated about what she really had to say. And so the “whatshereallysaid” bot was born. It might annoy the guys into shutting off the TWSS bot, or we might all learn about notable women. It’s a win either way, in my books!
I’d never written a bot, but how hard could it be? Python is my primary programming language these days, so I started searching for Python IRC bots. I tried a few different libraries before setting on twisted. I found a very nice example of a logging bot using twisted , and went from there.
Step 1: I copied the LogBot code into local files, and tried running it. Since the IRC channels I am normally in use SSL, I had to set up my own channel on a public IRC server for testing. It took a while to get the settings right, but finally I succeeded!
Step 2: I created a quotes file and put in a couple of my favorite quotes to test with. I created class to randomly select a quote from the file. I modified the log bot so that if you said its name, it would get a quote and post it to the channel.
Step 3: My ultimate goal was to put the bot in the work channel, so I needed it to support SSL. I found a stackoverflow thread about using a different connect function, one for SSL. Again, it took me some time to get the settings right, but eventually I got the bot to connect to the work IRC server.
Step 4: Now I needed more quotes! I found an excellent website of quotes by women and wrote a small script to scrape the quotes. I used the excellent python requests library. The script got the content of the main links page, and iterated over the links, using a regular expression to get the name and then following the link to the quotes page. Then it pulled the quotes out of the page content and stored them in a text file, one per line, with the woman’s name attached.
Step 5: I set up my own private channel and had my bot connect. I tested it to verify it was correctly noticing and responding to trigger phrases. Once I was confident of that, I tried in the general chat channel. Success!
Step 6: I cleaned up the code, removing everything unrelated to posting quotes, and adding a settings file. I made the trigger phrases configurable. I also added a virtualenv. I felt the project was now in good enough shape to post publicly, so I shared it on github.
Step 7: At this point, I was still running the bot on my own computer, which was clearly non-ideal. I wanted it to be in the channel all the time, not just when I was logged in! I found a python daemon implementation and made my bot run through that. Perfect! Now I could run it from any VPS rented from a big server provider like Hostgator (and save a few bucks on their HostGator Black Friday Deal) or on one of the development servers at work.
Step 8: Unit tests, of course! No project is complete without them. I only have tests on the quote selector so far, but I am going to set up some mock tests on the bot logic.
Originally, I had called the project “whatshereallysaid”, but as I was working, I realized this could be much more broadly applied. Anyone could clone my github project and configure a bot to respond to whatever annoying comments they had to put up with. Inspired by “take back the night” and “hollaback”, I decided to rename the project “talkbackbot“.
I finished all this rather late on Friday night, so there was nobody around. I waited with anticipation to see what the response would be. To my surprise, many coworkers complimented me on creating the bot, including some I would never have guessed would appreciate it. Some people even retweeted my announcement that the bot was available on github.
It has been fascinating to watch the ongoing reactions. There have been complaints that we have too many bots in the channel now. There have been complaints about it spamming the channel. There were several “Make them shut up!” responses. These are not reactions I have seen the other bots elicit, certainly not with such intensity. One person even complained about the name being too long, though to his credit he realized right after he said that that several other people in the channel also have very long handles.
To me, all of this seems like typical geek behaviour: something is making them uncomfortable, and so they attack it on “rational” grounds. Most likely, they aren’t even aware of the gut reaction fueling their logic. Interestingly, the intensity of emotion seemed to carry over into subsequent discussions, including one about women in the Python community. For the most part, I have not responded to the comments. I did shorten the bot’s name to “twsrs”, and I pointed out that it’s trivially easy to have the bot not say anything: don’t say TWSS.
In any case, I feel that I have succeeded in constructively disrupting an aspect of my work culture that made me uncomfortable. This is the first personal project I’ve ever thought of, coded, and made public, and I am pretty excited about it! It makes me so happy every time the other bot says “that’s what she said,” and my bot responds with something like:
Our struggle today is not to have a female Einstein get appointed as an assistant professor. It is for a woman schlemiel to get as quickly promoted as a male schlemiel. ~ Bella Abzug
One of the most hilarious responses was a guy saying he was going to say TWSS a lot more, so he could get more quotes! I decided to add another trigger phrase just for him, since the last thing I want to do is encourage *more* TWSS!