From 3M Health Information Systems
AI Talk: The role of humor
I am a member of the Association for Computing Machinery (ACM) and I get all kinds of emails from them. Recently, I received two back-to-back emails with a curious theme: The role of humor in the design of intelligent systems! Let’s examine what is happening in the AI landscape with respect to humor.
The first email was about a curious competition titled “HAHA” which stands for “Human Analysis on Human Annotation.” This competition is part of an effort to study the Iberian language, an ancient language with origins in Spain. The competition explores whether humor can be detected and classified in Spanish language tweets. Which begs the question, what about English joke detection? Turns out there is a lot of research on this front as well. Here is a recent research article which analyzes tweets for sarcasm, irony and other humor. Who knew?!
Teachable Humorous Assistant
The second ACM email referenced a research paper on the role of humor in a virtual assistant, but with a twist. The research was conducted by a multi-disciplinary team of researchers at Waterloo University. The virtual agent is the learner and the teachers are 58 participants, ages 18 to 35. The task? How to classify rocks as metamorphic, sedimentary or igneous. The research basically explores if humor can serve as an effective medium for enhancing learning using teachable agents. The participants attempt to teach a virtual agent and in the process learn themselves. I like that concept very much, as that is precisely what I do! If I am not familiar with a subject, I offer to teach that subject to some poor souls (I usually get away with it too, as I have a “professor” moniker in my name).
The goal of the research is not just to facilitate learning by teaching, but to figure out what role humor plays in this process. So, how does this work? The team of researchers hired an army of writers to come up with jokes about rock classification. As I was researching this article, I found out that there are actually four humor styles: affiliative, self-enhancing, self-defeating and aggressive. The research team focused on the affiliative and self-deprecating types in their experiment. They also created an avatar named “Sigma” who served as the trainable agent. Here is an example of an affiliative joke by Sigma: “What’s a rock’s worst enemy? Paper, haha!” and a self-deprecating one: “You know that feeling when you’re taught something and understand it right away?… Yeah, not me! Haha!” The teachable agent is implemented as a text-based chatbot. So, what were the results of the study? The affiliative-humor style works best, but the humor style has to be matched to the participant—a one-size-fits-all approach does not work.
There are a lot of fun bots out there. Some will tell you jokes. If you want an inspirational quote, there is bot for that. There is even one that will give you Bollywood entertainment! This website lists a whole bunch of bots. Many of these bots are heavily programmed and that’s why Amazon has the Alexa challenge (the fourth in this series) underway now. The goal of this challenge is to build a socialbot that is able to engage a user in 20 minutes of conversation on a range of topics! That will require significant advances in current bot technology.
Voice assistants and bots will be present in pretty much every aspect of our lives in a few years. They are already in our cars, in our homes, always ready to answer any random question. To me, this trend to explore humor and empathy in bots is a good thing. We could all use a helping hand from these bots and a smile while we’re at it.
I am always looking for feedback and if you would like me to cover a story, please let me know! Leave me a comment below or ask a question on my blogger profile page.
V. “Juggy” Jagannathan, PhD, is Director of Research for 3M M*Modal and is an AI Evangelist with four decades of experience in AI and Computer Science research.