Helping Students Understand the Weirdness of AI

[april_lidinsky]

I am definitely learning alongside my students as we consider the place of Artificial Intelligence (AI) in the writing classroom. I’m not an early adopter — far from it. But I am certainly open to AI’s generative possibilities. I’ve enjoyed hearing how students use large language models such as ChatGPT in almost all areas of their lives. Some use it to figure out what to make from the weird collection of ingredients in their apartment fridges. Some use it to find less confrontational ways to phrase difficult messages to friends. My nephew used it to generate on-the-spot ceremonial language for his sister’s wedding. He’s a coder with a realistic sense of what AI does and doesn’t do. He made an assessment: Wedding ceremonies are largely formulaic, and using AI got him out of a tongue-tied spot so that he could embellish with some personal touches.

Writing in our classrooms, though, has different purposes. I’m learning a lot from instructors in the “Bits on Bots” series, including Jeanne Beatrix Law’s recent post about the energetic student engagement she has witnessed with some generative classroom uses of ChatGPT. In another post from that series, Jennifer Duncan proposes an annotated bibliography assignment that guides students into meta-analysis of AI research results that moves students’ thinking to higher levels in Bloom’s Taxonomy, and pushes them into deeper thinking about how and why to use sources in their papers. That certainly seems like using AI as a “tool, and not as not a crutch,” as her students observed.

And yet I’m also learning from AI insiders how to explain to students some of the serious limits of AI, including the crucial insight that the “I” in AI is not “intelligence.” A machine learning algorithm can predict the next words in a standardized sentence, but it is not grounded in reality, in culture, or in the network of creative meaning-making that humans bring to the task of interpretation and argument. Helping students understand this can guide their decisions about how and when to use AI.

So, I highly recommend bringing into your classroom the student-friendly, accessible and very funny blog by AI researcher Janelle Shane, AI Weirdness.  Shane also has a TED talk, “The Danger of AI is Weirder Than You Think” that could be 10 minutes well-spent as a conversation-starter in class or as a spark to reflective writing.

Shane consolidates her argument in these “Five Principles of AI Weirdness”:

            * The danger of AI is not that it’s too smart but that it’s not smart enough

            * AI has the approximate brainpower of a worm

            * AI does not really understand the problem you want it to solve

            * But: AI will do exactly what you tell it to. Or at least it will try its best.

            * And AI will take the path of least resistance. 

Shane illustrates these principles in ways that will likely catch your students’ interest. For example, she experiments with AI to generate pick-up lines, including, “You look like a thing and I love you,” which became the title of Shane’s 2019 book. On her blog, she collects hilarious examples of the ways AI cannot replicate human creativity and experience when it comes to designing Halloween costumes (“Spartan Gandalf,” anyone?) or ice cream flavors (care for some “praline cheddar swirl”?).

Keeping up with the value and the limits of AI might be one of the most important conversations we have this year — for our students, and for ourselves.