Hot Garbage AI and Education
Published: 2024-03-09 5:09 PM
Category: AI | Tags: artificial intelligence, llm, education, feedback, comment
This week, Axios published an article about teachers turning to AI to give students feedback and then went on to extol the benefits of a new AI company purchased by Houghton-Mifflin Harcourt to push it into schools.
This is the hottest of all garbage.
Driving the news: Writable, which is billed as a time-saving tool for teachers, was purchased last month by education giant Houghton Mifflin Harcourt, whose materials are used in 90% of K-12 schools.
Teachers use it to run students' essays through ChatGPT, then evaluate the AI-generated feedback and return it to the students. "We have a lot of teachers who are using the program and are very excited about it," Jack Lynch, CEO of Houghton Mifflin Harcourt, tells Axios.
This is the worst case of AI in education that I've seen. The article spins it with the fact that "...teachers are already using ChatGPT and other generative AI programs to craft lesson plans," and that "diligent teachers will probably use ChatGPT's suggestions as a starting point."
I'm glad diligent people will probably read the feedback. HMH calls this "human in the middle AI."
Time, of course is the great PR payoff. Teachers are so low on time that the only solution mega-corporations with lobbying influence and political sway can come up with is to give me time to get to know my students by not reading their writing, you know, instead of working on behalf of teachers to reduce class sizes, improve working conditions, or limit the impact of high-stakes testing1 several times a year. All things that are shown to improve student learning outcomes and teacher effectiveness.
But don't worry, Lynch says that the goal is "to empower teachers, to give them time back to reallocate to higher-impact teaching and learning activities."
Unfortunately, Mr. Lynch doesn't realize that feedback is one of the most impactful teaching and learning activities I can engage students with. My time is not occupied with endless amounts of grading, which is what they're suggesting we use an LLM to do. It's spent on getting to know my students' thinking brains and how they interpret and interact with the world. Feedback is my way to engage - one on one - with a student, whether it is verbal in the moment or written on a submission.
An LLM is only able to give feedback on the combination of words a student produces - not the thought that went into those words or in the originality of ideas expressed by the words. If a student expresses something novel, the LLM is not going to be able to recognize the skills that went into the creation of the work. Relying on a model of any kind to give feedback tells our students that we don't care about originality of thought, just that they can regurgitate with full sentences.
As bad as Jack Lynch's take on AI for feedback as, Simon Allen, CEO of McGraw Hill follows up with a real "hold my beer" moment:
"The actual process of grading, we have simplified significantly," ... "You're not going to physically hand-grade every single essay or multiple choice activity. You're going to utilize the technology we've given you."
For as much as they want to help teachers out, they're really showing a ton of confidence in us.
- Accessed from Arizona State University Mary Lou Foulton Teachers College Education Policy and Analysis Archives.
Comments