West Michigan program tests humans and AI in predicting snow days

Artificial intelligence now can predict your next snow day. A program at one west Michigan school is putting the technology to the test.

The contest, Humans vs. The Machine, teaches math students about AI with real-world experience. Rockford High School AP statistics students test the likelihood of snow days in their district, comparing the effectiveness of human and AI predictions.

"This is just now entering into their world, and this is going to give them an introduction to what AI is, what AI is capable of and I think it will get them thinking into what else can we do with AI and how can we use it in a productive way," said Tina Shutich, Rockford High School AP statistics teacher.

The contest comes from a partnership between Rockford Public Schools, the Van Andel Institute for Education, a science academy in Grand Rapids, and one local parent's technology skills.

"Kids are always interested in snow days, it's a great opportunity to get a little math in there," said Ben Talsma, learning solutions specialist for the Van Andel Institute.

The institute then contacted district officials, who connected them with AP statistics teachers in late fall 2023. Throughout the contest this winter, students analyze the likelihood of snow days during winter weather advisory conditions.

The program uses two models, a human-based model and an AI-based model.

The human model was crafted by a friend of Talsma's and a former teacher years back, while the AI-based model comes from Steven Wangler, a software engineer and parent in the Rockford school district. Talsma learned of Wangler's model through local news reporting and reached out to collaborate.

The human model uses a crowdsourced method of statistical analysis to predict the likelihood of snow days based on the averages of four statistics categories.

The AI model, known as Blizzard, uses ChatGPT software to analyze the likelihood of snow days based on weather data and past snow days in the district.

Students and teachers had little prior exposure to AI but welcomed the opportunity to learn more about the technology and grow their skills utilizing it.

The move comes amidst the growing influence of AI technology in U.S. society, including in the education sphere. Across Michigan, companies and institutions have begun considering introducing AI across industries, including automobile manufacturing, craft beer brewing, advertising, sports, and more.

While the district doesn't plan to determine snow days based on the contest, it provides a valuable learning experience for students, many of whom will later use AI in their careers. Community members, including parents and former students, have reached out with interest after seeing the initiative on social media, Shutich said.

AI-based snow day prediction models are already in effect in the U.S., with online platforms and applications promoting the technology model. Platforms such as snowdaypredictor.com and willihaveasnowday.com operate a calculation method based on inputting the relevant ZIP code.

The students' initial excitement over potential snow days gradually grew into a genuine interest in exploring the possibilities of AI, Shutich said.

"We even had the discussion of later on, after we get through this season and maybe even during this season, how my students can look at the data that we have and how can we improve upon the model that we're using," Shutich said.

Under Humans vs. The Machine, a group of about 130 participating students in statistics classes works under the guidance of two teachers. Thus far, the program has featured four contests, as students narrow down the methodology to improve predictions and better understand the models involved.

The classes first ran the program before the school's winter break to test the process and again amidst current winter weather advisory conditions in the state.

Findings show contest results so far, comparing the two models on a scale of up to 100% accuracy in predictions. The above chart demonstrates how the models closed the gap over time, with the human model becoming closer to and then surpassing the AI model.

The above bar chart compares the accuracy rates of the two models in predicting snow days, demonstrating similar rates of up to 100% possible accuracy.

This program marks a trial run for the effectiveness of AI in predicting snow days, Talsma said. Community reactions range from fear to excitement, part of broader mixed sentiments on the rise of AI globally.

"It demonstrates the degree to which we're still exploring and feeling out the ways in which artificial intelligence can enhance our work and our abilities and the ways in which human cognition, human thinking still is required or necessary," Talsma said.

The AP statistics classes are planning to have Wangler come speak to students about his AI model soon, Shutich said. Program leaders keep in touch through a text chain that includes Wangler.

"This is an opportunity for students to take what they've learned and have a little fun with it, but then also see the use of statistics and see the use of artificial intelligence," Shutich said.

The program will run through the winter at Rockford High School and provides updates online. The Van Andel Institute hopes to expand the initiative, encouraging other schools to test the AI technology.

This article originally appeared on Detroit Free Press: Contest compares humans and AI snow day prediction skills