Improving on Theories of Learning with AI

This piece, by Onno Berkan, was published on 12/03/24. The original text, by Martijn Meeter, was published by Nature Computational Science on 05/24/24.

This Vrije University study explores how artificial intelligence can be used to discover the underlying patterns in how humans learn new skills. The researchers analyzed data from 6,000 people learning various cognitive skills, like mathematics, through an online learning platform, Luminosity.

The study employed a two-step approach using AI. First, they used a deep learning network to predict how well people would perform on their next exercises based on their previous performance. Then, they used another algorithm called symbolic regression to translate these predictions into mathematical formulas. These formulas were then compared to existing theories of learning.

One interesting discovery was that some skills followed a logarithmic learning pattern, meaning that progress happens quickly at first but gradually slows down. This makes intuitive sense. The researchers found that their AI-discovered formulas performed as well as or better than traditional learning theories.

However, there are some significant limitations to consider. First, while the algorithms found patterns that fit the data well, it is unclear if these patterns truly represent how humans actually learn. The researchers worked with simplified versions of the raw data, which might have made the patterns easier for the AI to detect than in real-world situations.

Additionally, there's uncertainty about whether these learning patterns apply to different learning situations. For example, how someone learns topography in a classroom might follow completely different patterns than those discovered in the Luminosity data.

Despite these limitations, the study represents an important step forward in using AI to understand human behavior. While we can't be certain whether there are universal laws governing how humans learn, this research suggests that AI might be able to help us discover such patterns if they do exist.

Want to submit a piece? Or trying to write a piece and struggling? Check out the guides here!

Thank you for reading. Reminder: Byte Sized is open to everyone! Feel free to submit your piece. Please read the guides first though.

All submissions to berkan@usc.edu with the header “Byte Sized Submission” in Word Doc format please. Thank you!