The Eighty Five Percent Rule
Psych 🧠 - 320/500
Reader,
Today’s research paper explores the concept of an optimal difficulty level for learning, proposing the "Eighty Five Percent Rule."
What is it?
It suggests that learning is most effective when the training accuracy is around 85%, meaning learners should make mistakes about 15% of the time.
The research focuses on binary classification tasks, where the goal is to categorize items into one of two groups (like distinguishing between 🐱 and 🐶).
This rule applies to various learning algorithms that use a method called stochastic gradient descent, which is found in both artificial intelligence (AI) and models of biological learning.
Read this far? You Deserve This Special Book of Psychology 💡
Claim The Psych Handbook featuring explanations of 150+ biases with emojis.
Get a Amazon Kindle copy from here.
What do I need to know:
Optimal error rate for learning: The study mathematically derives that the ideal error rate for learning is around 15.87%, or an 85% accuracy. This sweet spot maximizes the rate at which the learning algorithm adjusts its internal parameters to improve performance.
Learning speed at optimal accuracy: Learning at 85% accuracy is significantly faster than learning at a fixed difficulty level, especially for certain types of noise in the learning process.
Potential application to cognitive control: The mathematical analysis might extend to other brain functions like attention and cognitive control, which also influence the precision of neural representations.
Mathematical foundation for Flow: The research provides a possible mathematical explanation for the state of "Flow," where individuals feel optimally engaged and learn effectively when the challenge matches their skill level.
Key takeaways:
Training at the edge of competence is ideal: The Eighty Five Percent Rule formalizes the idea that learning is most effective when we are challenged but not overwhelmed.
Making mistakes is essential for learning: A certain amount of error is crucial for learning algorithms to adjust and improve.
Implications for teaching and training: Understanding this rule can help educators and AI researchers design more efficient learning environments and curricula.
great piece on why we should make mistakes!
I find that societal's need for win/lose, competitions and how media glorifies the winners don't help us learn the importance of making mistakes.
Too many times, this brings out the desire for vanity, jealousy, greed, the 7 sins in us, preventing this stage of learning.