Two Determinants of Dynamic Adaptive Learning for Magnitudes and Probabilities.

Area of Science:

• Cognitive Psychology
• Neuroscience
• Machine Learning

Background:

• Humans continuously update knowledge in a dynamic world.
• Adaptive learning rates are crucial but understudied, especially in probability learning.
• Distinguishing environmental changes from stochastic fluctuations is key for learning.

Purpose of the Study:

• To investigate dynamic learning rate adaptation in humans for both magnitude and probability learning.
• To identify the determinants driving these dynamic adjustments in different learning contexts.
• To compare human learning dynamics with normative theories.

Main Methods:

• Subjects provided real-time continuous reports of their learning rates during magnitude and probability learning tasks.
• Experimental design allowed for direct measurement of learning rate dynamics.
• Comparison of observed dynamics against predictions from normative models.

Main Results:

• Humans dynamically adapt their learning rates in both magnitude and probability learning.
• Following a change point, learning rates increase suddenly for magnitudes and gradually for probabilities.
• Change-point probability primarily drives magnitude learning dynamics, while prior uncertainty drives probability learning dynamics.

Conclusions:

• Human learning exhibits dynamic adaptation under uncertainty, aligning with normative theory.
• Magnitude and probability learning involve distinct adaptive mechanisms and driving factors.
• Findings inform research into the neural underpinnings of adaptive learning.