Related Experiment Video
Updated: Jan 15, 2026

Eye Tracking During Visually Situated Language Comprehension: Flexibility and Limitations in Uncovering Visual Context Effects
Published on: November 30, 2018
Curriculum effects in multitask learning through the lens of contextual inference
Sabyasachi Shivkumar1, Máté Lengyel2, Daniel M Wolpert3
1Zuckerman Mind Brain Behavior Institute, Department of Neuroscience, Kavli Institute for Brain Science, Columbia University, New York, USA.
Practice structure significantly impacts learning across motor, rule, perceptual, and machine learning tasks. Contextual inference offers a unified framework to explain trade-offs between learning speed and retention, potentially mitigating catastrophic forgetting in AI.
Area of Science:
- Cognitive Science
- Machine Learning
- Neuroscience
Background:
- The structure of practice, or curriculum, is crucial for multitask learning outcomes.
- A trade-off often exists between rapid learning acquisition and long-term retention in multitask settings.
- Blocked training can enhance acquisition but reduce retention (e.g., motor learning), while interleaved training may improve retention but slow acquisition.
Purpose of the Study:
- To propose a unifying framework, contextual inference, to explain the divergent effects of practice structures on learning across different domains.
- To investigate how task transition dynamics, contextual cues, and observation noise influence learning outcomes.
- To explore how principles from biological learning can inform machine learning to mitigate catastrophic interference.
Main Methods:
- The study proposes a theoretical framework, contextual inference, integrating task transition dynamics, contextual cues, and observation noise.
- It draws parallels between findings in motor learning, perceptual learning, cognitive learning, and machine learning (catastrophic forgetting).
- The framework aims to explain why blocked training benefits some learning types while harming others.
Main Results:
- Blocked training can lead to faster acquisition but poorer long-term retention in domains like motor learning.
- Perceptual and cognitive learning often benefit from structured, blocked training curricula.
- Machine learning exhibits 'catastrophic forgetting,' a phenomenon analogous to reduced retention in blocked motor learning.
Conclusions:
- Contextual inference provides a unified explanation for the impact of practice structure on learning across diverse domains.
- Understanding these principles can inspire methods to reduce catastrophic interference in machine learning.
- Leveraging insights from biological learning may enhance the robustness and efficiency of artificial intelligence systems.
Related Concept Videos
Purposive Learning
Associative Learning
Classical conditioning, also known...
Cognitive Learning
E. C. Tolman's theory of purposive behavior emphasizes that much behavior is goal-directed. He argued that to understand behavior, we must look at the entire sequence of actions leading to a goal. For instance, high school students study hard, not just due to past reinforcement but also to achieve the goal of getting into a good college.
Tolman introduced the idea that behavior is influenced by...
Observational Learning
Impact of Social Context on Individuals
Multi-input and Multi-variable systems
In the absence of...

