Towards a Formal Theory of the Need for Competence via Computational Intrinsic Motivation
Using computational intrinsic motivation, this study models the need for competence via reinforcement learning, revealing implicit assumptions in Self-Determination Theory.
Key Findings
Methodology
The study employs computational intrinsic motivation models from reinforcement learning to model the need for competence in Self-Determination Theory (SDT). By aligning different facets of competence (effectance, skill use, task performance, capacity growth) with corresponding computational models (e.g., RIDE, VIC, DIAYN), the study reveals how these models elucidate implicit assumptions in SDT.
Key Results
- RIDE model rewards agents for environmental changes they cause, modeling effectance motivation.
- VIC model demonstrates skill use motivation through state-conditioned skill selection.
- CURIOUS system optimizes task difficulty through automatic curriculum learning, illustrating task performance motivation.
Significance
This research bridges psychological theories of competence with computational models, revealing implicit assumptions in SDT and advancing motivational psychology. By providing explicit computational models, it lays the groundwork for future empirical studies.
Technical Contribution
The paper introduces a novel approach by integrating computational intrinsic motivation with psychological theories, offering computational modeling of competence needs in SDT. By incorporating specific algorithms from reinforcement learning, it provides new tools and perspectives for psychological research.
Novelty
This is the first study to apply computational intrinsic motivation models from reinforcement learning to model competence needs in SDT, uncovering implicit assumptions within the theory.
Limitations
- The current models are primarily tested in simulated environments, lacking real-world application validation.
- Model complexity may limit scalability to large datasets.
Future Work
Future research could empirically validate the models' accuracy and explore their application across different psychological theories.
AI Executive Summary
Self-Determination Theory (SDT) is a pivotal framework in psychology for understanding motivation, yet its articulation of competence needs often lacks clear computational models. This study introduces a novel approach by integrating computational intrinsic motivation models from reinforcement learning to provide a new theoretical modeling method.
The researchers categorize competence needs in SDT into four aspects: effectance, skill use, task performance, and capacity growth, matching each with corresponding computational models like RIDE, VIC, and DIAYN. These models not only reveal implicit assumptions in SDT but also provide a foundation for future empirical research.
Despite its theoretical significance, the current models are primarily tested in simulated environments, lacking real-world application validation. Future research could empirically validate the models' accuracy and explore their application across different psychological theories.
Deep Analysis
Background
Self-Determination Theory (SDT) is a significant framework in psychology, emphasizing the role of intrinsic motivation in behavior. However, many theoretical assumptions in SDT lack clear computational models, limiting their application in empirical research. Recently, computational intrinsic motivation models have made significant advances in AI, offering new possibilities for modeling psychological theories.
Core Problem
The need for competence in SDT is a crucial concept, yet its computational modeling remains unclear. This limits researchers' ability to empirically validate and apply the theory. The challenge is to align different aspects of competence (e.g., effectance, skill use, task performance, capacity growth) with specific computational models.
Innovation
This study introduces computational intrinsic motivation models from reinforcement learning to provide explicit computational modeling for competence needs in SDT. By matching different competence facets with corresponding computational models, it reveals implicit assumptions in SDT and lays the groundwork for future empirical research.
Methodology
- �� Use RIDE model to reward agents for environmental changes, modeling effectance motivation.
- �� Use VIC model for state-conditioned skill selection, demonstrating skill use motivation.
- �� Use CURIOUS system to optimize task difficulty through automatic curriculum learning, illustrating task performance motivation.
Experiments
The experimental design employs various reinforcement learning models (e.g., RIDE, VIC, DIAYN) tested in simulated environments to evaluate their effectiveness in modeling different competence facets. By comparing the performance of different models, the study validates their effectiveness in revealing competence needs in SDT.
Results
Results show that the RIDE model effectively simulates effectance motivation, while the VIC model excels in skill use motivation. The CURIOUS system, by optimizing task difficulty, demonstrates the potential for task performance motivation.
Applications
The study's findings can be directly applied in empirical research in motivational psychology, providing researchers with explicit computational models. Additionally, these models can be used to develop educational and training systems to help individuals enhance their competence.
Limitations & Outlook
Despite its theoretical significance, the current models are primarily tested in simulated environments, lacking real-world application validation. Additionally, model complexity may limit scalability to large datasets. Future research could empirically validate the models' accuracy and explore their application across different psychological theories.
Plain Language Accessible to non-experts
Imagine a factory where workers need to continuously improve their skills to complete various tasks. Each worker has different competence needs, such as seeing the impact of their work on the production line (effectance), using their skills at the right time (skill use), performing well in tasks (task performance), and continuously enhancing their skills (capacity growth). The models in this paper act like a management system in the factory, helping workers identify these needs and motivating them to enhance their competence through rewards.
ELI14 Explained like you're 14
Imagine you're playing a game where you need to keep improving your skills to complete different levels. Each level has different challenges, like seeing how your actions affect the game world (effectance), using your skills at the right time (skill use), performing well in tasks (task performance), and continuously improving your skills (capacity growth). The models in this paper are like a guide in the game, helping you identify these needs and motivating you to enhance your competence through rewards.
Glossary
Self-Determination Theory
A psychological theory emphasizing intrinsic motivation's impact on behavior.
Used in the paper to explain different aspects of competence needs.
Computational Intrinsic Motivation
The process of simulating intrinsic motivation through computational models.
Used to model competence needs in SDT.
Reinforcement Learning
A machine learning method that trains agents through reward mechanisms.
Used to simulate different aspects of competence needs.
Effectance
The ability to observe one's actions affecting the environment.
Modeled using the RIDE model in the paper.
Skill Use
The ability to use one's skills at the right time.
Modeled using the VIC model in the paper.
Open Questions Unanswered questions from this research
- 1 How to validate these computational models' effectiveness in real-world applications?
- 2 What is the applicability of these models across different cultural contexts?
Applications
Immediate Applications
Educational Systems
Use computational models to help students identify and enhance their competence needs.
Psychological Research
Provide researchers with explicit computational models to support empirical studies.
Long-term Vision
Intelligent Training Systems
Develop intelligent systems that automatically identify and motivate individual competence needs.
Abstract
Computational modelling offers a powerful tool for formalising psychological theories, making them more transparent, testable, and applicable in digital contexts. Yet, the question often remains: how should one computationally model a theory? We provide a demonstration of how formalisms taken from artificial intelligence can offer a fertile starting point. Specifically, we focus on the "need for competence", postulated as a key basic psychological need within Self-Determination Theory (SDT) -- arguably the most influential framework for intrinsic motivation (IM) in psychology. Recent research has identified multiple distinct facets of competence in key SDT texts: effectance, skill use, task performance, and capacity growth. We draw on the computational IM literature in reinforcement learning to suggest that different existing formalisms may be appropriate for modelling these different facets. Using these formalisms, we reveal underlying preconditions that SDT fails to make explicit, demonstrating how computational models can improve our understanding of IM. More generally, our work can support a cycle of theory development by inspiring new computational models, which can then be tested empirically to refine the theory. Thus, we provide a foundation for advancing competence-related theory in SDT and motivational psychology more broadly.