Plasticity as the Mirror of Empowerment
David Abel, Michael Bowling, André Barreto, Will Dabney, Shi Dong, Steven Hansen, Anna Harutyunyan, Khimya Khetarpal, Clare Lyle, Razvan Pascanu, Georgios Piliouras, Doina Precup
Abstract
Agents are minimally entities that are influenced by their past observations and act to influence future observations. This latter capacity is captured by empowerment, which has served as a vital framing concept across artificial intelligence and cognitive science. This former capacity, however, is equally foundational: In what ways, and to what extent, can an agent be influenced by what it observes? In this paper, we ground this concept in a universal agent-centric measure that we refer to as plasticity, and reveal a fundamental connection to empowerment. Following a set of desiderata on a suitable definition, we define plasticity using a new information-theoretic quantity we call the generalized directed information. We show that this new quantity strictly generalizes the directed information introduced by Massey (1990) while preserving all of its desirable properties. Under this definition, we find that plasticity is well thought of as the mirror of empowerment: The two concepts are defined using the same measure, with only the direction of influence reversed. Our main result establishes a tension between the plasticity and empowerment of an agent, suggesting that agent design needs to be mindful of both characteristics. We explore the implications of these findings, and suggest that plasticity, empowerment, and their relationship are essential to understanding agency
Ask about this paper
Your agent reads all of it.
Lune indexed this paper to the last equation, along with the top-tier papers that cite it. Ask a question and the answer quotes them.
Your agent calls
Luneget_paper_fulltext
Free to start. No credit card required.
Terminal
Install the CLIlune papers fulltext 614f3f7c-bfe2-4a3b-8e48-f1c8eac6b617Cited by top-tier papers2
- Training LLM Agents to Empower HumansEvan Ellis, Vivek Myers, Jens Tuyls, Sergey Levine et al.ICML 2026 · 4 citations
- Compositional Behavioral Semantics for State Abstraction in Reinforcement LearningYivan Zhang, Ziyan Luo, Manuel BaltieriICML 2026
Builds on8
- Explore, Discover and Learn: Unsupervised Discovery of State-Covering SkillsVictor Campos, Alexander Trott, Caiming Xiong, Richard Socher et al.ICML 2020 · 178 citations
- A Definition of Continual Reinforcement LearningDavid Abel, André Barreto, Benjamin Van Roy, Doina Precup et al.NeurIPS 2023 · 167 citations
- Understanding Plasticity in Neural NetworksClare Lyle, Zeyu Zheng, Evgenii Nikishin, Bernardo Ávila Pires et al.ICML 2023 · 162 citations
- Optimal Policies Tend To Seek PowerAlexander Matt Turner, Logan Smith, Rohin Shah, Andrew Critch et al.NeurIPS 2021 · 111 citations
- Settling the Reward HypothesisMichael Bowling, John D. Martin, David Abel, Will DabneyICML 2023 · 47 citations
Related papers
- Estimating the Empowerment of Language Model AgentsJinyeop Song, Jeff Gore, Max Kleiman-WeinerICML 2026
- Learning to Assist Humans without Inferring RewardsVivek Myers, Evan Ellis, Sergey Levine, Benjamin Eysenbach et al.NeurIPS 2024 · 16 citations
- Learning to Perceive the World Through Control: Empowerment-Based Representation LearningMahsa Bastankhah, Sophie Broderick, Benjamin EysenbachICML 2026
- AvE: Assistance via EmpowermentYuqing Du, Stas Tiomkin, Emre Kiciman, Daniel Polani et al.NeurIPS 2020 · 51 citations
- Measuring Goal-DirectednessMatt MacDermott, James Fox, Francesco Belardinelli, Tom EverittNeurIPS 2024 · 9 citations
