The DOI matches its Crossref record; 128 of the 128 references it cites were found in a registry or an index.
Predictive representations can link model-based reinforcement learning to model-free mechanisms
Open accessPublished version of a preprint
Abstract
Abstract Humans and animals are capable of evaluating actions by considering their long-run future rewards through a process described using model-based reinforcement learning (RL) algorithms. The mechanisms by which neural circuits perform the computations prescribed by model-based RL remain largely unknown; however, multiple lines of evidence suggest that neural circuits supporting model-based behavior are structurally homologous to and overlapping with those thought to carry out model-free temporal difference (TD) learning. Here, we lay out a family of approaches by which model-based computation may be built upon a core of TD learning. The foundation of this framework is the successor representation, a predictive state representation that, when combined with TD learning of value predictions, can produce a subset of the behaviors associated with model-based learning, while requiring less decision-time computation than dynamic programming. Using simulations, we delineate the precise behavioral capabilities enabled by evaluating actions using this approach, and compare them to those demonstrated by biological organisms. We then introduce two new algorithms that build upon the successor representation while progressively mitigating its limitations. Because this framework can account for the full range of observed putatively model-based behaviors while still utilizing a core TD framework, we suggest that it represents a neurally plausible family of mechanisms for model-based evaluation. Author Summary According to standard models, when confronted with a choice, animals and humans rely on two separate, distinct processes to come to a decision. One process deliberatively evaluates the consequences of each candidate action and is thought to underlie the ability to flexibly come up with novel plans. The other process gradually increases the propensity to perform behaviors that were previously successful and is thought to underlie automatically executed, habitual reflexes. Although computational principles and animal behavior support this dichotomy, at the neural level, there is little evidence supporting a clean segregation. For instance, although dopamine — famously implicated in drug addiction and Parkinson’s disease — currently only has a well-defined role in the automatic process, evidence suggests that it also plays a role in the deliberative process. In this work, we present a computational framework for resolving this mismatch. We show that the types of behaviors associated with either process could result from a common learning mechanism applied to different strategies for how populations of neurons could represent candidate actions. In addition to demonstrating that this account can produce the full range of flexible behavior observed in the empirical literature, we suggest experiments that could detect the various approaches within this framework.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
1 retracted
RETRACTED: A mathematical model explains saturating axon guidance responses to molecular gradients
Nguyen, H., Dayan, P., Pujic, Z., Cooper-White, J., & Goodhill, G. J. (2016) · eLife · doi:10.7554/elife.12248 · Cited by 1,221
RetractedReinforcement Learning: An Introduction
Sutton, R. S., & Barto, A. G. (1998) · IEEE Transactions on Neural Networks · doi:10.1109/tnn.1998.712192 · Cited by 27,225
Reinforcement Learning: An Introduction
(2005) · IEEE Transactions on Neural Networks · doi:10.1109/tnn.2004.842673 · Cited by 24,496
(2014) · Climate Change 2013 – The Physical Science Basis · doi:10.1017/cbo9781107415324.004 · Cited by 23,880
An Integrative Theory of Prefrontal Cortex Function
Miller, E. K., & Cohen, J. D. (2001) · Annual Review of Neuroscience · doi:10.1146/annurev.neuro.24.1.167 · Cited by 12,803
A neural substrate of prediction and reward
Schultz, W., Dayan, P., & Montague, P. R. (1997) · Science · doi:10.1126/science.275.5306.1593 · Cited by 9,388
Reinforcement Learning: An Introduction
Johnson, J. D., Li, J., & Chen, Z. (2000) · Neurocomputing · doi:10.1016/s0925-2312(00)00324-6 · Cited by 8,414
(1978) · UA Campus Repository (The University of Arizona) · Cited by 8,010
Cognitive maps in rats and men
Tolman, E. C. (1948) · Psychological Review · doi:10.1037/h0061626 · Cited by 6,641
McClelland, J. L., McNaughton, B. L., & O'Reilly, R. C. (1995) · Psychological Review · doi:10.1037/0033-295x.102.3.419 · Cited by 5,256
Research works printed but not linked: 3
- BW Balleine, 2009, Neuroeconomics, Neuroeconomics, 367
- Schlegel M, White A, White M. Stable predictive representations with general value functions for continual learning. Continual Learning and Deep Networks workshop at the Neural Information Processing System Conference. 2017. Available: https://sites.ualberta.ca/~amw8/cldl.pdf
- AC Schapiro, 2013, Neural representations of events arise from temporal community structure, Nat Publ Gr, 16
Cited by
The hippocampus as a predictive map
Stachenfeld, K. L., Botvinick, M. M., & Gershman, S. J. (2017) · Nature Neuroscience · doi:10.1038/nn.4650 · Cited by 1,000
Self-citationPrioritized memory access explains planning and hippocampal replay
Mattar, M. G., & Daw, N. D. (2018) · Nature Neuroscience · doi:10.1038/s41593-018-0232-z · Cited by 393
Self-citationThe successor representation in human reinforcement learning
Momennejad, I., Russek, E. M., Cheong, J. H., Botvinick, M. M., Daw, N. D., & Gershman, S. J. (2017) · Nature Human Behaviour · doi:10.1038/s41562-017-0180-8 · Cited by 390
Self-citationA map of abstract relational knowledge in the human hippocampal-entorhinal cortex
Garvert, M. M., Dolan, R. J., & Behrens, T. E. (2017) · eLife · doi:10.7554/elife.17086 · Cited by 379
Sequential replay of nonspatial task states in the human hippocampus
Schuck, N. W., & Niv, Y. (2019) · Science · doi:10.1126/science.aaw5181 · Cited by 304
Hippocampal contributions to social and cognitive deficits in autism spectrum disorder
Banker, S. M., Gu, X., Schiller, D., & Foss-Feig, J. H. (2021) · Trends in Neurosciences · doi:10.1016/j.tins.2021.08.005 · Cited by 271
Planning and navigation as active inference
Kaplan, R., & Friston, K. J. (2018) · Biological Cybernetics · doi:10.1007/s00422-018-0753-2 · Cited by 203
The sense of should: A biologically-based framework for modeling social pressure
Theriault, J. E., Young, L., & Barrett, L. F. (2021) · Physics of Life Reviews · doi:10.1016/j.plrev.2020.01.004 · Cited by 201
The successor representation: Its computational logic and neural substrates
Gershman, S. J. (2018) · The Journal of Neuroscience · doi:10.1523/jneurosci.0151-18.2018 · Cited by 191
“For simplicity of exposition, we have chosen to omit actions, though our treatment extends straightforwardly to handle actions (Russek et al., 2017).”
BackgroundMethodResultInfluentialSelf-citationLearning structures: Predictive representations, replay, and generalization
Momennejad, I. (2020) · Current Opinion in Behavioral Sciences · doi:10.1016/j.cobeha.2020.02.017 · Cited by 159
Self-citationThe Anterior Cingulate Cortex Predicts Future States to Mediate Model-Based Action Selection
Akam, T., Rodrigues-Vaz, I., Marcelo, I., Zhang, X., Pereira, M., Oliveira, R. F., Dayan, P., & Costa, R. M. (2021) · Neuron · doi:10.1016/j.neuron.2020.10.013 · Cited by 170
Spatial goal coding in the hippocampal formation
Nyberg, N., Duvelle, É., Barry, C., & Spiers, H. J. (2022) · Neuron · doi:10.1016/j.neuron.2021.12.012 · Cited by 141
Predictive maps in rats and humans for spatial navigation
de Cothi, W., Nyberg, N., Griesbauer, E.-M., Ghanamé, C., Zisch, F., Lefort, J. M., Fletcher, L., Newton, C., Renaudineau, S., Bendor, D., et al. (2022) · Current Biology · doi:10.1016/j.cub.2022.06.090 · Cited by 109
Interoception as modeling, allostasis as control
Sennesh, E., Theriault, J., Brooks, D., van de Meent, J.-W., Barrett, L. F., & Quigley, K. S. (2022) · Biological Psychology · doi:10.1016/j.biopsycho.2021.108242 · Cited by 107
Reward revaluation biases hippocampal replay content away from the preferred outcome
Carey, A. A., Tanaka, Y., & van der Meer, M. A. A. (2019) · Nature Neuroscience · doi:10.1038/s41593-019-0464-6 · Cited by 106
Episodic retrieval for model-based evaluation in sequential decision tasks.
Zhou, C. Y., Talmi, D., Daw, N. D., & Mattar, M. G. (2025) · Psychological Review · doi:10.1037/rev0000505 · Cited by 21
Self-citationPredictive Representations: Building Blocks of Intelligence
Carvalho, W., Tomov, M. S., de Cothi, W., Barry, C., & Gershman, S. J. (2024) · Neural Computation · doi:10.1162/neco_a_01705 · Cited by 15
Self-citationA probabilistic successor representation for context-dependent learning
Geerts, J. P., Gershman, S. J., Burgess, N., & Stachenfeld, K. L. (2024) · Psychological Review · doi:10.1037/rev0000414 · Cited by 14
Self-citation
Metadata
- Authors
- Russek, E. M., Momennejad, I., Botvinick, M. M., Gershman, S. J., & Daw, N. D.
- Venue
- PLOS Computational Biology
- Publisher
- Public Library of Science (PLoS)
- Published
- 2017-09-25
- Volume
- 13
- Issue
- 9
- Pages
- e1005768
- ISSN
- 1553-734X (ISSN-L)
- 1553-7358
- OpenAlex
- W2950347959
- Cited by
- 326
- References
- 128
- FWCI
- 16.40
- Type
- journal article
- Language
- en
- Update notices
- new_version (2017-10-05)
- Open access
- gold
- Funders
- National Institute of Mental Health (US) (T32 MH019524)
- National Institutes of Health (5R01MH109177-02)
- John Templeton Foundation
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.