Available online 7 December 2023

Author links open overlay panel, , , Abstract1

The Free Energy Principle (FEP) is a normative computational framework for iterative reduction of prediction error and uncertainty through perception-intervention cycles that has been presented as a potential unifying theory of all brain functions (Friston, 2006). Decision-making is an important cognitive faculty whose mechanisms must be explained by any theory hoping to unify the brain sciences. This challenge has been accepted by several proponents of the FEP (Friston, 2010; Gershman, 2019). We evaluate these attempts to reduce decision-making to the FEP, using Lucas’ (2005) meta-theory of the brain’s contextual constraints as a guidepost. We find variants of the FEP unable to explain behavior in certain types of diagnostic, predictive, and multi-armed bandit decision tasks. We trace the shortcomings to the theory’s lack of an adequate notion of utility, a concept central to decision-making and grounded in the brain’s biological reality. We argue that any attempts to fully reduce utility to the FEP would require unrealistic assumptions, making the principle an unlikely candidate for unifying brain science. We suggest that researchers instead attempt to identify contexts in which either informational or reward constraints predominate, delimiting the FEP’s area of applicability. To encourage this type of research, we propose a two-factor formal framework that can subsume any FEP model and allows experimenters to estimate the contributions of informational versus reward constraints to behavior.

Section snippetsIntroduction: For the Love of a Paradigm

Establishing a unifying theory, a “paradigm” in Kuhnian history of science, is a Holy Grail of any scientific discipline (Kuhn, 2012). As demonstrated by the success of atomic and evolutionary theories in physical and biological sciences, it is possible to have such unifying theories that transform piles of “incommensurable” evidence into additive parts of a bigger whole, not only guiding but also encouraging research in relevant domains. In the absence of a similar grand theory, the

Free Energy versus Utility in Decision-Making

Decision-making is a central human faculty. For the FEP to serve as a unifying principle for the brain, it needs to address its processes and outcomes. Valuation is at the core of decision-making, reflected in the presumption of subjective utility’s existence across the behavioral decision sciences (Barberà et al., 2004). For the FEP to unify the brain sciences, decision goals and their mapping to options according to this valuation process must be reducible to error and/or uncertainty

Free Energy and Utility in Decision-Making

If we concede that the FEP is an unlikely candidate for a unifying theory of the brain, what would the prospect be for moving towards a paradigmatic brain science? One possibility is that there will never be a unifying theory. Although the optimization approach to biology has been fruitful in explaining why brains developed certain properties (Varshney et al., 2016), some describe the mind-brain system as a kludge—a quick-and-dirty solution that is clumsy, inelegant, inefficient, difficult to

Optimality Versus Efficiency in Brain Science

A major appeal of the FEP’s computational machinery lies in the use of approximate Bayesian inference to estimate parameters in a manner that is optimal up to a given threshold. This allows normative models to be developed without adjusting the theory’s fundamental equations, the claimed aligning of which with empirical data provides the basis for supporting the ambitious Bayesian Brain Hypothesis (Friston, 2010). Would the adoption of a utility-aware approach mean the Hypothesis must be

Summary and Open Questions

Unifying theories, or at least unifying principles that can serve as their components, are the Holy Grails of every science (Kuhn, 2012). We examined the performance of one proposed candidate for fulfilling this role in the brain sciences: the Free Energy Principle (Friston, 2006). According to the theory, the cornerstone of brain processing is an attempt to reduce prediction error and uncertainty. Agents are driven to accomplish this through developing accurate models of the environment, but

Funding

This work was supported by the Department of Defense, Defense Advanced Research Projects Activity (DARPA), via Contract 2019-HR00111990067 to the University of Illinois Urbana-Champaign (PI: Aron K. Barbey). The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official policies or endorsements, either expressed or implied, of DARPA, or the US Government. The US Government is authorized to reproduce and distribute

Uncited reference

Collins and Frank, 2014.

CRediT authorship contribution statement

Babak Hemmatian: Writing – original draft, Writing – review & editing. Lav R. Varshney: Writing – original draft, Writing – review & editing. Frederick Pi: Writing – original draft, Writing – review & editing. Aron K. Barbey: Writing – original draft, Writing – review & editing.

References (60)R. Ratcliff et al.Diffusion decision model: Current issues and history

Trends in cognitive sciences

(2016)

Ashton Smith, M. (2023, September 10). The FEP+ Model of General Intelligence....Barberà, S., Hammond, P., & Seidl, C. (Eds.). (2004). Handbook of utility theory (2 volumes). Springer Science &...Barbey, A. K., Karama, S., & Haier, R. J. (2021). The Cambridge Handbook of Intelligence and Cognitive Neuroscience....Bouneffouf, D., & Rish, I. (2019). A survey on practical applications of multi-armed and contextual bandits. arXiv...Clark, A. (2015). Surfing Uncertainty: Prediction, Action, and the Embodied Mind. New York: Oxford University...A.G. Collins et al.Opponent actor learning (OpAL): modeling interactive effects of striatal dopamine on reinforcement learning and choice incentive

Psychological review

(2014)

A. Fengler et al.Likelihood approximation networks (LANs) for fast inference of simulation models in cognitive neuroscience

Elife

(2021)

A. Fengler et al.Beyond Drift Diffusion Models: Fitting a Broad Class of Decision and Reinforcement Learning Models with HDDM

Journal of cognitive neuroscience

(2022)

P.M. Fernbach et al.Asymmetries in predictive and diagnostic reasoning

Journal of Experimental Psychology: General

(2011)

P.C. FishburnSubjective expected utility: A review of normative theories

Theory and decision

(1981)

K.J. Friston et al.Reinforcement learning or active inference?

PloS one

(2009)

K. FristonThe free-energy principle: a unified brain theory?

Nature reviews neuroscience

(2010)

K. Friston et al.Free-energy minimization and the dark-room problem

Frontiers in psychology

(2012)

K. Friston et al.The anatomy of choice: dopamine and decision-making

Philosophical Transactions of the Royal Society B: Biological Sciences

(2014)

Gershman, S. J. (2019). What does the free energy principle tell us about the brain?. arXiv preprint...Gigerenzer, G., & Todd, P. M. (1999). Fast and frugal heuristics: The adaptive toolbox. In G. Gigerenzer, P. M. Todd, &...Gittins, J. C. (1989). Multi-armed bandit allocation indices. Chichester: John Wiley & Sons,...J.I. Gold et al.The neural basis of decision making

Annual Review of Neuroscience

(2007)

Guilford, J.P. (1967). The nature of human intelligence....View full text

© 2023 Elsevier Ltd. All rights reserved.