CoChat

Evidence for surprise minimization over value maximization in choice behavior

Open access

The DOI matches its Crossref record; 47 of the 47 references it cites were found in a registry or an index.

Abstract

Classical economic models are predicated on the idea that the ultimate aim of choice is to maximize utility or reward. In contrast, an alternative perspective highlights the fact that adaptive behavior requires agents' to model their environment and minimize surprise about the states they frequent. We propose that choice behavior can be more accurately accounted for by surprise minimization compared to reward or utility maximization alone. Minimizing surprise makes a prediction at variance with expected utility models; namely, that in addition to attaining valuable states, agents attempt to maximize the entropy over outcomes and thus 'keep their options open'. We tested this prediction using a simple binary choice paradigm and show that human decision-making is better explained by surprise minimization compared to utility maximization. Furthermore, we replicated this entropy-seeking behavior in a control task with no explicit utilities. These findings highlight a limitation of purely economic motivations in explaining choice behavior and instead emphasize the importance of belief-based motivations.

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.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/srep16575

Cites

47 of 47 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Difference between classical economic models of choice and active inference. On the left, an example of two offers is illustrated, where offer 1 (left, blue bars—middle panel) contains two possible outcomes and offer 2 (right, yellow bars—middle panel) contains four different outcomes with equal probability. Based on a given set of preferences (grey bars—top panel), the expected utilities of offer 1 and 2 are almost identical, whereas the entropy over outcomes is higher in offer 2. On the right, choice tendencies under classical economic models and under surprise minimization are illustrated, where denotes a (softmax-) choice rule. While classical economic models predict that behaviour will be governed by maximising expected utility alone (top right), active inference predicts that agents are additionally compelled to maximize entropy over outcomes (bottom right).
Figure 1 Difference between classical economic models of choice and active inference. On the left, an example of two offers is illustrated, where offer 1 (left, blue bars—middle panel) contains two possible outcomes and offer 2 (right, yellow bars—middle panel) contains four different outcomes with equal probability. Based on a given set of preferences (grey bars—top panel), the expected utilities of offer 1 and 2 are almost identical, whereas the entropy over outcomes is higher in offer 2. On the right, choice tendencies under classical economic models and under surprise minimization are illustrated, where denotes a (softmax-) choice rule. While classical economic models predict that behaviour will be governed by maximising expected utility alone (top right), active inference predicts that agents are additionally compelled to maximize entropy over outcomes (bottom right).CC BY · PubMed Central
Figure 2 Task. Subjects had to choose between two offers containing between one and four different snacks. Snacks could only appear once in each option, but the same could appear in each offer. Subjects knew that at each trial one of the snacks of their chosen offer would be selected randomly (without knowing which one) and at the end of the experiment they would receive one of the selected snacks. Thus, the number of snacks in the chosen option determined the probability of winning one of snacks. In this example, subjects had a 1/3-chance of winning a cookie in the left and a 1/4 –chance of winning a cookie in the right option.
Figure 2 Task. Subjects had to choose between two offers containing between one and four different snacks. Snacks could only appear once in each option, but the same could appear in each offer. Subjects knew that at each trial one of the snacks of their chosen offer would be selected randomly (without knowing which one) and at the end of the experiment they would receive one of the selected snacks. Thus, the number of snacks in the chosen option determined the probability of winning one of snacks. In this example, subjects had a 1/3-chance of winning a cookie in the left and a 1/4 –chance of winning a cookie in the right option.CC BY · PubMed Central
Figure 3 Model-based analysis: (A) Individual regression coefficients for entropy as obtained in a linear regression predicting choices. (B) Exceedance probabilities in random-effects Bayesian model comparison provide evidence for an active inference model (with individually estimated prior precision) over expected utility theory and equivalent models with a fixed precision. Note that these exceedance probabilities are computed with respect to the BIC-scores of the models and thus adhere to an accuracy-complexity trade-off.
Figure 3 Model-based analysis: (A) Individual regression coefficients for entropy as obtained in a linear regression predicting choices. (B) Exceedance probabilities in random-effects Bayesian model comparison provide evidence for an active inference model (with individually estimated prior precision) over expected utility theory and equivalent models with a fixed precision. Note that these exceedance probabilities are computed with respect to the BIC-scores of the models and thus adhere to an accuracy-complexity trade-off.CC BY · PubMed Central
Figure 4 Comparing active inference and value maximization predictions of choice. (A) illustrates a trial where two options differ in number of snacks but only marginally in their expected utility based on a subject’s individual preference of snacks (displayed in (B)). Active inference predicts a choice tendency towards the option with more snacks (shown in (C)) while value maximization predicts a choice preference for the option with (marginally) higher expected utility (shown in (D)).
Figure 4 Comparing active inference and value maximization predictions of choice. (A) illustrates a trial where two options differ in number of snacks but only marginally in their expected utility based on a subject’s individual preference of snacks (displayed in (B)). Active inference predicts a choice tendency towards the option with more snacks (shown in (C)) while value maximization predicts a choice preference for the option with (marginally) higher expected utility (shown in (D)).CC BY · PubMed Central
Figure 5 Control task. (A) In a control experiment, subjects were asked to make a choice between options containing colors rather than snacks. After every trial, subjects could see one of the colors of their chosen option on the screen in form of a square. Crucially, the number and type of colors differed between the two options, containing between one and four different colors. (B) In a simple regression analysis, we found a substantial choice bias to choose options that maximize the entropy over outcomes as predicted by surprise minimization.
Figure 5 Control task. (A) In a control experiment, subjects were asked to make a choice between options containing colors rather than snacks. After every trial, subjects could see one of the colors of their chosen option on the screen in form of a square. Crucially, the number and type of colors differed between the two options, containing between one and four different colors. (B) In a simple regression analysis, we found a substantial choice bias to choose options that maximize the entropy over outcomes as predicted by surprise minimization.CC BY · PubMed Central

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.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

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

Show all 47 references

Research works printed but not linked: 2

  1. Sun, Y., Gomez, F. & Schmidhuber, J. Planning to Be Surprised: Optimal Bayesian Exploration in Dynamic Environments. Artif Intell 6830, 41–51 (2011).
  2. Friston, K., Adams, R., Perrinet, L. & Breakspear, M. Perception as Hypothesis, Saccades as Experiments. Front Psychol 3, 151 (2012)

Cited by

Metadata

Authors
Schwartenbeck, P., FitzGerald, T. H. B., Mathys, C., Dolan, R., Kronbichler, M., & Friston, K.
Affiliations
  • Wellcome Centre for Human Neuroimaging (GB)
  • University of Salzburg (AT)
Corresponding author
Schwartenbeck, Philipp
Venue
Scientific Reports
Published
2015-11-13
Volume
5
Issue
1
Pages
16575-16575
ISSN
2045-2322 (ISSN-L)
OpenAlex
W2415954501
Cited by
96
References
47
Citations by year
2026: 4 · 2025: 2 · 2024: 4 · 2023: 18 · 2022: 21 · 2021: 7 · 2020: 12 · 2019: 9 · 2018: 6 · 2017: 4 · 2016: 9
FWCI
4.79
Type
journal article
Language
en
Open access
Licence
Locations
  • Scientific Reports (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • UEA Digital Repository (University of East Anglia) (acceptedVersion)
  • SDL (submittedVersion, cc-by)
  • UCL Discovery (University College London) (submittedVersion)
  • MPG.PuRe (Max Planck Society) (submittedVersion)
Funders
  • Wellcome Trust (091593, 091593/Z/10/, 098362, 098362/Z/12/Z, 091593/Z/10/Z, 098362/Z/12/)
  • University College London (098362/Z/12/Z)
  • Österreichischen Akademie der Wissenschaften
  • Max-Planck-Gesellschaft (098362/Z/12/Z)
  • Universität Salzburg
Topics
  • Social Sciences › Decision Sciences › General Decision Sciences › Decision-Making and Behavioral Economics
  • Social Sciences › Economics, Econometrics and Finance › Economics and Econometrics › Experimental Behavioral Economics Studies
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Neural and Behavioral Psychology Studies
Keywords
surprise minimization, entropy maximization, choice behavior, utility maximization, expected utility, adaptive behavior, binary choice, human decision-making
MeSH terms
Adolescent, Adult, Algorithms, Choice Behavior, Decision Making, Female, Humans, Male, Motivation, Psychomotor Performance, Reward, Linear Models, Models, Economic, Young Adult

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.