Experimentally revealed stochastic preferences for multicomponent choice options.

Abstract

Realistic, everyday rewards contain multiple components. An apple has taste and size. However, we choose in single dimensions, simply preferring some apples to others. How can such single-dimensional preference relationships refer to multicomponent choice options? Here, we measured how stochastic choices revealed preferences for 2-component milkshakes. The preferences were intuitively graphed as indifference curves that represented the orderly integration of the 2 components as trade-off: parts of 1 component were given up for obtaining 1 additional unit of the other component without a change in preference. The well-ordered, nonoverlapping curves satisfied leave-one-out tests, followed predictions by machine learning decoders and correlated with single-dimensional Becker-DeGroot-Marschak (BDM) auction-like bids for the 2-component rewards. This accuracy suggests a decision process that integrates multiple reward components into single-dimensional estimates in a systematic fashion. In interspecies comparisons, human performance matched that of highly experienced laboratory monkeys, as measured by accuracy of the critical trade-off between bundle components. These data describe the nature of choices of multicomponent choice options and attest to the validity of the rigorous economic concepts and their convenient graphic schemes for explaining choices of human and nonhuman primates. The results encourage formal behavioral and neural investigations of normal, irrational, and pathological economic choices.

On the importance of feedback for categorization: Revisiting category learning experiments using an adaptive filter model.

“…First, it is assumed that this system learns associations through dopamine mediated reinforcement processes (Ashby & Valentin, 2017). Second, it is assumed that dopamine pathways are relevant for reward processing because they code for unexpected errors (Pastor-Bernier et al, 2020; Schultz, 1999). Because the ALF model updates its coefficients when an error is made, Basal Ganglia circuitry seems like a reasonable biological structure to implement processes like those described in the current work.…”

Nutrient-sensitive reinforcement learning in monkeys

“…This approach views reward values as physiological signals that serve to maintain homeostasis. However, fat and sugar can be preferred even without corresponding nutrient deficits 22,23,44 ; therefore, the hedonic values of foods cannot be explained solely by homeostatic regulations of nutrient deficits. Future experiments could challenge the nutrient states of animals during food choices to estimate empirical nutrient-value functions from state-dependent choice patterns to refine these models.…”

“…For instance, monkeys spend more time in food patches associated with a high probability of nutritious foods (e.g., nuts) while ignoring more frequent low-nutrient foods (e.g., leaves). Primates, including humans, also exhibit individual subjective preferences for specific nutrients and sensory food qualities to regulate nutrient intake [17][18][19][20][21][22][23][24] . Thus, ecological data suggest that animals consider both the nutritional value of food and the food's availability.…”

Mechanisms Underlying Economic Revealed Preference Of Multi-Component Bundles In Primate Orbitofrontal Cortex

“…The BOLD signal variation across the subjects accounted for a great deal of the variation in WTP observed, providing evidence that, even in the absence of choice, the brain was evaluating the quality of the items presented. In recent human studies (Pastor-Bernier et al, 2020; Seak et al, 2021) stochastic preferences for bundles containing various juice components high-lipid or high-sugar containing milkshakes were presented passively viewing subjects that performed a BDM auction. ICs represented revealed stochastic preferences for the bundles presented systematically (Pastor-Bernier et al, 2020) and the BOLD signals in the reward-related brain structures, including striatum, midbrain and medial orbitofrontal cortex followed the characteristic pattern of ICs: similar responses along ICs, but monotonic changes across ICs (Seak et al, 2020).…”

“…In recent human studies (Pastor-Bernier et al, 2020; Seak et al, 2021) stochastic preferences for bundles containing various juice components high-lipid or high-sugar containing milkshakes were presented passively viewing subjects that performed a BDM auction. ICs represented revealed stochastic preferences for the bundles presented systematically (Pastor-Bernier et al, 2020) and the BOLD signals in the reward-related brain structures, including striatum, midbrain and medial orbitofrontal cortex followed the characteristic pattern of ICs: similar responses along ICs, but monotonic changes across ICs (Seak et al, 2020). These brain structures integrated multiple reward components into a scalar signal and reflected the synergies between the components, thus corroborating previous observations for bundles consisting of fat and carbohydrates (DiFeliceantonio et al, 2018).…”