Models of Economic Choice

The “standard model” of economic decisions posits that value itself is computed in the prefrontal cortex such as the orbital or medial prefrontal cortex. Such value information is then transmitted to motor-related brain regions such as the lateral intraparietal cortex and the cingulate cortex where the utility is integrated with effort cost and the final utility is computed, which guides the choice amongst multiple goods. Early evidence challenging this proposal gained support mainly from the observations that neurons in the orbitofrontal cortex (OFC) encode both the pre-and post-decision variables, which suggests that economic decisions may be resolved in the OFC. We are further addressing this question by recording from multiple brain regions simultaneously using a multi-attribute choice task.

Reference Frames for Value Representation

We are currently taking a deeper look into the flexible representations of value and its implications for the organization of neural circuits for economic choice. Choice tasks often afford multiple reference frames. In general, any characteristics of the choice task can provide a valid reference frame. In choice tasks where options differ by a particular determinant – probability, delay, cost, etc. – a valid reference frame may be that defined by that determinant. The fact that goods and values may be represented in multiple reference frames has far-reaching implications. One line of research suggests that OFC is criti­cal for constructing internal models of the environment to determine optimal behaviors that will yield specific out­comes. One characteristic of this theory is that it allows any task­ relevant informa­tion to be represented within OFC so that a given neuron might encode arbitrary variables under different task conditions. However, one recent study argued against this proposal by demonstrating that within commodity-based reference frame, OFC neurons consistently encode the same key decision variables under different behavioral contexts. This organizational stability could conserve the interaction of specific populations of neurons that perform the same computation across multiple contexts. We are pushing the limit of such organizational stability by testing neuronal coding in single vs multi-attribute decision context and across different reference frames.

Context-dependency in Value-based Decisions

Behavioral studies show that subjective value of any particular good can be affected by the "historical" context. In particular, higher effort can bias preference, increasing the value animals and humans ascribed to commodities previously acquired at greater cost/effort. Early studies showed that after rats were trained to obtain the same amount of food associated with different levels of effort, it takes more trials in extinction (both food and effort requirement were removed) for the animals to stop responding to the reward predicting cue previously associated with higher effort. Under the choice paradigm, shifts in preferences towards reinforcers following greater effort were observed in a variety of species. For example, when provided with a choice between cues previously associated with high or low effort, human subjects prefer visual cues previously associated with high effort. Such behaviors were described in terms of cognitive dissonance, stating that “individuals assign greater value to high-effort outcomes in an attempt to reduce the “dissonance” (conflict) that would otherwise occur by working hard to gain a goal.” However, an alternative hypothesis is that encoding of value is history dependent. Under current context, it suggests that the value of a reinforcer depends on the relative improvement in conditions leading up to the reinforcer. This hypothesis is consistent with growing behavioral evidence indicating that choice behavior in both animals and humans is often context dependent, for example, varying depending on the exposure to previous priming situations. It has been shown that cognitive dissonance mainly engages the dorsal anterior cingulate cortex and anterior insula while the OFC neurons encode subjective value of goods. Thus, the two hypotheses can be potentially disambiguated at the neuronal level. 

Neuroeconomics and Psychopathology 

NMDA receptors are increasingly implicated in psychiatric disorders such as schizophrenia. The NMDA antagonist treatment is considered a mechanistically relevant model for cognitive deficits of schizophrenia. Functional imaging studies suggest that, after systemic administration of NMDA antagonists, the most affected brain region is the prefrontal cortex (PFC). Animal studies also suggest that the PFC is a key region responsible for the expression of behaviors that are relevant to schizophrenia, including working memory and reward processing. Therefore, it is important to understand the impact of this treatment on prefrontal cellular mechanisms that lead to the disruption of these cognitive functions. To this end, under the economic choice task, we will record from the same population of neurons and compare their activities while the animals are ON and OFF acute NMDA antagonist treatment. This project is part of a key science project of which I am a principal investigator (Nonhuman primate whole brain connectome and disease models) funded by the Shanghai Municipality.