Before I deal with the question of how to connect internal and external evolutionary mechanisms, I want to show that evolutionary thinking is already emerging in current neuroeconomics. This relates to the strand of research that builds on neural networks to explain choice. The standard approach in neuroeconomics starts out from the idea of a modular brain and looks for localized mechanisms in which choice is realized in a serial way. The exemplary and seminal research along this paradigm was done by Padoa-Schioppa and collaborators and has been condensed in programmatic reviews (Padoa-Schioppa 2011). A core idea is that subjective value is embodied in certain brain regions (the orbitofrontal cortex) and that choice is governed ‘downstream’ by these neurons, such that, for example, there are no sensorimotor feedback circuits influencing subjective value. This would clearly distinguish ‘economic values’ from other forms of valuation in the brain and would thus be essential for establishing ‘neuroECONOMICS’ as a separate field of research. At the same time, this model would be congruent with the basic economic model of choice (which includes aspects such as transitivity). Consequently, Padoa-Schioppa christened the model as ‘goods based’: That means, the alternatives of choice are interpreted as ‘goods’ in the economic sense, and choice is based on an abstract and subjective representation of value aka ‘utility’.

Against this view, Hunt and Hayden (2017) have presented a model of distributed and hierarchical neural networks. This view goes back to Hebbian connectionism and therefore would be directly comparable to a Hayekian approach to neuroeconomics. The fundamental difference is that subjective value would be conceived as an emergent property of network dynamics which is grounded in regular neuronal activity of mutual inhibition that implies the recurrent comparison of activity levels in hierarchically structured networks in which those patterns of mutual inhibition would be further processed. This was exactly Hayek’s starting point. These comparisons happen across all brain areas, so that local specialization combines with cross-area connectivity. That reflects that choices are always multidimensional (such as involving taste, smell and colour of food). In addition, these networks are recurrent (a central point made by Edelman), such that sequences of neuronal activity are always fed back to the current sequence, in this sense involving processes of short-term memory. This catches the important property of real-world choices which constantly need to revaluate ongoing behaviour (for example, when to stop eating a meal). Against the sequential view, the notion of multiple time scales is added: Patterns of mutual inhibition, mapping and re-entry emerge with different time scales, which allows for complex comparisons of valuation that match with the needs of behavioural regulation. In sum, the model is following the general connectionist paradigm as developed by leading neurophilosophers and is eliminative with reference to notions such as subjective value.

This view does not preclude that at certain stages of the process, some areas or even single neurons would obtain a central, though transient function in further channelling the network dynamics. This points to the basic difficulty in neuroscience how to infer causality from certain interventions such as lesions. But the fundamental paradigm would be clearly different from the established one in neuroeconomics. This is recognized by Padoa-Schioppa. In the review Padoa-Schioppa and Konen (2017), the evolutionary approach is discussed under the heading of ‘distributed consensus model’. They reject the model mainly based on two observations, i.e. the evidence on modularization and the evidence regarding low or lack of motor feedback circuits. However, at the same time Padoa-Schioppa has turned to network models as theoretical grounding of his ‘goods-based model’. Rusticchini and Padoa-Schioppa (2015) build on a purely connectionist network model of perceptual decisions by Wang. This network endogenously reproduces the differentiation of neuron activities in different types, including subjective value representation. In fact, Hunt and Hayden cite this work as supporting their own alternative model!

How can we reconcile these different interpretations of connectionist, i.e. ‘Hayekian’ models of the brain in economics and neuroeconomics? I suggest the following. In Padoa-Schioppa (2011) the basic model grounds in the notion of ‘integration’ of multiple dimensions of choice into subjective value. One central argument in the standard model is that subjective value cannot be stored in memory because otherwise choice could not adapt flexibly to situational context. That implies, however, that integration must happen at very high speeds. If we approach this integration as an evolutionary process, we can still assume that ultimately the results may be represented in special, but transient and intermediary neuronal structures of the OFC that enable choice among the alternatives. In other words, the goods-based model would be reduced to a specific form and stage of algorithmic implementation of the more basal evolutionary process.

That being said, the standard model only analyses choices between clearly defined alternatives. Although this is indeed what economics also does in its standard model, many choices that are ‘economic’ do not fit that pattern, even though alternatives are considered. But choosing among different quantities of juice or clearly defined lotteries is very different from taking an entrepreneurial decision under uncertainty. I think if we widen the scope of economic decisions conceptually, the evolutionary approach becomes even more plausible.

Hunt, Laurence T. and Benjamin Y. Hayden (2017): A Distributed, Hierarchical and Recurrent Framework for Reward-Based Choice. Nature Reviews Neuroscience 18(3) 172–82. https://doi.org/10.1038/nrn.2017.7.

Padoa-Schioppa, Camillo (2011): Neurobiology of Economic Choice: A Good-Based Model. Annual Review of Neuroscience 34(1): 333–59. https://doi.org/10.1146/annurev-neuro-061010-113648.

Padoa-Schioppa, Camillo and Katherine E. Conen (2017): Orbitofrontal Cortex: A Neural Circuit for Economic Decisions. Neuron 968(4): 736–54. https://doi.org/10.1016/j.neuron.2017.09.031.

Rustichini, Aldo and Camillo Padoa-Schioppa (2015): A Neuro-Computational Model of Economic Decisions. Journal of Neurophysiology 114(3): 1382–98. https://doi.org/10.1152/jn.00184.2015.