Tuesday, January 27, 2009
A recent paper by Camillo Padoa-Schioppa and John Assad investigates the encoding of value in the brain for rhesus monkeys, a renewal of efforts to understand how and whether the assumption of transitivity, so crucial to the behavioral sciences, is in fact an accurate description of how individuals act when making choices. It builds on a previous paper (2006) of theirs in Nature, 'Neurons in the orbitofrontal cortex encode economic value.' The novelty in their (2008) research is that they provide evidence for value-encoding of juice consumption as menu-invariant, i.e. invariant to changes in the goods that are rank-preferred by the monkeys. I'm going to give you a brief summary and then make some points on where I'd be interested in seeing future research go.
There are three basic characteristics that they define. First, offer value responses which encode quantity or value of one or two offered juices (products). Second, chosen value responses which encode the value of the chosen juice (product) independent of the juice (product) type. Third, taste responses which are binary responses reflecting which one of two juices (products) are chosen independent of the amount of the juice (product). I have highlighted the notion of juice as product to foreground the idea of parallels between monkey juice consumption and the consumption of products by humans. Padoa-Schioppa and Assad assert that the second response, the chosen value response, is the most interesting because it is subjective and it ``represents a common unit for qualitatively different goods.'' (95) Their experimental methods are, thus, engaged with unearthing and isolating this chosen value effect. It is on this basis that we must assess their experimental methods and design.
The authors ran experiments with two (a male and a female) rhesus monkeys, measuring the neuronal output in the orbitofrontal cortex (OFC). The monkeys were prepared for the experiments using normal surgical procedures in preparation for neuronal recording. Juice was delivered to the monkeys in specifically measured quanta. The instructions, as such, that the monkeys were given were consistent with previous literature on the subject, with the monkeys selecting from a set of coloured squares indicating type and quantity of juice. The monkeys had an initial fixation point, after which a `go' signal was given: the introduction of two saccade targets adjacent to the coloured squares. Once the monkey had maintained fixation on the saccade target that they chose, juice would be delivered to them. Trials were aborted if the monkey broke fixation before the `go' signal. This procedure is highly detailed and thus easily replicable with appropriate resources.
The theory was that there are three potential neuronal responses: offer value, chosen value and taste responses. The argument was that the Orbitofrontal Cortex (OFC) neurons have previously responded to delivery of a given juice depending on the context (Tremblay & Schultz
1999). However, that study did not assess whether these encoded responses were invariant to changes in menu. They designed an experiment in which they could isolate whether preference encoding is independent of menu choice and thus whether any (A,B,C) triple selected for an experimental session would allow for valid encoding of value in the OFC, value which would be consistent across menus. The crucial underlying distinction though is that the set of juices given fits the ranking A > B > C, with any one juice falling into a class A, B or C. This brings us to the question of how the juice rankings are derived, for which we do not have the data. Nevertheless, the final result was that 1 unit of an A class juice was equivalent to 1.3 units of a B class juice and one unit of a class B juice was approximately equivalent to 3 units of a class C juice. Finally, they showed that 1 unit of a class A juice was roughly equivalent to 4 units of a class C juice. Hence you have approximate value indifference (1 unit A = 1.3 x 3 ~= 4).
Thus the authors cover a broad base of theory related to preference transitivity and show that with distinct and well-defined preferences, values are encoded and the structure of these values reflects the mathematical conditions of value transitivity and indifference transitivity. This is very interesting and definitely warrants further research in other monkeys and for a larger sample of individuals, rather than it being some quirk of the two monkeys in the experiment.
Would it not be cool, ladies and gentlemen, if instead of using disjoint sets of ranked goods they doing something that attempted to replicate what Simonson and Tversky (1992) (and most recently reinvigorated in Dan Ariely's book Predictably Irrational) discussed in house choice - seeing two potential consumption goods and then introducing what would be called an 'irrelevant alternative' in economic theory and seeing how the type of 'irrelevant alternative' (or the degree of similarity of the irrelevant alternative to one of the actual choices) affected value encoding. To me, this would bridge the gap from what would be a traditionally 'behavioral economics' assertion (for which we require, I believe, more evidence) and that of a good neurocellular economics discussion. A question for further research I suppose.
Camillo Padoa-Schioppa, John A Assad (2007). The representation of economic value in the orbitofrontal cortex is invariant for changes of menu Nature Neuroscience, 11 (1), 95-102 DOI: 10.1038/nn2020