Neuroeconomics

Lesions and the Valuation of Information

Information economics is based on the premise that information is a scarce commodity. Because its acquisition is costly, the economic actor weighs the value of an additional unit of information against his/her highest valued alternative. In this sense, information is no different than any tangible good that can be bought and sold. The September issue of Science includes a paper by Rudebeck, Buckley, Walton and Rushworth that integrates neuroecon and info-econ via a lesion study on macaques:

We assessed how selective lesions of ACC gyrus (ACC_G), ACC sulcus (ACC_S), or lateral orbital and ventral prefrontal cortex (PFv+o) affect the way macaques value social information.

And how were the monkeys’ valuations measured?

Measurements were made of latencies to pick up food items in the presence of fear-inducing stimuli (toy snakes) in experiment 1, social stimuli (short films of other macaques) in experiment 2, or neutral control objects. The latencies indexed the macaques’ assessment of the value of obtaining further information about the stimulus before reaching and reflected their relative valuation of the stimulus in contrast to the incentive of the food.

When confronted with the fear-inducing information,

[Control animals with no lesion] were consistently reluctant to take food from above a moving toy snake, but the case was significantly different for both the PFv+o group and the ACC_S group…There were no significant differences between the ACC_G and controls.

The behavior of the ACC_G group, however, was significantly different from the other three groups when social information was presented:

[Controls] were slower to pick up food in the presence of a large staring male, a female macaque with visible sexual perineal swellings, and a midsized macaque making affiliative lip-smacking gestures…The ACC_G group remained uninterested in any of the images of other macaques…By contrast, comparable modulations of reaching latencies were seen in PFv+o and ACC_S groups as in the controls, and there were no significant differences between any of these three groups. 

The results enable the researchers to draw the following conclusions:

The ACC region critical for mediating the valuation of social stimuli appears to be the ACC_G immediately rostral and dorsal to the genu of the corpus callosum and includes areas 32 and rostral area 24. 

Complete ACC_S lesions only affected the mild fear task significantly and not the social task. 

(Note that Brodmann’s area 32 refers to the dorsal ACC, while BA 24 refers to the ventral ACC. The experiment described here suggests that the lesions to the dorsal ACC and the ventral ACC may have differential effects that sum to produce the observed social deviancy.)

If the latency measure legitimately represents the monkeys’ opportunity cost of acquiring more information, these results suggest the following economic interpretation: (1) An ACC_G lesion decreases monkeys’ willingness to pay for social information; (2) Lesions of the ACC_S and lateral orbital and ventral prefrontal cortex increase monkeys’ willingness to pay for information relating to fearful stimuli.

Orbitofrontal cortex specific to time preference

An article in Neuron this week finds a population of OFC neurons that fire when receiving a reward after a short delay, and a smaller population that fires when receiving a reward after a longer delay. Authors claim that this is contrary to the concept of the OFC being an all purpose utility totalizer; the smaller population which is sensitive to delayed reward assists in learning behaviors which yield a distant reward and the larger population of neurons provides stronger reinforcement for immediate reward. The issue also has a nice review which incorporates the findings into the larger body of literature.

Are we bad forecasters?

Predications of emotional impact weigh heavily on decisions. In fact, people avoid risk even when faced with the prospect of large gain (PDF), predicting loss will hurt them much more than an equal gain will please them. If that is true, this phenomenon (termed loss aversion) is simply a rational product of accurate affective forecasting.  Currently, research seems split on this question.  Studies have indicated that loss induces more intense neural activity, indicating that our forecasting may be valid.  However, behavioral economics generally proposes that we are bad forecasters, and studies show that we consistently overestimate the intensity of emotion from life tragedy.

In a new study, participants effectively minimized impact of loss after a game of luck using various coping mechanisms, such as dissonance reduction, self-affirmation, motivated reasoning, and positive illusions (PDF). Researchers found that “there was no evidence that losing actually had a greater emotional impact than winning,” showing we are indeed poor loss forecasters.

So what of studies reporting more intense neural response to loss?  Those studies could be measuring only very immediate effects before subjects could employ the above coping mechanisms.   However, until further research is done, it seems that the current study should be taken simply to show that, as with other losses, we have effective methods of coping with monetary loss that we do not factor in to our affective forecasts.

Interestingly, participants were far better at assessing levels of positive affect from winning; in one of the two games participant guesses were spot-on, and in the other were very close.

Intuition as effective decision making system

The hippocampus is well known as the primary seat for memory.  Previous research suggested that this memory agent is also crucial for making relational (transitive) judgments, though this has been debated by researchers (PDF).  A new study in Psychological Science may help to resolve this question by illustrating that it is not necessary for such judgments or probabilistic learning.  In the study, hippocampal deactivation not only didn't prevent, but actually enhanced, participants' ability to make transitive inferences.  Researchers propose that the basal ganglia dopamine system, known to work in competition with the hippocampus, is the primary actor in implicit learning and was allowed fuller riegn while the hippocampus was deactivated.  Results indicates that the hippocampus is only critical before intuitive learning can kick in over multiple experiences, such as for early stages of explicit learning or when applying new information to novel concepts quickly.

This is interesting also because it illustrates two memory systems utilized during decision making.  And who knows - certain tasks are better learned and performed using reinforcement and intuition, so drugs such as the benzodiazepine used in this study may become learning tools.

Anterior cingulate cortex in decision making

Integrating past rewards and losses into risk and payoff calculations is fundamental to most decisions, and the dopaminergic system has been implicated as the primary actor in this process. However, a new study in Nature Neuroscience suggests that one player, the anterior cingulate cortex (ACC), may be more important in this process than previously thought. Once seen as a simple error-detection device used when presented alternative choices, Kennerley et al. indicate that the ACC used to guide choices by assessing past experiences and understanding the value of options.  In fact, the previous understanding of ACC function may be off, as lesioned monkeys did not perform significantly worse on error detection. 

Kennerly et al. also note that control monkeys (i.e., monkeys without lesions) only corrected their errors 68% of the time, leading them to conclude that "errors and rewards do not naturally operate like the explicit sensory cues that instruct actions in a conditional learning protocol."  Rather, negative outcomes are seen as evidence and weighed against recent history to determine the value of competing alternatives - and that is just what the ACC seems to handle from this study.

Although lesioned monkeys were unable to incorporate reinforcement history into their decisions, Kennerly et al. are quick to point out that their simian subjects were still able to make rational decisions in certain cases.  These situations typically involved selecting between two present visual stimuli that differ in work and reward, which indicates that the ACC is just one part of a larger system including the orbitofrontal and prefrontal cortices.

Decision Science News

I've just had a look at Dan Goldstein's Decision Science News.  He has a very nice informative blog about current research and news in the decision sciences.  For examples, see his blogs on What is Behavioral Economics? or What is Neuroeconomics? , but also some of the more intersting topics like,  Honesty, No longer the best policy?  I wonder in turn how this work relates to the work by Uri Gneezy called "Deception: The role of consequences."

Addiction and Cue-Triggered Decision Processes

Very interesting to see this article from Bernheim and Rangel in December's AER on addiction and decision making. In addition to coming up with a nifty model presenting the notion that drug use among addicts is a "cue-triggered mistake", the authors then expand into policy prescriptions based on their model.

My confusion, however, starts in the abstract: "use among addicts is frequently a mistake" To back this up the authors give us a story about how we can't infer preferences from observed behavior:

There are plainly circumstances in which it makes no sense to infer preferences from choices. For example, American visitors to the U.K. suffer numerous injuries and fatalities because they often look only to the left before stepping into streets, even though they know traffic approaches from the right. One cannot reasonably attribute this to the pleasure of looking left or to masochistic preferences.

Well, sure, the first time you step into the street I could call that a mistake. But someone who keeps getting run over repeatedly has some additional problems. The authors go into a Hedonic Forecasting Mechanism which is disrupted by the addictive substance to skew the perception of the addict that he's going to get much more pleasure from the addictive substance than he actually will.

Maybe.

Another explanation might be that when someone is addicted to a drug they get an abnormal amount of pleasure from it (as opposed to someone who is not addicted) due to the relief from the craving. In that case there is no skewed perception at all. The person is forecasting correctly that he will greatly enjoy getting the monkey off of his back.

This also fits in to what we know about the effects of attentive/reward drugs on neurotransmitter receptors. When your synapse is flooded with a particular neurotransmitter, the neuron can remove some of the receptor sites that neurotransmitter is binding to. In a nutshell, that means that such a neuron is less receptive to that neurotransmitter.

Fewer receptors mean that when you're not on drugs you're actually feeling worse, which is what results in cravings. The choice becomes one of inter-temporal substitution. Drugs now which will have an immediate relief of symptoms and generate direct reward, or abstain, suffer now, and get an indirect reward sometime in the future.

So getting back to this concept of skewed perception, the non addicted user might go from 0-60 on the Hedon-O-Meter, but because of the down-regulation of receptors the addict is starting at a negative value. Suppose that utility is function of the cubed root of opioids:

The non addicted user has a normal level of 0 for opiates and is getting a 'net' utility of 0 from them. Let's say the drug taken increases opiates by a level of 40. The user now gets utility of 3.4. Repeated use, however, down-regulates the opiate receptors so that when the user is not taking drugs, they are experiencing a -20 level of opiates and a utility of -2.7 (withdrawal). This user takes the same dose and moves from -20 to +20, and a utility of +2.7, but the total change in utility is from -2.7 to +2.7, or 5.4 which is greater than 3.4 received by the non-user.

The model of Bernheim and Rangel has recovered addicts (their receptor levels have returned to normal) being stimulated by usage cues (hanging out with drug buddies, Jimmi Hendrix, etc...) activating the motivational circuitry to seek out drugs. I guess so, but maybe the activation is in response to something else. If you buy into the notion that addicts use drugs in response to cravings, it could be that the cues are associated with the craving rather than just the motivation. As any smoker trying to quit will tell you, they start to crave cigarettes when they are presented with smoking cues (usually hanging out in a bar). It makes much more sense that it is the craving that stimulates the motivational behavior rather than the cues alone. Even though the 'underlying preferences' may be to stay clean, the individual experiences disutility from the craving now and is motivated to satisfy it.  And if the craving is stopped by using the drug,  that's not a mistake, it's the a rational decision to satisfy a craving.