Tuesday, November 16, 2021

Secret Science Club Zoom Lecture Recap: Our Gambling Brains

Tonight, my great and good friends at the Secret Science Club are presenting a lecture by neuroscientist and biomedical engineer Dr Sridevi Sarma of associate director of the Institute for Computational Medicine at Johns Hopkins University.  Today's lecture was the annual SSC/Dana Foundation collaboration.

Dr Sarma's subject was the role of the brain in gambling, and the choices that one takes while betting.

Human decisions vary even when the options stay the same- even going to the same store with the same ice cream flavors, we choose different flavors.  This is subtly different in a casino, where a person receiving the same cards might bet different amounts, even though the odds are the same.  Emotions, confidence, motivation, all these dynamic motivational biases, change over time.  What is going in the brain as these internal biases change?

We access human brains through non-invasive ways, such as non-invasive functional MRI.  Snap decisions, such as slamming on brakes while driving, are made in hundredths of a section- how does one get a snapshot of a quick decision?  

SEEG implantation is used to aid patients with drug-resistant epilepsy- seizure onset zones can be located and removed- invasive monitoring needs be to used to locate these zones- for ten patients, electrodes are implanted through holes drilled in the skull to make intracranial EEG recordings.  The electrical signals have to be examined to locate the pathological section of the brain.  While in the hospital waiting until a seizure occurs, subjects get bored, so they are provided with games, including games in which they can 'gamble'.  Their neural activity while gambling is monitored.  Why are some people more comfortable with risks?

The gambling task assigned by Dr Sarma elicits internal bias- the game is akin to War, but with an infinite, though limited deck with only five cards (2,4,6,8,10).  The patients are dealt a card, and must choose to bet 5 or 20 virtual dollars on whether or not their card is higher than a second dealer (computer) card.  If a draw ensues, no money is awarded or lost.  If, say, a player receives a 2 or a 4, it stands to reason that the lower bet would be made, an eight or a ten would probably result in a 20 dollar bet if someone wants to maximize their reward.  If a 6 is played, a risk averse person would bet five virtual dollars.

Decision strategies vary across subjects and trials.  Most subjects bet in a fashion similar to that of the Zoom lecture card- with high or low cards, decisions are made quickly- with the ambiguous 6 card, the decision is made more slowly.  Oddly, not everyone bet twenty dollars on a ten card or five on a two card- sometimes, they bet five on a ten or twenty on a two, even though ten is a sure win and two is a sure loss.  Some players elected to make poor, irrational bets.  

Across the ten subjects, there were three types of gamblers- static gamblers make the same decision on each card, always betting high on an eight or ten, low on a two or four, and low on a six.  Typical gamblers bet high on eight or ten, low on two or four, and flip-flopped on a six.  Dynamic gamblers changed their bets on every card, even it it didn't make sense- perhaps these irrational choices were due to emotions.  This has implications on gambling addiction, perhaps a subject feels lucky- 

Dr Sarma had a hypothesis about dynamic gamblers: an internal bias is causing variability within subjects and can predict behavior.  How can we test this hypothesis?  Computational modeling can help.  A gambler on a winning streak can experience an emotional bias which changes strategy- past results can be recorded and models based on it.  Behavior with fluctuate with internal states and cards receieved.  There are two strategies in a betting decision- a logical input decision and an internal state decision, and the internal state decision should be based on past results, such as winning streaks.  Internal state cannot be measured, but models can be extrapolated based on the past. This model is a state-space model of a system.  In this casee, the probability of a high bet on a 6 card is plotted against a bias signal over time- high bias signals tend to result in high bets on a 6.  Among the ten patients, there was a spectrum between logical players, who don't allow internal bias to alter their behavior to biased players, who are flip-flopping on all cards.

After this, an attempt was made to quantify the gamblers- some gamblers believed in a hot hand fallacy, allowing winning streaks to bias their opinions, while some believed in a gamblers' fallacy, with winning streaks causing pessimism, even though the odds aren't altered.

What separates humans from the rest of the animal kingdom?  Neuroscientists will answer 'cerebral cortex', Jerry Seinfield said 'money'.  Money causes people to act irrationally.  Why do people gamble poorly?  What makes them make bad decisions?  Something in the brain must be encoding bias.  Neural recordings were made of the ten patients as they gambled.  Multiple signals were measured in their brains, and the regions of the brains activated by the trials were marked.  Frequency domain analysis was made- the brain is measured much like a radio- different brain areas work at different frequencies, and a spectrum map can be made over time.  The power of neural signal oscillations and their frequencies can be used to determine which areas of the brain are active.

The insular cortex, which is involved in emotion, is involved in encoding changes of bias state.  The left side of the brain tends to increase activity when bias is low, while the right side of the brain increases when bias is high.  Encoding bias occurs at high frequency.  Bias is a right-left push-pull system.  Dr Sarma joked that it is like the old trope of the devil and the angel whispering in a character's ears.  

If a researcher knows where bias is encoded, can she decode the bias?  Can she predict intent from the bias signals of the brain?  Can bias be estimated from the results of reading a brain?  Yes, emotion can be tracked without asking a subject how he or she feels.

Dr Sarma thanked a bunch of colleagues and sponsors, including neurosurgeon Dr Jorge González-Martínez of the Cleveland Clinic, who has performed a lot of surgery on drug-resistant epilepsy patients.

The lecture was followed by a Q&A session.  The first question involved the effect of having fat stacks on decision making- Dr Sarma noted that the mere fact of having a lot of money can increase bias, as one's cash makes them more comfortable with risk.  Another question involved the subjects- these were all subjects with brain pathologies, and this was taken into account when bias encoding maps were made.  Regarding TBIs, if someone is young, their brain will rewire to some extent, with less of an incidence of this among older patients with less plastic brains.  

Dr Sarma cited Patient H.M., who underwent radical surgery for epilepsy which impaired his short-term memory, but not his long-term memory, and Phineas Gage, who had a traumatic brain injury which resulted in personality changes.

In the test, the individual with the highest executive function was playing the most rational game, and won the most money, but did not pay much attention to the game.  The individual who flip-flopped the most also paid little attention to the game.  None of the testing occurred while the patients were seen to be in danger of a seizure.  Even when inter-seizure neural activity spikes were observed, that data was discarded.

Another questioner asked if she thought that a quick course on probability would change behavior- Dr Sarma believed so.  Another question regarded the role of boredom- are subjects doing the irrational thing because they are bored?  Attention in signaled in the brain by an alpha power signal, and it was determined that bias is not a byproduct of lack of attention.  The best player was not paying much attention.

Bias is distributed throughout the brain- the push-pull system involved in go/no-go decisions is not located in one space. Depression may have played a role, one-third of epilepsy patients suffer from depression.

Another questioner asked if this information could be used to help in investment decisions.  Dr Sarma noted that there are ethical considerations.  Could brain stimulation be used to alter decisions?  Magnetic stimulation (TMS) could be used, but would that be ethical?  Could altering risk-seeking behavior have military applications?  She is not involved in that sort of work.

Addiction involves a hijacking of the risk, reward system- we feel good when we are rewarded, but the rewards aren't always the ones that were selected for evolutionarily.  Most of our evolution didn't occur while monetary systems were in existence.

Are these discrete brain studies being collated to make a grand map of the brain?  Dr Sarma noted that hers was the first study of gambling behavior and brain activity.  Other studies have involved memory, emotion.  All parts of the brain are being mapped out, the major conferences have about 30,000 attendees, so it's difficult to make one superimposed map of the brain.  Access to human patients is limited, but the NIH has special panels that evaluate studies of the patient population.  Dollars are being thrown at it, ten years ago such studies were not funded on the same level.  The community is more accepting, small studies are more common, and government funding is at a high.  The puzzle pieces are coming together, but a big project is in its infancy.

Right now, most of Dr Sarma's research is involved in improving the treatment of epilepsy patients.  While she is studying decision making and motor function in these patients, most of her work is modeling brain activity before the onset of seizures.  She is bringing computational power to clinicians, working on 'heat maps' of the brain to pinpoint seizure activity.  She is also studying chronic pain- can neural patterns in patients with chronic pain be restored to a more healthy state?

Once again, the SSC has delivered a fun, informative lecture.  Kudos to Dr Sarma, and Margaret and Dorian and the good people of the Dana Foundation.  While I haven't been able to find a video of this lecture subject, here's an interview with Dr Sarma about individualized medicine:

Pour yourself a nice beverage and soak in that SCIENCE!

No comments: