Space Cynics

(courtesy of the Old Space Cadet)

A recent article (1) by Davide Marchiori and Massimo Warglien and an accompanying commentary (2) by Michael Cohen describes an apparent human characteristic that could explain the dorkiness of space cadets.

I have often wondered why otherwise intelligent people (space-cadets) can be so obtuse and uncooperative in furthering their goal of getting the human race to be space-faring. The characteristic trash talk describing other space cadets and other efforts by various space cadets has always appeared to be self-defeating and almost descriptive of a circular firing squad. Certainly, there is enough native intelligence and skill for a group of committed space cadets to work together and accomplish their goals, but they invariably fail to do so and discussions ultimately fall into acrimonious arguments about trivia.

John Forbes Nash (the central character in the movie “A Beautiful Mind”) had postulated that learning is driven by feedback based on optimizing expected gains in mixed-strategy games. His equilibrium predictions provide poor fits to actual data and real players. Apparently, Nash was wrong.

Another class of decision-making strategies is based on regret rather than optimizing expected outcome. That is, minimizing regret. In this instance, regret is defined as the difference between the best possible outcome and a potential outcome.

Consider a highly simplified medical example.

A patient considers which of two proposed surgical procedures to accept. The first has an 80 percent success rate and the second has a 60 percent success rate.

Maximizing expected gain according to Nash would cause the patient to accept the first procedure.

However, assume that the first procedure has a 5 percent mortality rate and the second has a one percent mortality. A person minimizing maximum regret would then accept the second procedure because the first procedure is five times more likely to kill him than is the second procedure.

A person looking to minimize overall regret would have to weigh two relative costs: That of death and that of a failure of the surgical procedure to provide a cure. The patient would weigh the relative consequences and decide accordingly. Procedure One has an 80 percent success rate implying a 20 percent failure rate along with a 5 percent mortality. The second procedure has a 40 percent failure rate along with a one percent mortality. This presents the patient with the following possible outcomes as part of the decision-making process:

Procedure 1
Cure the Condition 80%
Live with Condition 15%
Die 5%

Procedure 2
Cure the Condition 60%
Live with Condition 39%
Die 1%

Refuse Surgery
Cure the Condition 0%
Live with Condition 100%
Die 0%

If the patient judges that dying is no more than twice as bad as living with a failed procedure, he will determine that Procedure One has a 15 percent live failure and a 5 percent fatality or three to one failure to death ratio. Procedure Two has a 39 percent live failure and one percent fatality or 39 to one failure to death ratio. Based on his judgment that dying is no more than twice as bad as living with the condition, the patient would choose Procedure One. The person seeking to minimize the maximum regret, dying, would be most rational in rejecting both procedures. These are examples of regret-driven decision making processes.

Marchiori and Warglien examined simulated learning by many different neural network models and found that the models that most closely mimicked actual games were regret driven. Therefore, the predictions of outcome are based on learning by looking backwards and being driven by regret rather than by looking forwards to expectations of gains. As stated by Cohen: “Choices in economic games are predicted better by models that look back at what might have been, instead of looking forward to maximum gain.”

This mechanism may explain why space cadets act in a way that seems destructive. Instead of decision-making by forward-looking maximizing of expected gains – that is, deciding among options that maximize the rate at which humans approach space-faring capability, their decision-making is regret-driven.

The space cadets look backwards and have a concept of what might have been, or the maximum potential gain that humanity could have had – the pointy-nosed spaceships all over the solar system within a half century of space flight. They then compare what might have been to the current state as they perceive it. Their regret is the difference between the ideal potential (fantasy) and a current fantasy based on their perception of reality. That causes them to minimize the regret by substituting their fantasy position for current reality and thereby argue unrealistically for their own fantasy position.

No single person can be current in all aspects of space-associated science and technology. Therefore, every Space Cadet will have gaps in his knowledge base which can be filled with optimized fantasy or rationalization. Since that fantasy position is idiosyncratic and not perfectly correlated with the fantasy positions of other space cadets, they fight vigorously in support of their own positions and reject any effort at cooperating with others to minimize their regret or avoid threatening their view of reality.

If this hypothesis is viable, then the bickering, refusal to cooperate in furthering a common goal, and refusal to accept current reality is hard-wired into the behavior of at least a subset of the human race. That subset seeks idealized frontiers of one type or another. If they are exposed to spaceflight during their formative years as in Heinlein’s science fiction, they are either driven to or attracted to spaceflight. That is, they dive into their own distorted navels.

A recent article by Chandrasekhar and coworkers (3) reported areas of brain activation documented by fMRI during regret and rejoice. It appears that the medial orbitofrontal cortex, left superior frontal cortex, right angular gyrus, and left thalamus are activated during regret. Furthermore, the right inferior orbitofrontal cortex, presupplementary motor area, and both anterior and posterior cingulates are activated during both regret and rejoice. The authors suggest that the latter areas may be activated by surprise from realization of relatively unlikely events as well. Could this provide a neurological basis for regret-driven decision-making? If the cingulates are activated by surprise as well, space cadets may have theirs lit up markedly by an alt.space success.

My friend and fellow space cadet Sam Dinkin (4) has thrown some alternative hypotheses into the mix. One is that people approach causes in order to belong to a morose support group rather than to solve real problems. A second alternative hypothesis is that both parties know that they are dealing with a low probability game and therefore don’t spare the feelings of others because they know it doesn’t matter. I might note that these functional explanations are consistent with anatomical findings of the Chandrasekhar study.

Now if we could just define some experiments to perform on space cadets that would allow us to differentiate between these hypotheses.

References:

Davide Marchiori and Massimo Warglien: “Predicting Human Interactive Learning by Regret-Driven Neural Networks” Science, pp. 1111-1113, February 22, 2008.

Michael Cohen: Comments on “Predicting Human Interactive Learning by Regret-Driven Neural Networks” Science, pp. 1052-1053, February 22, 2008.

Pammi Chandrasekhar, C. Monica Capra, Sara Moore, Charles Noussair, and Gregory Berns: Neurobiological Regret and Rejoice Functions for Aversive Outcomes” NeuroImage 39:1472-1484, 2008.

Sam Dinkin: Personal communication, March 6, 2008.