Spock and Data look at CPAN
In the two Star Trek series, the characters of Data and Spock pose a question: What would it be like to be able to make decisions on a largely, or even purely, logical basis? Not widely known is that this is a question with an answer based on evidence.
Because of tumors, surgery, and accidents, some unfortunate individuals lose all or most of their ability to make decisions based on emotion. Particularly striking are patients with ventromedial (VM) damage. Patients after VM damage seem to function normally. Their IQ is not impaired, and is sometimes above average. VM patients also do well on multiple-choice tests aimed at measuring their ability to make the kind of decisions needed to cope with everyday life. But VM patients do not make real life decisions well. They cannot hold jobs. They cannot manage their own financial affairs. Everyday decisions, even ones they make successfully on multiple-choice tests, they are unable to make in real life. When family is not available to care for them, people with VM damage live out their lives in institutions.
Data and Spock could not function as Starfleet officers. Why? Antonio Damasio studied VM patients for many years, and he gives his answer in the book Descartes' Error. Written for the non-specialist, the book reveals how we actually make complex decisions, including complex technical decisions. It is well worth studying.
Imagine looking for a module on CPAN. You look at what the modules claim to do. You note the author's name. Do you know him from elsewhere? What did you think of his other work? You look at the documentation. Well written? Does it give you the feeling that the writer knows what he is talking about? You read the reviews. You assess not just what is said, but how the review says it. Is the reviewer thoughtful or reckless? Well-informed or a newbie? Bigoted or open-minded?
These decision variables, according to Damasio, are matched to "somatic states". Somatic states can be thought of as snapshots of the body state, internal and external. Body chemistry is a major component in a somatic state. Somatic states embody (pun intended) emotional value judgements. Somatic states are the variables in the forebrain's decision-making calculator.
Our somatic state calculator is able to make complex decisions in situations where the data is not just uncertain, but deeply uncertain. Real-life uncertainty cannot be characterized as a Bell curve distribution of well-defined variables. Real-life uncertainty goes deeper than that. Deep uncertainty cannot be plotted as a distribution. In a deeply uncertain situation, we are uncertain, not only about the shape of a probability distribution, but also about the meaning and significance of the distribution itself.
Even a relatively structured real-life decision, like selecting a CPAN module, is deeply uncertain. The human brain can make quick and relatively accurate decisions in the face of deep uncertainty. An unemotional thinker could not effectively choose among CPAN modules, and could not make accurate decisions about which programming language to use. Data, Spock or a computer might be able to deal with technical subproblems if the subproblems were structured to eliminate deep uncertainty, but their logic would get lost in the technical decisions that actually occur in real-life.
Damasio theorizes that the basic "variables" of our decision making are body-images. He uses this hypothesis to outline a mechanism to explain how human decision-making came to be through evolution. Earlier brains (perhaps pre-human) would have started by associating somatic states with actions and/or things in their environment. Perhaps their memory would have contained action-to-somatic-state pairs like these:
"Tried to eat porcupine" => "Still hungry and boy do I hurt"
While informative, this by itself does not tell an organism whether it should avoid all porcupines, or whether it should merely be more cautious when they are on the menu. It would be an evolutionary advantage to be able to analyze fact-to-somatic-state data entries like the above and calculate with them in complex ways. Logic would be one tool for doing that. A creature with some degree of ability to reason logically would have a survival advantage.
Logical decision-making seems to have evolved from emotion. It remains rooted in it. The role of emotion in decision-making is not a useless vestige like the appendix. We can't cut it out without destroying our decision-making mechanism.
Closing apologies: I hope Star Trek fans will forgive my simplified picture of Spock and Data. While Spock and Data do not have the usual humanoid emotional lives, they do have emotional lives, emotional lives which play major roles in some episodes. In fact, the two series seem to spend more time on Spock's and Data's emotional issues than on those of anyone else on the Enterprise. For Spock and Data to function as well as they do, they would need to be emotional. It might be that, even in fiction, there is no way of getting around this.
Damasio's theory is complex, and like Spock and Data, it resists simplification. I hope my summary will encourage others to read Damasio directly. Descartes' Error is written for a general audience, and most of it is about his somatic marker hypothesis.
I blog about Perl, with a focus on parsing and Marpa, my parsing algorithm based on Jay Earley's.
Dammit, Spock! I'm a programmer, not a neuroscientist!
In another sense, both Data and Spock are described as super-intelligent, so I can imagine both would open all modules, read their content, run a super-fast calculations of efficiency and appropriateness and choose the one with the highest score. Data could actually, while reading, create a test-suite in his "brain" and run all the modules against that to see which fares the best.
Aside, we make decisions based on emotional bias, and we make logical reasoning to excuse our decision.
VM patients actually do the kind of data analysis Erez suggests. But they have no criteria for "fares the best" or "highest score". And without access to their somatic states, they have no way of developing criteria or of doing the other kinds of meta-reasoning that effective decisions require.
On p. 193-194 of Descartes's Error, Damasio tells of giving a VM patient a choice of two dates (both about a month away) for his next appointment. For the next half-hour, the patient analyzed the decision out loud:
"[...] previous engagements, proximity to other engagements, possible meterological conditions, virtually anything that one could reasonably think about concerning a simple date. [...] It took enormous discipline to listen to all of this without pounding on the table and telling him to stop, but we finally did tell him, quietly, that he should come on the second of the alternative dates. His response was equally calm and prompt. He simply said 'That's fine.' Back the appointment back went into his pocket and he was off. This behavior is a good example of the limits of pure reason."
In common with other VM patients, this patient could not, without emotional input, keep the decision making process directed to its goal, or even put a time limit on it. It seems likely that, if this patient were superintelligent like Data or Spock, he would have covered more ground in the half an hour, but he would have been no better at making an effective decision.
A fascinating, fascinating discussion. Thanks for sharing this with us.