The Science Virus, Part 2: Machine Learning Into the Destruction of Science
The Science Wars Part XII
"Watch your thoughts, they become words;
Watch your words, they become actions;
Watch your actions, they become habits;
Watch your habits, they become character;
Watch your character, for it becomes your destiny." -Frank Outlaw
Find other articles in the Science Wars article series here.
Please tell the children that what matters first and foremost in their education is that they learn to love.
This is the Story About How Machine Learning Drives the Education System to Destroy Science
This article recently showed up on one of my numerous discussion groups, and it occurred to me how terrifying this is with respect to what I call the Science Virus.
Because…of course they are.
You are meant from the headline onward to accept the grotesque and absurd educational system, goals included, indecipherable though they may be. At no point should you question the value of the system—just think upon this emerging oddity. Remember that as we move forward.
I first described what I call the Science Virus in an article back in January.
The primary idea is that the publication of incorrect knowledge—including potentially entire sub-branches of scientific literature, and including fabricated results—can become accepted through the peer-reviewed system of publication, making it economically harder to unwind and correct the direction of science. The implications are obvious, and more frightening than ever given the way we've seen policy steered toward moneyed interests.
What I'm even more worried about at this point is the way in which proxy trust in the science can miss this poisoning of the well entirely—potentially resulting in an out-of-control feedback loop destroying any body of scientific knowledge not specifically promoted by money. I'll explain this worry as best as I can, but first, I want to backtrack to revisit the schooling of childhood. And power pose. Power posing will definitely improve the quality of this article.
Cheating at School: A Controlled Feedback Loop
In 2010, Psychologist Peter Gray published an article in Psychology Today explaining how schools are a breeding ground for cheaters, and that the "best" students cheat the most. The story is fundamentally about brainwashing the specialists into a particular form of nihilism.
Our system of compulsory (forced) schooling is almost perfectly designed to promote cheating. That is even truer today than in times past. Students are required to spend way more time than they wish doing work that they did not choose, that bores them, that seems purposeless to them. They are constantly told about the value of high grades. Grades are used as essentially the sole motivator. Everything is done for grades. Advancement through the system, and eventual freedom from it, depends upon grades.
Students become convinced that high grades and advancement to the next level are the be-all and end-all of their school work. By the time they are 11 or 12 years old, most are realistically cynical about the idea that school is fundamentally a place for learning. They realize that much of what they are required to do is senseless and that they will forget most of what they are tested on shortly after the test. They see little direct connection--because there usually is none--between their school assignments and the real world in which they live. They learn that their own questions and interests don't count. What counts are their abilities to provide the "correct" answers to questions that they did not ask and that do not interest them. And "correct" means the answers that the teachers or the test-producers are looking for, not answers that the students really understand to be correct.
Hopefully you can already connect the dots to the strangely disturbing state of specialization of our era. The article points to anonymous questionnaires in which 98% of students admit to cheating, with 70% becoming frequent and blatant cheaters. But where he really shakes up the audience is where he demonstrates the connection to Science.
The Continuity Between Cheating in School and Cheating in Science
Let's take the example of Bob, who decides at some point in college to go on to become a scientist. He makes this decision not because he really loves science or has some burning questions that he wants to answer through scientific methods. His own sense of curiosity was drilled out of him long ago. Rather, he decides to become a scientist because (a) he has always done well in science classes (only partly by cheating), (b) others have encouraged him to become a scientist, and (c) he sees that scientists have relatively high status and he would like that. In his gut, he doesn't really quite even know what it means to be a scientist, but he thinks it would be a good career.
So, Bob applies to and gets accepted into a graduate program in science leading to a Ph.D. Now, as a graduate student, he is in some sense "doing" science, as he carries out the research he must do for his doctoral dissertation. But is it "real" science, or is he still a student going through hoops? He finds that as he works on his research project--a project that was designed more by his advisor than by himself--he is not getting quite the results that his advisor expected. The advisor seems disappointed and is lavishing much more attention and praise on another Ph.D. candidate who is getting strong, positive, publishable results.
Bob gets worried about his future. He's working hard and, through no fault of his own, it's not paying off. So, the old habit of cheating returns. By manipulating just a few numbers, in some of his data sets, he turns statistically insignificant results into significant ones--results that lead to a much-praised dissertation and to a number of publications in prestigious scientific journals.
This article was a followup to one Gray had written about a young scientist's suicide upon his narcissistic mask covering a career of fabricated results being revealed by his inquisitive graduate students.
In 2012, the New York Times published an article about cheating at one of the world's premier high schools: New York City's prized magnet school, Stuyvesant. Stuyvesant has likely graduated more Nobel Laureates and elite university professors than any other high school—maybe by a wide margin.
“I’m sure everybody understood it was wrong to take other people’s work, but they had ways of rationalizing it,” said Karina Moy, a 2010 graduate of the school. “Everyone took it as a necessary evil to get through.”
Karina doesn't get it. She can't get it because she is herself institutionalized. They cheat because the system is set up such that a student either cheats a step or three up the ladder or gets passed by inferior students willing to cheat in order to clear every hurdle and jump through every hoop. By the time these students are done with freshman year at Stuy, most students have their exploratory nature beaten out of them except perhaps in their chief areas of creative focus where they surpass grade metrics easily.
That same year, Alex Tabarrok at Marginal Revolution examined cheating in terms of Bryan Caplan's signaling theory of education, which is that education is about signaling to employers, not about improving productivity gains—which is perfectly consistent with what I'll call my nihilism theory of education. From Alex,
The cheater-detector arms-race is interesting but also makes me think about the signaling theory of education. Cheating works best if the signaling model is true. If education were all about increasing productivity and if employers could measure productivity then cheating would be a waste of time. As illustrated by Mr. Smith, however, at least some students care about the A that cheating produces more than the knowledge that learning produces. Mr. Smith must believe either that education (in at least this class) doesn’t increase productivity or that employers don’t learn about productivity. Employers have big incentives to learn about productivity so my bet is on the former.
If students perceive the situation correctly we also have an interesting hypothesis: students should cheat more in those courses that offer the least productivity gains. Studies on cheating find mixed results across majors, with some finding that business majors cheat more and others not, but these studies are cross sectional, i.e. across individuals. A better test of the theory that I propose would look at cheating by the same individuals across courses. Absences should also be higher in courses with lower productivity gains.
Also during 2012, a cheating scandal rocked the halls of Harvard University before being quickly forgotten. The Last Psychiatrist called the whole scandal "stupid" while explaining to us exactly why we need to pay it great attention!
Before everyone rushes to their predetermined sides, can we ask why, when there are cheating scandals, they are almost always in introductory classes? When the stakes are lowest?
75% of the students in these kind of courses get As and Bs because of Grade Inflation. I'd put big money down that if I used a crayon to draw an elephant and a donkey I'd get at least a B+ with the margin comment, "Interesting take, could you elaborate?"
And yet the students here felt compelled to cheat. Take a minute away from your self-righteousness and put yourself in their shoes. Did they not think they could get an A on their own? Or.... is "cheating" the only way to create the kind of answer that the professor wants?
"Using in-text citations to support your answer" is the standard way academics pretend at knowledge, and it is always a trick, it doesn't allow the reader "a better understanding of your thought process," it is an appeal to authority (Salmon 2006) masquerading as critical thinking (Ennis 1987). But it sure makes grading easier, here is the answer key: >5 references: A. 3-4 references: B Etc.
If I gave this test to other government professors not affiliated with the course, I'm sure they'd have good answers-- but would it be "what the professor is looking for?" That's the phrase that alerts you to the fact that the class isn't designed for you to learn but for him to teach. All for the fair market price of $2000.
In the system, to cheat is to perform even more for the brainwashing agent at the head of the classroom. That and it's a signal to employers that you read an absurdly long set of instructions to reach the point at which you're incentivized to cheat.
The Atlantic had its own cocktail party approved take on academic cheating before the whole topic was to be forgotten for another generation. It was perhaps the most bluntly nihilistic imaginable, ending with a scenic description and totalitarian memory loss through the process that is somewhat reminiscent of the closing pages of Animal Farm.
After hanging up, Nancy sat down on the curb. Above her were leafy trees and crisscrossing wires; on the street one block over, the honking of cars. She was alone in the city. She couldn’t remember if that was how she’d imagined this life would turn out. Maybe she’d had other plans, but none of them mattered, because back then, we all thought we knew just what we wanted, and who could say now that we hadn’t gotten it?
Ron Unz blames the situation on the winner-take-all Western economy that masquerades as the American Meritocracy. This is certainly true for some people, but I'll draw from the intelligence community's description of the motivations of double agents: M.I.C.E.
Pick your rat poison, so to speak.
From Catastrophe to Worse
The situation was bad enough when artificial intelligence arrived on the scene.
Let us be fair in what I'm calling artificial intelligence. We're not talking about human-level cyborgs or smartphones just yet. We're talking about the machine learning algorithms that make up many of today's killer apps on the World Wide Web—and Google in particular. I'm not here to repeat the Vice article's warning of whizzy new Open AI paper-generating code. If I have a warning, it's this:
These students naturally believe the artificial intelligence toys they use enhance them as academics. After all, they are rewarded for their acceptance of augmentation. But these AI toys are controlled by a few technologists, and the information they serve comes from the poisoned wells of Science and other knowledge domains. This student augmentation serves not to accelerate the education of the "best" students, but rather to expedite their brainwashing and cognitive dissonance. These students are then predisposed not to correct errors in the knowledge domains, and that only encourages any new poisoning of the well that results in somebody's profit.
Put simply: the use of machine learning algorithms in such a flawed educational system is a recipe for the totalitarian destruction of science. The implications could not be larger.
From the article:
“Because I used Open AI I didn’t feel the constant anxiety of needing to focus all my time on writing it,” AeUsako_, who also asked to use their online pseudonym, told Motherboard.
The interviewed student answers from the pseudo-identity of their online handle. How apropos. Five points to Digital Slytherin for demonstrating the lack of transparency in the entire process.