by R. Passov
During my third year as an undergrad — I took six+ years in all — I ran into an old friend (cocaine) and failed all of my classes. Failed by not taking my finals which proved a blessing. After recovering from the shock of having done something I had sworn many times I’d never do again, a kind counselor offered to change my F’s to incompletes.
And, she suggested filing a ‘petition 631’ which would allow me to gain credits through negotiation with a willing professor. The hope was that with a little direct supervision, I’d finish college.
I was a (lapsed) physics major at the time, working in the laboratory of Dr. A. Theodore Forrester (formerly Finkelstein.) Alongside an ex-tank commander from the Israeli army, I labored on a containment device for a plasma from which an ion beam was to be focused toward hydrogen, inside a tokamak, so as to create enough pressure to induce fusion. Or something like that.
But I didn’t comprehend the physics and was even further away from understanding how to use the lathe with the tolerances prescribed by Dr. Forrester and so asking him to act as a personal guide seemed more than a long shot.
I had taken introductory economics, taught by a professor who was blessed with more than enough mirth to keep a hall of two hundred students enthralled. But in his office he was retiring, not interested in helping a struggling student.
The other economics course, Industrial Organization, I chose because it fit around my work schedule. Around twenty of us met three times a week to listen to a young professor talk way over our heads.
I tried the readings, tried the lecture notes that you could buy from the student book store, tried the arguments over and over in my head, all the way to a B+, of which I was proud. I earned that grade.
I believe I poked my heard into his office as I sensed sternness masked the uncertainty of being new to his profession. Inside his office, I sensed he really did want to smile.
Remember me, I said? I was in your Industrial Organization class.
After listening in a very professional manner to a student in need of an alternative path, he offered to work with me and suggested I read a book on a subject of interest, then lent me his copy. I scoured that book, struggling through the lingo and the models, finally stumbling upon something of interest.
We attacked the idea, developing a model to the point where it seemed to say something interesting. Then, the Professor told me it was time to write the paper. A few weeks later, I dropped off a draft.
Not long after, we sat behind his desk. With doctor-like patience, he explained the concept of a paragraph, how to frame ideas, how to check for changes in tense, how to bring the reader through a complicated discussion. I remain his student.
Many years later, Steve invited me to join a working group that meets bi-monthly to discuss ‘Decision Making Under Uncertainty.’ For the most part, we are a collection of former students who teach or practice somewhere between data science, economics and computer science.
Each participant submits one or more papers. We vote and the most popular wins. The last two reads, as well as the upcoming read, have been my suggestions:
- The Anchoring Bias Reflects Rational Use of Cognitive Resources; Lieder, Griffiths, et.al, Psychonomic Bulletin & Review, May 2017
- Rationality and Intelligence: A Brief Update; Stuart Russell, from Fundamental Issues of Artificial Intelligence, pgs 7 – 28, Springer, 2016; and
- Counterperformativity; A. Bamford & D. Mackenzie, New Left Review, 113: 97-121. (2018)
In a letter I plan to share with Steve, here’s why I’ve suggested these papers (though if I had a do over, I would put Russell ahead of Griffiths.)
I remain in your camp, believing that the value of an equilibrium argument lies not in equilibrium being obtained but rather in the mental discipline of attempting to identify the forces at work.
Same for the pedagogy of rational expectations, whether presented in finance or economics. I saw it as necessary to admitting mathematical argument. Griffiths supports this in the opening remarks to his paper: Rational decision making allows for a process to be constructed upon logic and logic opens the front door to mathematics.
Thomas Kuhn, writing in The Structure of Scientific Revolutions, describes the self-reinforcing processes that develop post the acceptance of a paradigm: An election process that restricts questions to only those in support of further development of what has already been accepted. 1
Eventually, according to Kuhn, something comes along that cannot be fit comfortably into the domain of the existing paradigm. At this juncture, Kuhn argues the new paradigm and the existing one have no institutional means through which to resolve their competition and so the outcome is revolutionary.
This almost happened in economics when Tversky and Kahneman showed us Prospect Theory and the collection of anchors and biases that defy rational decision making. 2,3 The result, as put by Griffiths – economics got rocked, creating the substrate for a classic paradigm shift.
But surprisingly, that’s not what happened.
Economists came to their own rescue, recasting our biases as heuristics; techniques extracted through an ecological process that are fit-for-purpose. No process is revealed, just the acceptance that the mind cannot defy the rules of the natural universe: it cannot reason, whatever its process of reasoning might be, infinitely fast. Therefore, it is, it must be, a constrained optimization engine. Consequently, anchors and biases have been recast as rational short cuts to decisions that, on average, have been right. 4,5
Things could have stayed on this path, economists redeeming their own foundations by reinterpreting the work of iconoclasts not to arrive at a new paradigm but instead to further the existing one.
However, in their efforts to save their foundation, economists let the guard down, the one that Milton Freidman hoped would keep economics apart, as its own science, and not subsumed into applied mathematics. The guard, according to Freidman, was continuing the development of theories in economics that arise in advance of math. 6
While the economists were working to salvage their foundation, the artificial intelligence community, with Russell at the vanguard, set out to define intelligence. Since perfect rationality leads to an explosion of computation, intelligence in a machine, they’ve concluded, is a matter of ‘bounded optimality.’
Bounded optimality is a heuristic; a path through the puzzle that suffices.
The computer scientists are sneaking in. The mind, they say, is indeed a computer and once the costs of time and the expense of changing the state space are accounted for, bounded optimization works well as a paradigm of intelligence.
Following on from Russell, Griffiths, et. al. “…illustrate general properties of resource-rational information processing.” And while they do not claim “… the brain implements the sampling algorithm we analyzed…” they do write their “… qualitative predictions … characterize bounded rationality for a more general class of cognitive architectures. Importantly, this class includes biologically plausible neural network implementations…
I once jumped out of an airplane. As I approached the ground, I kept my focus on the horizon. This allowed me to rely on my autonomic nervous system, an embedded function that runs as a utility. While begging the question as to how particular functions get embedded and why, an action potential obviated the need for any meta processes, saving time and resources.
I suppose if we are systems built to optimize our survival subject to constraints we can make some sense of this. We have capabilities in layers and there is a logic to the hierarchy, as survival tasks are embedded while math (most likely) is not, and this hierarchy of embedding implies a recursive-ness (which some believe gives rise to consciousness.)
There is a second recursive-ness here. If the paradigm is that humans are similar to their machines, then we’re constraining ourselves to proving this. According to Kuhn, the terms we use define the problem; the tools we build, fitted to discover what we believe to be there.
The third paper offers examples of Performativity, a fancy word for the apparent fact that humans create systems that modify their own behavior. Humans engineer environments in which they operate. These engineered spaces contain feedback loops which appear to play upon the same forces that, overtime, cause the development of our heuristics.
The Black Scholes options pricing model is an imagined solution to an imaginary problem; the price for the chance to be better off in some uncertain future. And yet, an industry of seemingly immeasurable size continues in an unremitting feedback loop, marginally aiding the allocation of capital in good times, severely exacerbating the hiccups in bad times.
But we’ve gone further than developing imaginary solutions to imaginary problems; we now have a vast data collection engine operating through algorithms that are designed to produce outputs where the producer is us: we have built an engine in which we are immersed, the purpose of which, at the highest level of abstraction, is to condition us and it’s working.
If this is true or even partly true then it seems economics will have to go down a level, closer to the granularity of thought, finally mixing psychology with computer science, perhaps forever subjugating itself.
- The Structure of Scientific Revolutions; Thomas A. Kuhn, University of Chicago Press, (1962.)
- Judgement Under Uncertainty: Heuristics and Biases; Tversky, A; Kahneman, D (1974) Science, 185
- Prospect Theory: An Analysis of Decision under Risk; Kahneman, D ; Tversky, A,. (1979) Econometrica 47.
- Fast and Frugal Heuristics: Simple decision rules based on bounded and ecological rationality. Liu, Y., Gigerenzer, G., & Todd, P. M. Journal of Psychological Science, 26, 56–60. (1999)
- Models of Ecological Rationality. D. Goldstein, G. Gigerenzer, Psychological Review, Vol. 109, No 1, 75 –90. (2002)
- Essays in Positive Economics. Milton Friedman, The Methodology of Positive Economics, 3–43, (1953).