A recent post over at R. Joseph Hoffmann’s blog reminded me of a post over at Less Wrong: Train Philosophers with Pearl and Kahneman, not Plato and Kant. Luke proposes that philosophy is a diseased discipline and should be training people to think critically using modern methods of rationality (e.g. Pearl), not to think critically using ancient methods of rationality (e.g. Aristotle):
Philosophical training should begin with the latest and greatest formal methods (“Pearl” for the probabilistic graphical models made famous in Pearl 1988), and the latest and greatest science (“Kahneman” for the science of human reasoning reviewed in Kahneman 2011). Beginning with Plato and Kant (and company), as most universities do today, both (1) filters for inexact thinkers, as Russell suggested, and (2) teaches people to have too much respect for failed philosophical methods that are out of touch with 20th century breakthroughs in math and science.
So, I recommend we teach young philosophy students:
more Bayesian rationality, heuristics and biases, & debiasing less informal “critical thinking skills”;
more mathematical logic & theory of computation, less term logic;
more probability theory & Bayesian scientific method, less pre-1980 philosophy of science;
more psychology of concepts & machine learning, less conceptual analysis;
more formal epistemology & computational epistemology, less pre-1980 epistemology;
more physics & cosmology,less pre-1980 metaphysics; [me: metaphysics etymologically means “after physics” so I would think that metaphysics should only be engaged after one has a firm grasp of plain ol’ physics]
more psychology of choice, less philosophy of free will;
more moral psychology, decision theory, and game theory, less intuitionist moral philosophy;
more cognitive psychology & cognitive neuroscience, less pre-1980 philosophy of mind;
more linguistics & psycholinguistics, less pre-1980 philosophy of language;
more neuroaesthetics, less aesthetics;
more causal models & psychology of causal perception, less pre-1980 theories of causation.
Of course, the reason that Luke proposes these things is because he is the CEO of an organization that is focused on building friendly AI. Which means he has to know how to correctly design a thinking machine by the very nature of his work. As Luke says elsewhere, the tool we use to philosophize is the brain, and if we don’t know how our tool works we will use it poorly. Which is why philosophers should be studying how brains actually work, making cognitive science a mandatory field of study for those who want to “use their tool properly”.
I have a post that summarizes a bit of Kahneman’s thesis in Thinking, Fast and Slow that I wrote a while back.
Luke’s second point, that we should be Bayesian rationalists, not “term logic” rationalists is probably (heh) more debatable. Again, Luke is coming from an AI framework which doesn’t afford the simplicity of deduction’s major premise, minor premise, conclusion; of the binary true/false of basic logic. An AI would be programmed using mathematical logic for determining the best course of action. And since we don’t live in a world of deductive certainty, a person who is programming a brain would use a method that can handle uncertainty, which would be a Bayesian form of rationality.
This is one of the reasons why I started looking at some basic logical fallacies that aren’t necessarily fallacies from a probability context. An argument from silence, for example, fails as a deductive argument. But it is not a failure or probability; it would actually be a reasonable inference. That being said, simply parroting “that’s a logical fallacy!” is pretty much a worthless form of engagement with reality, since almost all of our decisions are ones that are made from a frame of uncertainty. The funny thing is, evolution programmed our brains for rudimentary Bayesian rationality. Our education system of rote learning helps to dismantle it.
Luke then goes on to state:
So, my own “intro to philosophy” mega-course might be guided by the following core readings:
- Stanovich, Rationality and the Reflective Mind (2010)
- Hinman, Fundamentals of Mathematical Logic (2005)
- Russell & Norvig, Artificial Intelligence: A Modern Approach (3rd edition, 2009) — contains chapters which briefly introduce probability theory, probabilistic graphical models, computational decision theory and game theory, knowledge representation, machine learning, computational epistemology, and other useful subjects
- Sipser, Introduction to the Theory of Computation (3rd edition, 2012) — relevant to lots of philosophical problems, as discussed in Aaronson (2011)
- Howson & Urbach, Scientific Reasoning: The Bayesian Approach (3rd edition, 2005)
- Holyoak & Morrison (eds.), The Oxford Handbook of Thinking and Reasoning (2012) — contains chapters which briefly introduce the psychology of knowledge representation, concepts, categories, causal learning, explanation, argument, decision making, judgment heuristics, moral judgment, behavioral game theory, problem solving, creativity, and other useful subjects
- Dolan & Sharot (eds.), Neuroscience of Preference and Choice (2011)
- Krane, Modern Physics (3rd edition, 2012) — includes a brief introduction to cosmology
Looks like I’ve got some reading to do!