Ask LessWrong: Design a degree in Rationality.

by XFrequentist1 min read13th May 201138 comments


Personal Blog

You are a tenured professor at a medium-sized public university. The Interdisciplinary Gods have smiled upon you, and you have been handed an operating grant, office space, and broad design powers to create an advanced interdisciplinary degree1 in "Rationality" (you suspect your department head reads LessWrong). Your students will come from a broad range of disciplines, and you cannot assume that they will posses any particular prior knowledge.

Candidates in your program could take courses in any department, as long as you have personally approved a course as eligible for credits. [ETA] All admitted students will be awarded tuition wavers and living wages. A compelling ROI calculation was a requirement for admission, and all students have demonstrated some impressive real-world accomplishments.

You thumb through your University's course register2, seeing a long list of courses in a variety of disciplines: Anthropology to Writing and Humanistic Studies. Without some constraints, you think, this degree will be incoherent.

Which do you include?

[1] To avoid tangential conversation, don't worry about what sort of degree. This could be the course load for a PhD/MBA/MA/etc.

[2] If it helps you think this through, use the MIT OCW listings to make suggestions. HT: nerzhin

38 comments, sorted by Highlighting new comments since Today at 7:06 PM
New Comment

Programming, Probability, and Cognitive Science are all useful, but I think that there's a lot being missed.

Writing: Practice dissolving questions, noticing the dark arts, and clear expression in a format natural to humans. Cover a wide variety of memes, and come up with a bigger world model and better understanding of other people's viewpoints. Other people are different, and you will not be able to effectively interact with them until you understand where they're coming from.

Economics: Economics is fairly counterintuitive, and has the benefit of being useful. It explains some of how the world actually works, and what kind of interventions are effective in it. Lots of economics ideas pop up in LW conversation (marginal utility, expected value, etc.)

Research: Every year, students seeking this degree attack another field, working with scientists in the university in not directly rationality related fields. If results aren't produced, the program is doing something wrong.

Presentation: Personality skills like fashion, improv, etc. to help rationalists effectively interact with other people.

Electives: Group projects based in other fields that the students are interested in. Emphasize effective coordination, communication, and results. Harvey Mudd's Clinic Program seems like a good place to start.

I once briefly pursued an advanced interdisciplinary degree. With the benefit of hindsight, I'd say that, generally, pursuing an advanced interdisciplinary degree -- in particular, taking out a student loan to pursue an interdisciplinary degree -- is prima facie evidence that your thinking is irrational.

All admitted students will be awarded tuition wavers and living wages. A compelling ROI calculation was a requirement for admission, and all students have demonstrated some real-world badass achievements.

Which classes are worthy of such a potent batch?

Fair enough! So, in the spirit of things, I would start with Psychology courses that demonstrate just how irrational and unreasonable human minds are, not only as the result of pathology or brain damage, but just in the normal course of life. Nothing new or unfamiliar to anybody who has read the sequences.

Separately, I'd suggest these high-achieving students be allowed to test out of any required courses in the program, to avoid redundancy and save time.

And psychology courses would include stuff on perception and how it can be tricked, such as optical illusions.

Oops. I just proposed a solution before talking about the problem.

What are the goals of this program?

The goal of my post? To get people brainstorming, so they could give me some ideas for something like this.

The goal of the degree in the scenario? To create formidable rationalists who would go out and solve big problems in the real world (thereby bringing prestige to the institution).

If I were this professor I would make every effort to design it only as a minor. Even better if I were allowed, I would just make it two one-semester courses. This is for two reasons - the first is that rationality should be used for things, and so planning for someone to major in rationality would be silly. The second is that the core content of rationality really isn't that big - it's just a bit hard.

So what courses would I list as prerequisites for my rationality courses? A statistics class, and that's it. Philosophy, neuroscience etc. are icing on the cake and can be applied with short readings.

Bayesian statistics, right?

Any sort - as long as it teaches you the standard notation and basic identities (including Bayes' theorem, so in that sense Bayesian :P ). My reasoning is that I'd have to spend two or three weeks covering the basics of probability if I didn't require it, and statistics is a useful life skill so requiring it wouldn't waste very many peoples' time.

Of course at some point I could mention that probabilities usually come from reasoning from incomplete information, which is a property of you and not the world. That really only takes a paragraph if you're already covering map and territory.

Can I propose entirely new classes to add? I think a good rationality degree would need at least a few.

Existing classes on math (probability theory and Bayesian statistics, information theory, etc. and possibly a new one on decision theory), computer science, cognitive science (to introduce heuristics and biases and to encourage naturalistic algorithmic thinking about one's own mind), and maybe evolutionary psychology would lead into new classes on reductionism and technical thinking. A second semester of reductionism would cover standard philosophical questions, as practice, but there would be a much higher standard of reduvtive technical scientifically-informed explanation than philosophy courses usually set. Any other philosophy courses would not spend much time on the subject often known as "philosophy" but better described as "history of philosophy", e.g. reading (and somehow taking seriously) the opinions of dead dudes who didn't know anything about cogsci or compsci or probability/decision theory or physics. We may look at old philosophers to find examples of what not to do, but that's about it. Though I think there would be room for further philosophy courses that go deeper into epistemology, ethics, metaethics, ontology, etc. with the same high standards of reasoning.

A basic class on modern AI should probably fit in there somewhere, to further encourage and demonstrate technical thinking about mental algorithms, and maybe some physics, with the same intention as the quantum physics sequence.

Anyway, I think that would be a good start for an epistemic rationality program. I'm much less able to envision an instrumental rationality program, though I'm inclined to think that if someone does come up with one that works well (perhaps drawing from Rationality Boot Camp after a few years of experience and refinement), it should be taught (at least in part) to most or all students, rather than just as part of a specialized program. I'd love to see more people learning reductionism and cogsci and Bayesian reasoning and such, but realistically, high-quality instrumental rationality instruction would probably be significantly more appealing and useful to most people.

Can I propose entirely new classes to add?


I'm much less able to envision an instrumental rationality program

This will be achieved through the practicum, in which students will take a level in Badass [trope]!

Since we're taking students from varied and heterogeneous backgrounds and it's an advanced degree, I'd have a list of required topics, with the students being able to place out of the area of their undergraduate study (if their undergrad major covered one of the topics).

Core areas would include:

  • Probability/statistics

  • Mathematics (at least through basic calc and linear algebra)

  • Computer science (at least basic programming, algorithms, and software architecture)

  • Natural science (chemistry OR biology OR physics)

  • Research experience in a natural science or engineering lab of choice

  • Psychology (emphasis on cognitive biases and memory)

  • Anthropology

  • Philosophy (overview course on historical perspectives)

Also, added seminar courses with mini-units to tie subjects together and place them in context.

Electives would be open-ended, pending an essay to justify their selection.

Anyone have thoughts on whether a business or economics course should be included? I considered that, but I have not taken a formal course in those topics myself, and so don't have a good estimate of their actual utility.

Why programming or other computer science? It is important to be able to know how to use a computer, but beyond that, what's the benefit (unless the student is going to be a professional programmer)? Perhaps it trains the mind in a useful way?

Exactly that. Being able to think in explicit algorithms is extremely useful for decoding your own thoughts and being able to actually change your mind.

Cheeky answer: I ignore all enrolled students, and simply award the degree to anyone who succeeds in the real world, however they do it.

I will give a more academy-oriented answer that ensures the degree recipient has actually assimilated rationalist thought modes and practices after I've had more time to think about this.

simply award the degree to anyone who succeeds in the real world, however they do it.

How about converting credit hours into a point based system (x points = 1 credit hour towards degree)?

Have the students identify a project/goal to pursue which they will submit to you, assign maximum points for rational actions that turn out well; decrement points for clearly harmful actions; assign half points when they have devised a good rational argument for pursuing the action ahead of time, even though it didn't work out so well.

The only unclear case is when it works but 'shouldn't have', which often times deserves a reward, because it isn't good to criticize people when they are right, but what if they simply bought a lottery ticket on a whim and one of their goals was to become rich? What if they admit they had no extra-normal reason to think that this method would work? Should you award them points or not?

Cheeky, but a fair point.

I edited the scenario slightly, so that students had been pre-screened for some indication of potential in the real world.

I'm pretty sure we've talked about similar things before - the closest thing I could find after a quick search is a list of math prerequisites.

A suggestion: to make this concrete, we could identify specific courses on MIT's open coarse project. Then people could actually "get" this degree in some sense.

I'll start with the obvious: Math 18.05, Introduction to Probability and Statistics.

I'll start with the obvious: Math 18.05, Introduction to Probability and Statistics.

It seems obvious that street fighting mathematics would be highly appropriate.

Upvoted, absolutely!

[-][anonymous]9y 2

I'm a bit late to the party, but wanted to comment because I'm sort-of actually doing this. I'm a third year Information Technology major (a jack-of-all-trades computing major, plus some small group communication classes) at Rochester Institute of Technology. I wanted to do a custom minor in rationality, but the liberal arts college doesn't have custom minors, so I'm taking my "rationality minor" informally.

I won't be able to take the exact classes on this list because of logistics. Next year, RIT is switching from quarters to semesters, so some of these classes will probably be eliminated or merged with others during that. I also plant to spend one of my three remaining grading periods studying abroad. So this list shows the sort of classes I want to take, but not necessarily the ones I will actually be able to take. (Also, there are no links because the course registration system is behind a password and the non-password version is a pain to navigate.)


  • Persuasion (taking now)

  • Rhetorical Theory

  • Interpersonal Communication or Small Group Communication


  • Behavior Modification

  • Cognitive Psychology

  • Judgement and Decision Making

Philosophy: I actually have just about no idea what philosophy I'll be taking. I'm hoping I'll have a better idea what areas to focus on after taking the intro course. I'll also be asking the advice of a rationalist friend who has taken some philosophy at my university (and so probably knows which classes and professors are worth taking). At a glance, this is what looks possibly useful:

  • Theories of Knowledge

  • Philosophy of the Mind

  • (formal) Logic

Other things that look useful:

  • History of Science (I've heard that Kuhn is the textbook)

A typical minor at RIT is a list of almost all the mid-level classes from a single department that says to choose any five. A rationality minor would probably have to be "choose any 5, and at least one from each category" to ensure variety.

I haven't included programming and discrete math on this list because my major requires them, but I'd recommend them to someone else doing this who wasn't a computing major. In absence of Bayesian statistics classes, a frequentist statistics class should probably be included, but again, it's already required for my major. (RIT does teach Bayesian statistics, but only as a small topic within other classes, and I'd rather not take even more frequentist statistics. I would take Bayes as a part of a data mining class if I got the chance, though.)

I have heard from people who have been to graduate school that undergraduate degrees are "just for learning how to think", so I could see a rationality major being useful for students who plan to go on to graduate school, or as a double major, if it becomes an option but lacks scholarships. (I could see a giant minor, or a major only available as a double major, working well.) What exactly it would include would depend on what the typical student already knows, and is already taking for other graduation requirements.

(edited for formatting)

In addition to "normal" classes, I'd add some game-like challenges, that can earn extra Quirrel points or whatever.

For example: the class is split in teams of five. Using only the contents of a standard kitchen, each team must find a way to move a egg as far as possible without breaking it (catapult, chariot on wheels ... the exact rules should be such that creative cheating could be possible, though the obvious cheats (hard-boiling the egg) should be banned). Points are awarded depending on how far your egg got relative to the other teams.

Variant number one: there is a traitor in each team. Traitors get points proportional to how badly their team fared (Possibly, teams may get extra points for identifying the traitor).

Variant number two: before launching the devices, each team presents it's device to the other students. Then for each device, each student records a probability that it will be among the first three, and a probability that it will be among the last three. Points are awarded according to how well you guessed, and to how many people in other teams bet your device would win (so you have an incentive to oversell your device).

Variant number three: each team builds an egg-moving device, but it is launched by another team. You must fit the device and any instructions inside a shoe box, and the other team has exactly one minute to launch the egg. Points are split between the builders and the launchers.

Variant number four: While building the device, the students don't know which of the above variant they're playing, they don't know if there's a traitor, if someone else will be launching the device, whether they should prepare an advertisement speech, etc., they're only told once the device is done.

... actually, if I had to design a degree, what I'd actually do is first list a lot of challenges like that (and rule variants), ask friends to design some too, and for each one list what kind of skills it helps build, what kind of skills it helps test (for this example: the base exercise is teamwork and creativity and craft, adding traitors means you have to balance he teamwork with paranoia), and then make a more careful selection of challenges.

Actually, a good homework question for a lot of subjects would be "How would you design a little game or challenge that uses this skill or knowledge you just learned" - just thinking about that would probably make the knowledge stick a bit better, in addition to giving you a huge "suggestions" box.

I'm in an interdisciplinary honors program at the moment and love it. It's not specifically anything to do with rationality, but I am using the system to level up as fast as I can. My own thoughts:

  1. This shouldn't be the only program the students are in. I'm in an interdisciplinary honors program AND I'm studying chemistry. Hopefully, this program would train rationalists to actually go into a field of their choosing and make a substantial difference, so they really do need to also be studying the field they intend to go into.

  2. Students should have at least one class per semester in a small group setting exclusively with other students in the program. The professors of these classes should be the best faculty that can be rounded up. These classes would probably be focused on applications of rationality, taking the knowledge that students have learned in their other classes and learning to actually use it on problems. Additionally, classes like this would help the students in the program develop closer ties to each other.

  3. Students should get at least an introduction to biology, chemistry, and physics. They certainly don't need to become experts in all of those fields, or even in one of them, but they should at least have some basic foundational knowledge in all of the major sciences. They should also have a psych course about heuristics and biases.

  4. They should study math, at least up to statistics/probability theory. This will presumably include an education in calculus and algebra.

  5. They should take at least one class on computer programming, ideally in a language that's very useful (I don't have a whole lot of experience in programming, but from what I know, I would recommend Python. I'm open to suggestions though from people who know more than me)

  6. The program I'm in has a required class for all freshmen, one semester, that teaches classical logic, Bayes theorem, Game Theory, and Decision theory. Our professor was a standard causal decision theorist, so I disagreed with some of the things he taught, but generally it was a good class, and I think something analogous would be a good class for the students in this program.

  7. Students should take courses on some, but probably not all, of these topics: evolutionary biology, economics, evolutionary psychology, neuroscience, etc. Set this up as a "take of classes from the following list" in the course catalog.

[-][anonymous]10y 1

"The program I'm in has a required class for all freshmen, one semester, that teaches classical logic, Bayes theorem, Game Theory, and Decision theory."

Really? Cool. May I ask which university and professor, or would that Bruce Wayne your Batman too much?

I'm sure that if someone really wanted to figure out who I was, it would be pretty easy. So, if you want to figure out what it is, I'll make you do two google searches (I just tried it, it's easy). The two things you will need to know is that the professor's name is Joshua Dever (I don't think he's teaching the course now, because it isn't on his website), and that the program that I am in is an interdisciplinary honors program.

If anyone reading this is a high school student looking at colleges, I highly recommend that you consider this program

I would start with a computer science or software engineering program, trim a bit, and fill the remaining space with custom-built rationality courses.

I would also reserve at least one course-slot per semester reserved for presentation of miscellaneous of topics that are too small to fill a whole semester, and allow any professor to claim a short block of that time, and assign readings and exercises for it. The best lessons are those with a high value-to-length ratio, which means a short length, which means not being long enough to fill a course.

Like I asked the dude above - why computer science or software engineering? I don't know any programming languages, but I'm guessing they might help someone think logically. Perhaps a dedicated logic course would be better for that, though?

Of the top of my head, there are three reasons to at least a little programming, aside from the practical benefit. First, it's a domain with immediate feedback. Mistakes are much more apparent, so you can cycle through making mistakes and learning from them quicker. Second, it teaches precise thinking. Implementing something requires a level of understanding and precision that it's easy to think you have with other subjects when it isn't really there. Third, it integrates well with other subjects like logic or math.

What is the purpose of getting this degree? making money or researching tough rationality related problems?

The latter. These are to be the guys (n'gals) that the rich donor funds to work on tough problems.

If you figure out which courses are most likely to make you be the rich donor, I'd be very interested to hear it!

making money isn't generally that hard, it's more the time commitment that gets most people. the careers with the strongest average salary outlooks have well defined paths into them. the traditional doctor and lawyer paths suck due to very high education costs, winner takes all effects (big drop off in salary outside of top 10 schools for both professions), and strong competition.

low risk high reward: any engineering discipline. medium risk very high reward: math or physics degree and go into financial analysis.

both those scenarios seem to me to involve less work than I see my peers putting into law and medical school (followed by residencies/apprenticeship).

[-][anonymous]10y 0

Carnegie Mellon University has a program as close as any I've come across thus far.

Ph.D. in Logic, Computation, and Methodology

They also offer an undergraduate program, B.S. in Logic and Computation.

ETA: Also, they regularly permit students in the Ph.D. program to concurrently achieve Master's in different fields, allowing even more flexibility while still remaining relevant to rationality-relata. For example, students and alumni have gotten the M.S. in Mathematics, and in Machine Learning, among other areas.

Is a one-size-fits-all list really a sound constraint on the desired solution? ISTM that the curriculum likely to suit each student will be idiosyncratic.

How do you plan to evaluate the quality of proposed solutions?

I like how you think!

Is a one-size-fits-all list really a sound constraint on the desired solution? ISTM that the curriculum likely to suit each student will be idiosyncratic.

The task was to pick options that would constitute the set of eligible courses. The student would pick their course load from within that set. A "Directed Reading in X" course would be a fine answer, IMHO.

How do you plan to evaluate the quality of proposed solutions?

I don't - this is a brainstorming exercise!

I'm not sure I'm getting the point of the exercise qua exercise. If this were a real-life opportunity my first inclination would be to explore how far I could stray from conventions and normal habits of thought. E.g. give the students themselves the task of designing the program for optimal results (which is why it'd help to define the results sought). So maybe the first thing they'd study would be the psychology of learning; math and probability theory would take a back seat to that, or would be taken as tools to understand the research on learning.

There are things it is rational to want, no matter what else you want. If you do not precisely know what you want, then the first thing you want is self-knowledge.

I like how you think!


Can you tell me more about "the psychology of learning"?

That's a broad question. Specific examples: what is it about my brain that makes some topics (e.g. math) hard to absorb and others (e.g. languages) easier. How does the traditional lecture format compare to so-called experiential learning (students are set a problem to solve and given access to conceptual tools that may help solve it), in terms of retention, comprehension etc.

Constructivism to name only one theory holds roughly that knowledge is always created anew, not transferred. (Note in particular the section "Influence on computer science".) Consider how this theory would change your predictions on what kinds of settings work well or poorly for learning.