Manifesto

Introduction

MIT allows its students to take any course and any number of courses from the curriculum with or without credit, where taking a course gives you access to all of the course’s learning materials. This opportunity inspires an amazing, empowering feeling that you can learn anything you want to, as long as you put in the required time and effort.

The main thing I took away from my college courses wasn’t content knowledge, but instead learning how to learn. By the time I was graduating, I felt like as long as I had good enough learning resources, I could learn anything. And I felt like if I could learn anything, I could do anything. There are many additional elements that are often prerequisites for learning and doing things – opportunities, networks, financial security, time, good health, a good K-12 education, and more – but access to high-quality learning resources, whatever form they may take, is an essential piece.

Knowledge empowers people. For many topics, and for many sequences of topics, the best way to learn them is through curated learning materials from a college course. However, while comprehensive, high-quality learning resources for primary and secondary-school students are widely available online (and put to good use by the homeschooling community!), analogous resources for college-level material are relatively scarce on the internet.

With college becoming prohibitively expensive in the United States, the use of artificial intelligence changing a lot of jobs, and the future of work looking uncertain, there is a need for an efficient and effective system that enables people to affordably gain the knowledge they need and prove what they know.

For the student looking for additional learning materials to supplement the badly taught course at their university, for the motivated high school student diving deeper into their interests, for the high school graduate building a portfolio of independent projects through self-study and going directly into the workforce, for the working professional reviewing concepts they have to apply at their job, for the working professional trying to change careers, for the teacher looking for inspiration for a course they are designing, and for many other kinds of people, high-quality college-level open education resources (OER) are an invaluable resource and a first step towards addressing this need.

What we do

SOUL aims to accelerate, experiment with, and build a stronger culture of open education at institutions of higher education, starting with MIT. We want to live in a world where anyone can learn anything from anywhere for free or at a low-cost. To that end, our main goal in the short term is to open source (e.g. via a Creative Commons license) learning materials from as many college courses as possible in a way that makes those courses “realistically learnable”. By “realistically learnable”, we mean that someone who meets the course’s prerequisites can realistically teach themselves all of the course’s content using the provided learning materials.

To achieve this, for each course, we try to put up as many materials from it as possible: videos and notes of lectures and recitations, homework assignments and solutions, exams, Q/A between students and course staff, etc – ideally everything the residential course has (and often with improvements: for example, we may rewrite unclear homework solutions and write hints for each problem to help someone get unstuck without looking at the solution).

What’s so exciting about this goal is that it’s very feasible, and we are uniquely capable of achieving it.

The course publication pipeline, at a high level, involves three things: (1) recruiting professors (getting them to agree to put their teaching materials up), (2) content collection (collecting existing and/or creating new course materials), and (3) processing (editing/subtitling video recordings and obtaining permission to use external copyrighted material).

These points are just a snippet of the many things that make it possible for us to scale up to open sourcing hundreds of classes. For more detail, see our projects.

Right now, we are prioritizing quality over quantity and focusing on building a robust infrastructure and workflow that will allow us to scale up high-quality production in the future. We are also prioritizing which courses to put up first: (1) courses corresponding to a major’s or minor’s curriculum, because we think the value of a complete course sequence is much greater than a smattering of unrelated courses, and (2) a few laboratory-intensive and discussion-focused courses that are underrepresented in OER and are uniquely difficult to translate effectively into an online setting.

Open sourcing an entire college curriculum is an exhilarating challenge that is actually very feasible for us. With the right strategy, we could make hundreds of realistically learnable courses accessible online for a fraction of the budget required to support something like OCW, all within just a few years. For instance, we are on track to make MIT’s entire undergraduate economics curriculum freely available online by 2025.

Anyone trying to innovate in the education space should be wary of overpromising and should acknowledge that the situation is probably more complicated than they realize. We are doing our best to be mindful of the limits of technology and of any single solution in disrupting education. Our approach involves systematic experimentation and, at every step, actually talking to and working with all kinds of students, teachers, administrators, and employers to make sure what we are building is sensible.

Vision

If the entire college-level curriculum was realistically learnable from freely available online resources, what would the world look like?

The resources on their own would be extremely useful, but we see them as a first step towards a larger goal. What we’re really excited about, and what we think will have a very significant impact, is the ecosystem that would grow around these resources:

Community teaching assistants: people offering tutoring services and people creating video walkthroughs of homeworks and problem sets
We are trying to make our learning materials realistically learnable without a TA by providing hints, solutions, Q/A, and more that learners can reference to get unstuck. But sometimes this isn't enough. Having a teaching assistant or tutor who can interact with you synchronously (and one-on-one, in-person) is ideal, and I'm sure we'll see people offering to do that (for money, probably). What I think will be more scalable than one-on-one tutoring is video walkthroughs of all the homework and exam problems. A good example is someone like Self-Taught Physicist, who does walkthroughs of OCW problem sets on YouTube. This kind of thing has seen success in MIT's residential version of 6.036 Introduction to Machine Learning class, which has a video explanation walkthrough for every homework problem, including all its subparts. When I TAed 6.041 at MIT, I would do recorded video walkthroughs of the exam afterwards to help people understand what they got wrong on the test and for future iterations of the class to use as extra reference study material. There might be some care required to make sure the walkthrough is error-free, but I think the community could vet that. One way to think about this is Sal Khan's Schoolhouse World, but for college-level OER content. We also know from David Joyner's distributed classroom that hiring large armies of TAs out of the student population in the previous semester's course offering can work extremely well. Community TAs will make a big difference in helping people learn from OER, but we won't see them until OER content is more comprehensive and high-quality – TAs (unless they are super highly qualified) often need good materials to base their teaching on, and it's also not a good use of their time to help out with a subpar course.


AI tutors: maybe not that soon, but eventually
There has been a lot of hype recently over large-language models, such as the latest Chat-GPT[x]. In the education space, people are interested in their potential to be tutors and provide feedback at scale. For example, Khan Academy is investing into this heavily with their one-on-one AI tutor called Khamingo. Another example is the integration of AI tutors in Harvard's CS50 to eventually approximate a 1:1 student-teacher ratio, as reported by the Harvard Crimson. The research sphere is also making progress – see this paper from Stanford. The impact of AI tutors will be greater if there are more learning resources available for people to learn from (and to train the AI on), and in turn the impact of OER will be greater if good AI tutors exist.


(Paid) exams that provide certification: not a new idea, of course!
This isn't a new idea. EdX, Coursera, MIT's MicroMasters, and many other programs offer this. The main thing that differentiates our OER from MOOCs is that MOOCs substantialy more time, effort, and resources to develop. We're more focused on polishing up courses that already exist, so we can get courses up much faster and at scale. It would take – at the rate things are going so far – more than 50 years for MIT to create MOOCs for most of their main courses, but it would take us only a few years to put up polished versions of existing materials for these main courses.


Alternative ways to demonstrate knowledge: addressing the signaling value of a college degree
A big part of the value of going to college is the degree. The degree has two functions. First, it certifies what you know. Second, it has a signaling value: in pursuing more education, people who know they are more capable signal to potential employers that they are the more capable workers. Recently, we are starting to see more people demonstrating knowledge and capability through alternative paths such as personal projects and internships, especially in fields like computer science and engineering. I think one reason we aren't seeing much innovation in this direction is that people, to a great extent, have to go to college to learn what they want to learn (especially for subjects like the life sciences). If we had large numbers of people learning what they need for a job without going to college, I think we would see more people innovating new ways to demonstrate and certify knowledge. There are many concrete examples of credentials no longer being required, such as Pennsylvania eliminating the requirement of a four-year college degree for the majority of its jobs in the state government, giving people more flexibility for learning without going to college.


Local instantiations: synchronous, in-person classes based on OER
OER is great, but anything virtual may not always be an ideal substitute for a synchronous, in-person experience. Instructors (e.g. at community colleges) could instantiate OER classes in-person in their city to provide the more natural, interactive experience that is important for learning. This could be particularly important for discussion-heavy classes and classes that involve lab equipment. Furthermore, instructors at community colleges (CCs) are often overworked and don't have time to design new classes, so prepackaged curricula could be useful to them (some work would be required to package this in the right way). Another nice thing about instantiations in CCs is that some material has to be licensed and costs money to access. CCs could provide a way for people to pay just a small amount and get access through their institution. One interesting thing about instantiations of OER is that the number of people who can teach someone else's material is much greater than the number of people who can create new learning materials. For example, I'm not smart enough to write a book like Prof. Bertsekas's and Prof. Tsitsiklis's Introduction to Probability, which forms the basis of MIT's 6.041 probability class, but once the book and its associated course materials exist, I can definitely teach the course well because I have a strong grasp on the material. You also see examples of this on YouTube: lots of people make educational videos that draw significantly from other sources, and these people usually learned the content for themselves from these other sources initially.

Frequently asked questions

  1. We thought MOOCs would solve the problems in education, but they didn’t. Will OER have the same issue?

    I think this is a valid point. But I don’t think MOOCs failed. It was naive to assume they would “solve” education: if you actually went and talked to the people that the leaders of the MOOC movement thought they would “save”, this was clear. Now we know that we should view MOOCs as an experiment. We learned many things: one of the highlighted insights was that the people who end up succeeding in MOOCs are those who already had a pretty good K-12 education and circumstances that endowed them with the ability to succeed in MOOCs. The problems in education are systemic. We should not expect any single endeavor to fix them.

    I don’t think we’ve made enough progress in OER to have properly tested what difference they could make – we haven’t given them a real chance yet. For example, OCW has a lot of courses, but most of them are outdated, most of them are not realistically learnable, and the course offerings represent some smattering of random courses – there’s no sequence of complete courses equivalent to a major or minor from the undergraduate curriculum. Despite this, there’s still so much evidence that OCW makes a difference.

    We would like to solve all the problems in education, but it’s naive to think we will. The right thing to do is to make an informed decision about what things to experiment, to try them, and then to iterate and move closer and closer to the right set of solutions. OER is a necessary but not sufficient condition to “solving” education.

  2. Are professors at universities even allowed to put their teaching materials online under an open license?

    At many universities, including MIT, the professors own their teaching materials. So, as long as they are willing to put their materials up, we are good to go. But why would the university be ok with all its course materials being freely available online? If that happened, why would anyone pay tuition to enroll in the university? Would the university change its policy for who owns course materials?

    From the many professors and administrators I’ve talked to at MIT, I gather that MIT, at least, doesn’t seem to be concerned about this. In fact, MIT seems to be trying hard to push open learning forward, as you might infer from its many open learning endeavors, from its OER endeavors like OCW to its certificate-granting MicroMasters programs. The real value of being at MIT comes from becoming friends with extraordinary people, access to research opportunities, close contact with world-class professors – things that won’t be devalued by MIT’s core learning resources being freely accessible. Furthermore, enrolling as a tuition-paying student awards you a degree, which, at least at the moment, is a necessary signaling mechanism in today’s job market. It is likely that smaller, less-known universities would be threatened by this kind of thing and would not participate – and that’s ok. We don’t need every university to participate, just a few. We could also imagine a future where community and state colleges are mandated to provide all their learning resources as OER.

  3. The college experience teaches you so much more than just what’s on a syllabus. Is OER enough?

    The college experience differs so much depending on where you go to college and what you make of it. OER won’t be a perfect substitute for college but the ecosystem built around OER might. In any case, people will need many different kinds of experiences to learn all the sorts of things they want to learn. Some things you can’t learn well in college and they are better learned through a job or some other experience. But there’s a reason why a lot of fundamental things are taught through college courses – it’s a good way to learn the material.