Home
About ADUni
Courses
People
Colloquia

Introduction

In 2000-2001, 34 talented and motivated college graduates attended a one-year intensive complete undergraduate computer science program for free. The program was an experiment in curriculum design, free education, and the effect of the Internet on the future of education. ArsDigita University (ADU) was the brainchild of entrepreneur, Philip Greenspun, and the ArsDigita Foundation.

The concept of ADU was to offer bright and motivated college graduates an undergraduate level major in computer science in one year with minimal interruption to their lives. Normally, these people's options would be limited. Most universities either do not allow college graduates to enroll in an undergraduate degree program, and taking the courses a-la-carte would take four years or more. Master's degree programs are not an option because a more extensive preparation is required. ADU offered a quality (MIT caliber) undergraduate computer science education, free of charge, preparing its graduates for jobs in the industry or graduate school in computer science. Most important is that ADU was free. We advertised "free education for the bright and motivated."

Admissions

Enrollment was restricted to students holding at least a Bachelor's degree or the equivalent in any field outside of computer science. The admissions requirements were: 'a confidence-inspiring transcript', SAT scores greater than 1350, and a persuasive essay explaining reasons for wanting to join. Flyers were sent to more than a hundred top schools all over the country, articles appeared in newspapers, and a posting appeared on http://slashdot.org. There were over 350 applications for 35 slots. Each application was evaluated by three people who decided to phone-interview a subset of 50 candidates. Each interview was conducted by two people and consisted of two parts. One person conducted the interview while the other recorded details, registered impressions, and assisted with a smooth conversation. The first part of each interview was to confirm and discuss motivation for the program. The second and more important part was a puzzle given to the applicant, which he/she was instructed to try and solve on the spot. The applicant was told to work with the interviewers as a team to solve the puzzle. We used many different puzzles, but a typical example is the well-known bank-vault puzzle (a distributed systems problem). In this puzzle, n people are asked to agree on a protocol, such that all of them act in unison to turn their keys to open a bank vault. The protocol must assume that the people are standing shoulder to shoulder in a line, and each person can communicate only with the people to his left and right, but all have common access to a light blinking at regular intervals.

The idea was not to test how fast or whether someone was able to solve the puzzle, but instead to see how the person engaged in problem solving, especially in a team setting. We looked for assertiveness, organization, cleverness, communication skills, logic, and persistence. No matter how far someone got in the problem, there was always a new angle to the puzzle that was harder. For example, in the bank vault puzzle, we begin with five people in a line, move to n people, restrict the amount of memory for each person, allow a different communication topology etc. No one ever solved all the variations; it was not expected that they would. Applicants were told that they were welcome to email in their ideas, if something came to mind after the interview was over. Almost every candidate chose to do this. Out of 50 interviewed, 40 candidates were accepted and 36 chose to attend.

By the end of the year, two students had dropped out, seven finished the program without completing all of the requirements, 27 completed all of the program requirements, seven of these with honors.

Students

The graduates of ADU included lawyers, writers, speakers, composers, poets, dancers, teachers, stage managers, sys-admins, engineers, technicians, physicians, musicians, statisticians, psychologists, anthropologists, social workers, and a spy. Their ages ranged from 22 through 68, with the median age 30, and mean 33. Students came from all over the United States, and a few from Europe. There was a great range of experience with computers, philosophies, and personalities. There was no correlation between how successful a student was, and their age, background, or gender. The two people who dropped out did so for personal reasons.

Curriculum and Schedule

The curriculum at ADU was linear, with one course per month, each month corresponding more or less to a typical semester course. Every student took every course in the same order at the same time. ADU opened officially each day at 9AM and closed at 9PM, although each student had 24-hour access and the lab was often full late into the night and sometimes into the early morning. A typical day at ADU started at 9:30 AM with a 100-minute lecture. Then came a break for lunch followed by an early afternoon recitation session at 1:30 PM lasting another 100 minutes. Recitation varied from informal problem review sessions, to presentation of new supplementary materials, to discussion groups, depending on the course and the lectures. The afternoons and evenings were used for problem sets, with teaching assistants available at all times for help. Saturdays were off and Sundays were used for exams, review, or supplementary lectures. There was a great deal of collaborative work. This was encouraged by the staff and was effective in making the best use of the compressed schedule.

Unlike a typical undergraduate model, where the course runs over a long time period but students do not see each other often. We had a model where the courses run for short periods of time, but the students saw each other at least 10 hours every day. It was efficient to allow natural interaction and collaboration just as one might have at the workplace, allowing students to leverage the expertise of their colleagues thereby increasing productivity.

A typical 4-credit semester month had 40 hours of lectures and 40 hours of recitation contact time. This is comparable to the total amounts in a course spread out over a semester. Some courses were run for one, two, or three weeks and were equivalent to one, two and three semester credits.

All lectures and recitations were videotaped and are available free to the public under an open source license at aduni.org. The problem sets, solutions, exams, and lecture notes are also available, as are student evaluations and suggestions for every course.

The curriculum at ADU consisted exclusively of computer science courses and related mathematics. Some courses were selected directly from successful self-contained courses at MIT, and some were designed from scratch specifically for the program. The overall plan is roughly in line with the ACM 2001 guidelines, modulo the unusual time constraints. A detailed list of courses with brief descriptions and requirements is shown below. The following terminology is used. An assignment is a homework that can usually be done in four hours, and should be completed that day. A problem set is a homework that is more involved, requires mild research or planning, and is expected to take 20 hours of work. A project is a homework that involved major planning, revision, execution and documentation. A typical project takes 80 hours and runs through the month.

Course Descriptions

Summary

The curriculum was organized so that the courses alternated between programming and mathematics. This feature worked well, allowing students to exercise and practice different skill groups. There were two one-week breaks in months three and six, respectively, that were crucial for rest. There were 39 total weeks of courses (equivalent to 39 semester credits) distributed into the following categories:

Mathematics: 10
Software: 13
Hardware: 3
Systems: 6
Theory: 7

Because every person took the same courses in the same order and at the same time, there were opportunities to build connections between the courses. For example, projects built in the Java software engineering course could be expanded on in Algorithms. Database projects could consist of optimizing previously designed web sites. Discussion of recurrence equations in discrete math could build on experience with recursion in Scheme. Control over the prerequisites affords greater flexibility in the design of a course. Hence the sequence of courses was more unified than otherwise possible in a regular college.

Physical Space

The physical space at ADU was crucial to its success. It created an environment in which the social and academic center of the students' (and staff's) lives was in the lab, rather than the dorm, pub, or library as it might be in a more traditional school. This not only encouraged collaboration, but also allowed for more flexible pedagogy. The contrast with a typical college where classmates see other twice a week for 90 minutes and perhaps at an occasional arranged study-date, was significant. For example, at ADU we never gave an exam in the lecture hall. Each student simply worked at his or her own workstation in the lab. On-line access was available during lab and exams, without any special arrangements. The distinctions between exam, assignment, problem set and project, sometimes got blurry, for better or worse. Ad-hoc group review sessions were easily organized by the teaching staff and by students for themselves.

ADU had three separate spaces: a lab, a classroom, and an administration area. The lab was a large open room with forty stations. Each station was equipped with a 3' by 6' table, a small file cabinet, and a Linux workstation with Internet access. Each student had his own textbooks and space. There were a few tables up in front of the room reserved for teaching assistants and bookshelves for a small library. In the center of the room there was a movie theatre style popcorn machine, a water cooler and a portable white board. The lab was the nerve center of ADU. It was filled with students from 9-9, and was rarely empty at any time of the day or night. For many students it was their home for the year, and they left only to sleep.

The classroom was across the hall from the lab in a wide room with three rows of fifteen chairs each. There was a blackboard, a video screen, and a videotaping station equipped with a camera and computer. There was a dual-boot Windows/Linux machine with Internet access that could be projected on the screen. Lecturers used a combination of media depending on the topic being presented, and their personal preferences. Across from the lecture room was a small separate conference room used for small tutorials, study sessions or review. Lectures were always delivered to all 35 students in the main lecture room. Recitations were either done the same way, or the class was split into smaller sections, using the lab, lecture, and the small conference room.

The administration area was between the lab and the classroom. The faculty and administration offices, and a communal kitchen were housed here. This area was open and divided into office cubicles, except for one separate walled office for private meetings and conferences.

Staffing

ADU had one full-time faculty member and three full-time teaching assistants (TAs). Many part-time faculty and teaching assistants were hired from MIT, and nearby high tech companies. The full-time TAs had minimum of a Masters' degrees in Mathematics or CS, and all faculty members were Ph.D holders or candidates in CS. In addition to serving as a resource for students, the TAs assisted with, grading, tutorials, problem set design, recitation sections, and website maintenance administration. Each course had one lead faculty member in charge of lectures, lecture notes, problem sets and overall course design. He/she supervised the recitation leader and TA's providing appropriate material and suggestions. The plan initially was to expand the fulltime faculty to three and avoid part-timers. This would have been a good idea, as finding quality part-time people who are willing to spend the intense time demanded of the job was difficult and time consuming. The program could be run with three full-time faculty and three full-time TA's. Depending on the day, some TA's worked late morning through mid-afternoon, and some worked early-afternoon, through late evening.

The amount of effort and commitment of the part-time faculty varied greatly, and was proportional to the quality of the course. The more time taken to design a course specifically for this time frame and for these students, the better the course was. If a course was cut and pasted from another school and molded to fit, it did not succeed as well.

There was a full-time administrator who was assisted by two volunteers to handle everything about the day-to-day logistics, including guest speakers, admissions, publicity, web postings, payroll, events, and evaluations.

Colloquium

A nice feature of ADU was its monthly colloquium. Every month we invited a prominent speaker from industry, academia, or other parts of the greater computer science community to speak to our students. Our location and mission allowed us to recruit some superb speakers, including Gerry Sussman, Richard Stallman, and Mike Sipser. The colloquia were all recorded and are available for free on the http://aduni.org web site. A standard university CS department would never have the time, interest or resources to round up a consistent set of speakers for an undergraduate audience.

Evaluation

The ADU idea has some clear benefits over the standard way to teach computer science. The important question is whether one year is enough for a student to absorb such a magnitude of material. Did the students get what they came for? What will they be doing in the next five years? I believe these questions are linked. Can we can send students to graduate school, and have them succeed? Can we send students to the work force, and have them produce? That is what will determine whether this idea really works. Interested readers can learn about the current status of ADU's first class at aduni.org.

Most of the students felt that the program was delivered as promised. However, it might be able to work better. Future consideration needs to go to tweaking the curriculum slightly, and challenging the broader assumptions. Should we have courses run in parallel? Should the courses run for longer? Should there be electives? Evaluation and critique of these topics by the students can be found on the web site.

Distance Learning

Originally, ADU was intended to be a distance-learning program as well as on-site. Questions of distance learning always came up when discussing ADU, especially because our materials are available for free on the Internet. ADU, however, was primarily an on-site concept. There was no ADU without the day-to-day physical interactions, the lab, and the personal contact. One story brings this point home.

One of our best students, a physician by training, and a brilliant yet modest man, had to leave Cambridge after 2.5 months for personal reasons. He had the year free, and decided to continue the program remotely using our web site and email. I told him that would be acceptable, and furthermore that it would be a great way to test the feasibility of simulating the program virtually by distance learning. I asked him to keep a diary. This was a person who worked well on his own, who already had made friends with many class members, and who was as bright and motivated as any of his classmates. After three weeks, he stopped sending in problem sets and emails. He apologized and said that he could not maintain the concentration, or sustain the effort required, to succeed in the program without being surrounded by the environment. Experiences such as this, made the administration downplay the distance-learning goal.

On the other hand, many people have made use of our videos and other materials provided on-line, if only internally and without direct interaction with other students and staff. Our bandwidth load for the video server averages 12G per day. Moreover, we have received many fan-mails from such people and some requests to mirror our materials.

ADU was a social experiment as much as an academic experiment. Does society have a responsibility to offer education to people who otherwise cannot afford the time or money? Do the economic realities of the software job market call for the industry to sponsor future such programs? These are hard questions that we do not address.

Future Plans: ADUni as Pilot Study

ADU can be used as a pilot study for an ambitious but exciting model of education in computer science. This new model would address the need for more qualified technical people in the workforce as well as the shortage of graduate students in computer science. It will efficiently tap into the reservoir of talent that currently dream of computer science as a career option, but see only vocational programs offered, or multi-year commitments.

The budget for ADU with 35 students was about 1 million dollars. It is hard to justify this expense on a per student basis no matter how valuable the experience was to each of them. There is, however, a great pool of untapped talent that can be used to fill programming positions, as well as future Ph.D programs. How do we reach this market efficiently?

An Independent National Post-Bacc School of Computer Science

I have a vision for an independent national program that can leverage our first year's start-up effort and experience, and be more efficient economically. Some of these ideas came out of a discussion with Ellen Spertus of Mills College. The model would be a national post-bacc undergraduate CS program accredited by CSAB/ABET modeled after the ADU pilot year. Students would apply and list geographical preferences. Universities and colleges all over the country sponsor from 6-20 students.

The students do not receive a degree from the sponsoring institutions. The sponsoring institutions provide full-time lab space and administrative support. Professors/Professionals mentors sign up to mentor particular courses whose materials are provided to them from our pilot study year. The mentors do not necessarily come from the sponsored institution. The program coordinates a match between mentors and courses, for each location. The hosting institution provides from their own staff, a fulltime mentor and a teaching assistant, who work with the students all year and provide continuity. A central web server (aduni.org) is used for coordinating a cohesive structure among the satellite sites, and for on-line help and information sharing. Two semi-annual events are scheduled for the students to present projects and share experiences.

It is a dream, but the educational details have been designed and tested. The pilot year's graduating class will provide the reputation. All it needs is manpower, money, organization, and support from the computer science community.

More than any educational effort, it can bridge the communication gap that arises between academics and industry. It can open computer science up to the largest untapped market in the world - bright professionals with a new career dream.

Site last updated: 10 August 2001
Comments: webmaster@aduni.org