The Robustness Principle

Another week, another book read. This time it was The Seventh Sense by Joshua Cooper Ramo. The premise of the book is that we are at the edge of a new revolution, governed by understanding networks and navigating complex systems, and that we need to develop a seventh sense to survive or thrive. I found the first half of the book gimmicky-sounding, like one of those self-help books that promises you the secret in life to get ahead. This led to skimming over certain paths. Things get more interesting in the second half when the author gets into detailed examples of technology. I learned a lot about internet protocols, computer viruses, and the “disappearing problem” in artificial intelligence (AI) research.

There are many vignettes in the book, but the one I want to discuss today is by Jon Postel with regard to transmission control protocol (TCP): “Be conservative in what you do (or send), be liberal in what you accept.” Since I’ve been reading and thinking about robustness and its connection with higher education, I’m going to try and push this Postel’s rule a little further and see how it might work.

If you read higher education news regularly, there has been a fair bit of chatter about how to break the “iron triangle” of higher education: Access, Cost, Quality. The dilemma is that you can choose to do two well at the cost of the third. (Cooper Ramo has a similar triangle for networks – you can read his book if you’re interested to find out more.) Do you want wide access at low cost? Quality will suffer. How about high quality and wide access? That will cost more than a fortune. Could you retain high quality at lower cost? Possibly, but this will only be available to a small segment – and who will choose the beneficiaries?

The U.S. system of education is modular in three ways: (1) there are a large range of institutional types that cater to different segments of the population, (2) courses are arranged in modular units allowing for some mix-and-match catering to different goals/students within an institution, (3) the modularity allows for mid-stream changes both intra- and inter-institution. (There are some very specialized institutions attached to specific career goals but we won’t consider these for now.) At many colleges and universities, you do not have to declare a major for admission. This creates some messiness for administrative planning; you might say it is less efficient. However, the advantage is flexibility – robustness perhaps.

Since chemistry is what I am most familiar with, I will use that as an example. Note that the chemistry curriculum tends to be more hierarchical compared to many other areas so there will be significant differences, but the general principles may still apply. General Chemistry, the first-year college sequence (in two semesters or three quarters) has the same core content. Textbooks have likely played an important role in the “standardization” of this curriculum. Different departments and instructors will have different emphases. They may have different lab experiments, but the core goals are similar across programs. This means that it is relatively easy to transfer credit across institutions. Our department is liberal in what it accepts coming in, while the program we deliver in General Chemistry is conservative and does not change much from year-to-year. Something similar takes place in Organic Chemistry in the second year, and Physical Chemistry in the third or fourth year.

There are elective classes that may vary a fair bit from one institution to another depending on whether they are offered as a one-semester or two-semester sequence (ditto for quarters): analytical chemistry, inorganic chemistry, biochemistry, advanced organic chemistry, interdisciplinary courses (bioorganic, organometallic, biophysical, physical-organic) and more specialized topics (instrumental analysis, computational chemistry, electrochemistry, materials, polymers, etc.) Different institutions may have different program requirements, but we will accept transfer credit that count towards the major, i.e., there will be X number of elective units in chemistry that constitute the requirements. Our program, like many others in the U.S., is accredited by the American Chemical Society (ACS). The ACS has certain core requirements that must be fulfilled but there is also some flexibility in how you fulfill them. The system, I would say, is overall robust in chemistry.

Being in a liberal arts college at a private institution, we are costly. I’d like to think we are high-quality – at least our students who go on to graduate school tell us that we prepared them very well. As to access, I don’t know of any online programs in the undergraduate college that may provide wider opportunities. Our department has a number of students who transfer in. While we accept their credits, many of them struggle through the program suggesting that there is some mismatch. Some have no problem adapting, and others take a semester or two before they find their footing. This is one way to reduce the overall cost to the student (in terms of tuition). There are financial aid packages, but those are limited and therefore access is constrained.

Is there a robust solution to the iron triangle? I’m not sure, and I don’t think the solution lies in selective liberal arts college as an archetype. In combination with other institutions and initiatives, perhaps. The larger, state public institutions may be the place to break the iron triangle – but they are beset by their own problems, not least of which is diminishing state funding. Can the programs be transformed to run “lean” without sacrificing quality? Difficult to say. I’ve been mulling over whether an online-hybrid of sorts can handle some of the basics encountered in General and Organic chemistry. Certainly adaptive learning platforms claim some success in this area. But a high quality platform is not cheap, even if it may provide widespread access. Would a college institution host and maintain such a platform at very low cost to students (as opposed to a for-profit ed-tech provider)? I don’t know.

The problem with complex technological systems as they grow (inevitable according to Cooper Ramo and Ellul) is that we humans understand less and less, and that blurs the meaning of robustness. When AIs are tasked with making determinations, what do robustness rules look like? It would be scary to aim purely for efficiency. Biologically, at least, that leads to a loss of robustness such that could result in extinction from significant environment change. Then again, we human-cyborgs and our machine counterparts seem to be the most successful so far in altering the environment to suit our needs, at least in the short-term. Will our robustness be measured in the hundreds of years since the industrial revolution? It is a mere blip in the eons of time.