My summer paranoia: computers will replace teachers in higher education

I had a strange dream last night. I was in discussion with an editor from a publishing company about the draft of a new book I was writing. As in all dreams it wasn’t clear what was going on in the discussion but then I realised he wanted me to change what I was writing to make the case that computers can replace teachers in higher education. He told me that his CEO and a number of CEOs from other companies all thought this was the right way to go, and were trying to influence the market to accept this. I was so upset that I woke with a start.

Now you could say I shouldn’t drink tequila before going to bed, but this dream was not at all unrealistic in the light of events over the summer.

No, they really ARE trying to get you

Let’s start with xMOOCs and automated marking and peer review to get around the awkward point that one instructor cannot provide adequate feedback to thousands of students. No problem: a combination of big data collection and analysis and multiple-choice testing will solve most problems, and the ones that it won’t solve will be solved by dumb students marking less dumb students.

Then there’s the Republican Party of Texas whose election platform contains the following (p. 12):

Knowledge-Based Education – We oppose the teaching of Higher Order Thinking Skills (HOTS) (values clarification), critical thinking skills and similar programs that are simply a relabeling of Outcome-Based Education (OBE) (mastery learning) which focus on behavior modification, and have the purpose of challenging the student’s fixed beliefs and undermining parental authority.

(It also reads as if they oppose clarity in sentence construction, but that’s another matter.) However, if you can’t teach critical thinking skills through automated computer-based teaching, get rid of the requirement for critical thinking. Brilliant!

This was followed by the California State University system deciding to outsource online learning to a commercial publisher.

Then it turns out that the California two year college system has undergone nearly $1 billion of cuts since 2008, resulting in a waiting list of 470,000 students who cannot get into classes. Talk about creating demand for automated courses.

Still in California (do they have too much sun there?) Stanford University has just created a new Vice-Provost for Online Learning, who turns out to be a computer scientist (as are all the people heading up Coursera and Harvard/MIT’s EdX). Who needs someone with expertise in teaching for positions like this?

Lastly, in Canuckistan, the Ontario government is looking for more ‘productivity’ from the post-secondary institutions, and is asking how online learning can lead to improved productivity. In this case, that’s a good question; it’s the answers people may come up with that scare me.

Do we really need teachers in post-secondary education?

At least these developments are forcing an examination of something that most of us have taken for granted – so let’s examine it.

Here we need to look carefully at the language we use. One thing that struck me when I emigrated to North America nearly 25 years ago was that in Britain, those that delivered teaching in universities and colleges were called lecturers or professors. In North America they are called instructors. Obviously, lecturers lecture, professors profess and instructors instruct. But we talk about university or college ‘teaching’.

Now having a background in primary school (k-7) teaching, I always associated ‘instruction’ with a didactic form of teaching, where the instructor determines what content will be learned, delivers information, and the student gobbles it up and is tested on how well he or she has ‘understood’ and ‘remembered.’ This may be a fair description of a lot of post-secondary education, but in my mind it isn’t and never was ‘teaching’. To develop critical thinking skills, professors did more than lecture: they discussed and talked with students, marked assignments and gave detailed feedback. The tried to help learners learn. In other words, they taught.

Yes, this was possible 50 years ago because we had an elite system, and few students per professor. Now we have, especially in North American Tier 1 public research universities, very large classes and many students per professor (made worse in first and second year by tenured professors focusing mainly on research and graduate education). So we have fallen back almost completely on ‘instruction’ rather than on ‘teaching’ in undergraduate programs. But, given the demands of a knowledge-based society, what we need is less instruction, more teaching, and in particular a different quality of student learning.

Instruction is easier to automate than teaching

If the focus is on a didactic model of instruction, then it does become easier to automate. Choose the ‘best’, i.e. most knowledgeable, professor in the field, record their lectures, and set multiple-choice computer tests and assessment, with automated feedback, and bingo, you can teach millions with the same teacher. If you knew nothing about education, like the Republicans in Texas, this would be an ideal way of avoiding grown-up things like paying taxes or tuition fees. As they say, if you can be replaced by a computer, you should be.

However, whether this will produce graduates with the skills and knowledge needed in a knowledge-based society is another matter.

What computers find difficult in teaching

Let’s define productivity in educational terms: it’s achieving the same or better quality outcome in learning at less cost. By definition, it means doing things differently (sometimes called innovation). If we want better quality outcomes then we also need to define outcomes. I put high value on, yes, critical thinking skills, evidence-based decision-making, independent thinking, self-management of learning, responsibility, and ethics.

To successfully achieve such learning outcomes, learners either have to be incredibly self-motivated and already highly knowledgeable (i.e. already well educated), or they need an environment that supports the development not just of these outcomes, but also the development of their thinking and decision-making. This requires fostering or supporting their motivation to learn, dealing with gaps in knowledge or lack of learning skills, providing timely feedback, and above all providing guidance, criteria and direction to ensure that they meet the necessary standards to operate effectively in the real world. This is what I call teaching, and much if not all of it is difficult or impossible to automate.

It should be remembered that a very behaviourist form of computer-based learning, called ‘programmed learning’, existed before the Internet. For over 30 years computer scientists, working in the field of artificial intelligence, have been trying since then to improve on ‘programmed learning’, which failed to deal adequately with the development of cognitive thinking skills beyond the level of comprehension and memory. Despite substantial research investment over many years, AI has proved to be less than successful in the teaching domain. It is very hard to replicate the complexity of a skilled teacher who has to deal with many different variables and factors in real time. It is not just about processing speed and data management, but also about building relationships with students, making intuitive judgments, and being able to handle qualitative issues such as beliefs, values, and the personal feelings of students. Computers are not good at this (and nor are many ‘instructors’, to be fair.)

I first became excited in online learning in the late 1980s, when the Internet was being developed, because it offered the possibility of communication at a distance, a problem that was not well managed in distance education at that time. Thus the computer was not replacing the teacher, but providing teachers with tools they could use to teach even students who were remote. In the early days of the Internet, online teaching was called deliberately ‘computer-mediated communication’ or CMC, to distinguish it from programmed learning. Now the computer scientists in xMOOCs are trying to drive us back to programmed learning.

Online learning can improve productivity, but not through automation (and it ain’t going to be easy)

It’s no good cutting costs if you don’t get the desired outcome. The best example is the construction industry. Cheap construction often has expensive consequences. So we have to be sure that if we are seeking increased productivity, the outcomes are at least as good, if not better, than before the intervention.

The main ways that online learning can improve productivity are as follows:

• using online learning, rather than building new campuses or physical facilities, to expand access. This option has a fairly narrow range. There has to be enough existing facilities already in place, and the increase in enrollments will need to be accompanied by changes in delivery, with a move to hybrid or fully distance learning in some courses or programs to free up facilities for the new students. This requires some fairly sophisticated juggling of classes, redesigning courses, and careful planning for it to work. It also assumes that numbers in post-secondary education need to increase;
• use of shared materials, and not just open educational resources, but developing courses or programs that can be used across several institutions. However, course development constitutes only about 15-20% of online teaching costs. The big cost remains delivery.
• economies in the delivery of programs by maintaining content quality through the use of tenured or research professors for the design and development and monitoring of course delivery, but reducing delivery costs through the use of well-trained online adjunct professors and automated marking where appropriate. Again, this requires some investment, particularly in training, that will to some extent offset the lower cost of adjuncts
• course designs that move the work away from the instructor to the student. Examples are collaborative learning, development of self-management learning skills, problem-based learning. Even here, though, instructor presence is essential for success.

All these activities to increase productivity without losing quality require careful planning and quite a lot of training. We are not helped by a total lack of research into the costs of various modes of delivery, which means we have poor or no data as a foundation for such changes. Nevertheless, redesign of teaching and a strategic approach to online learning could lead to savings of up to 10% on the present system without a loss of quality. But there’s no silver bullet if we are to get the kinds of graduates we need in a knowledge-based society, and good teachers (as distinct from instructors) will remain critical for success.

And with the return of real students and real ‘instructors’ next week (at least in Canada), maybe reality will return and my nightmares will go away.

Questions

1. Anyone else share my paranoia? (about computers replacing computers – anything else, see a doctor)

2. Do you believe we should replace teachers (or instructors) with computers? What are your reasons?

3. Can online learning improve productivity in post-secondary education without getting rid of most instructors?

4. Can you recommend a good doctor?