Rambles around computer science

Diverting trains of thought, wasting precious time

Tue, 29 Aug 2023

Teaching loads in computer science: it's not just lectures

About four years ago, when I had spent a year at the University of Kent in my first job as a lecturer (or “Assistant Professor” for North American-style readers), I had to fill in a probation review form. One question asked me to reflect on my teaching load over the previous year, which had consisted of 1.4 courses or “modules”. One module lasts for a 12-week teaching term, with one reading week, and students typically take four modules at a time. I said that in my view, 1.4 modules had been a heavy load for a new lecturer trying also to do research at the same time. I still think I was correct to say so.

My head of department's response was unsympathetic. “You only had to give thirty lectures. Normally it'd be thirty-three!” This made me unhappy, partly on a personal level—why ask for a reflection if you already know the “right answer”?—but also because counting lectures is “obviously” not a good measure of load. Since then, my move from Kent to King's has provided another study in this.

(It's been a couple of years since I wrote about my then-impending move. I realise I haven't yet written anything directly about my teaching experiences at King's. I'll get into that a little in this piece, but not a lot.)

To set the record straight, here I'll enumerate some other dimensions by which teaching loads differ. These might be useful for those weighing up job offers: it's worth getting a feel for how the institutions compare on as many of the dimensions as you can.

All this deserves some caveats. Of course there will be plenty of other ways in which institutions differ. Also, the loads at any given department are a moving target... I'm told my department at Kent has managed to reduce loads significantly since I left, while King's had been reducing them significantly immediately prior to my joining. At King's right now, the load is generally favourable compared to my time at Kent, both on a lecture-count basis and on many but not all of the other factors. However, there is one big exception, bringing me to the first of my many dimensions....

Number of students, as reflected in assessment and marking

This is the obvious one but it needs saying. All else being equal, marking is usually linear in the number of students. All else is never equal, however. Only slightly less obviously....

Amount and manner of assessment, marked by whom

If there is less coursework, there is less marking. If exams are done electronically and can be marked wholly or partly automatically, there is less marking (but this comes with downsides). If teaching assistants (TAs) can mark coursework, there is less marking for the lecturer, provided that directing the TAs and moderating the results is not too much work... this can be significant. All this varies hugely between institutions (and indeed departments).

Use of coursework

Although this is a special case of the previous one, it bears repeating that departments and programmes vary hugely in both how and how much they use coursework for assessment. The weighting of coursework can be a red herring load-wise, since just because something has a low weighting doesn't mean it's a small piece of work. Bigger coursework means more marking, and often disproportionately since coursework is usually harder to mark than exams. Also, whereas exam-marking happens in a consolidated, frenzied period during which most other business is paused and external commitments can easily be blocked out, coursework comes in amid other stuff going on.

Complex rubrics and submission schedules can also ramp up the work. For example, at Kent, when marking an undergraduate or master's project, the submission consists of not only a dissertation, but a corpus (code and/or data), one or more individual reports (many in the case of a group project), and various progress reports along the way. As someone with very limited desk- and head-space, I found that this plethora of materials made the process more time-consuming than it needed to be.

Student level

Again this is “obvious”, but a preponderance of relatively weak students, who therefore need a lot of help, can become a major ingredient of load. At the other extreme, the most able students can be needy in their way, though it's a nice problem to have. In theory, the course should adapt to stretch students “just enough” whatever their level, but this falls down when the spread of abilities is large.

Absolute class size matters too, because when the number of students is large enough, the spread of abilities is always large, almost independent of entry requirements. Put differently: if the invariant is that 10% of admitted students slip through the “equippedness net” that entry or progression requirements were supposed to take care of, it matters whether that 10% adds up to 4 students or 40. Critical mass effects mean the latter can draw much more attention, not just in absolute terms but also per capita.

The amount of direct contact with students

If you have few or no TAs, you as lecturer are the TA. You are the only human interface the student has with the course. This has consequences far beyond marking. There were few or no TAs on all my courses at Kent, although some other courses did have decent provision; it often reflected the age of the module spec, with higher levels of provision a vestige of bygone, less penny-pinching days. At King's there generally is decent TA provision, but scale can make this hard to manage, the TA bag is very mixed (many are themselves undergraduates; some very good, in fairness, but not all), and there are also rumours that the money might get taken away. Some days at Kent I spent hours answering student e-mails or dealing with spontaneous drop-ins. This brings me to....

Cultural distantness of the lecturer

Institutional culture affects the threshold that students feel they must cross before approaching a lecturer. It is related to the TA issue, in that students might approach TAs first (only if they have them!). But it varies even aside from that. If students are told there's an open-door policy and they can drop in to their lecturers' offices whenever (as was the case at Kent), they will do so! Conversely, if they absorb a sense that lecturers are busy and distinguished people who have limited time for them, they will be more conservative. (It also helps if, as at Cambridge, lecturers are kept behind a locked door....) I'm not saying which of these is “better”—obviously, access to lecturers can be a boon to students, but it comes at a price.

Student dependence and learned helplessness

Even in addition to the combination of student level, TA provision and lecturer distantness-or-not, a department's culture can make students more or less self-reliant. At Kent there was an oft-remarked vicious circle of expectations around hand-holding and “being nice”, which amplified both the office drop-in tendency and frequent generalised demands for “help” of unspecified nature. At King's there are similar effects in evidence. It continues to be a bit mysterious to me how exactly these expectations came about and what might lessen them... I suspect the style of first-year teaching has an outsized influence, but it can't be just that.

Late submission, resubmission, resitting

A disproportionate marking-related load comes not from exam scripts or coursework submissions per se, but from handling those scripts and submissions that do not arrive or do not pass. When departments have more generous policies around late submissions and extensions, this causes greater fallout onto academics, who are forced either to mark in multiple batches, or to work with a narrower marking window overall. Reassessment multiplies the number of assessments to be written, and adds to the number of context switches (getting into marking mode and out of it).

In Cambridge—the most pared-down, research-focused regime I know—these simply don't happen, at least for computer science students (and, er, with the caveat that my knowledge is possibly out-of-date). If you miss a deadline, the late-submission penalty is severe and you don't get another chance. There is only one exam period a year, and if you're not fit to sit your exams, you either get a special result that year (“deemed to deserve honours” a.k.a. DDH) or you fail, depending on whether your term-time work has been up to scratch. At other universities I've worked at, any module's assessment is done at least twice per year, with all the doubling that entails.

Per-student casework, as it falls on the academic

Things like submission, marking, extensions, mitigating circumstances, release of marks, appeals and so on can generate a surprising amount of student-by-student “casework”. Mainly this is when students end up in odd or difficult situations. In addition to the inherent task of marking resit papers or what-have-you, there is the added task of regularising the student's situation, for example by obtaining exemptions (or advising the student how they can do this), or simply by forwarding information that has got lost owing to the student's unusual path. Academics invariably do get pulled into this, to a varying extent; the lurking variables, aside from the number of students, are about processes themselves (which can be heavy or light) but also about how they are administered and communicated.

If support staff are adequately resourced, academically literate, communicative and procedurally empowered to solve problems by themselves, they can take the casework off the academic. If information such as internal web pages communicates effectively with students how they can help themselves, demand can be dissipated at source. However, when these are not the case, academics get sucked in, under either our lecturer or tutorial hats. We can end up as a hapless e-mail go-between, forwarding student queries to the admin office, unable to help but also unable to extricate ourselves. If the bureaucratic machine appears to be stonewalling a student, perhaps from underresourcing and/or lacking the right systems (e.g. a ticketing system for student requests), often they will re-approach the academic rather than wait for their case to reach the front of the invisible queue. This generates more requests on the system in general and more tasks for the academic; last summer it was a major problem for me at King's.

(I can't fault the student's tactics here: an academic advocate can help cut through bureaucracy and sometimes prevent unfair outcomes. Nevertheless, this is a high-cost way for the institution to enable this outcome; a more empowered and better-trained administration is a better option.)

Misconduct levels and processes

Misconduct cases are a particular kind of casework, but worth covering specially. A Kent colleague used to bemoan that “we incentivise cheating” because the penalty is a mark of zero if caught. So the expected return on a plagiarised submission is positive academic credit. At King's, the incentives are not too different, although perhaps slightly less bonkers because repeat offences are (I'm told) treated especially seriously. I'll believe that when I see it. In any case, reporting and seeing through misconduct cases places a large burden on the academic; the incentives are either to let it slide or to design assessments so that cheating is undetectable (note: not impossible! the former is feasible but the latter usually isn't).

It seems universal across (at least) CS departments that the cost of dealing with increased misconduct levels has not been adequately “priced in” as student numbers have been ramped up. On top of that, the greater use of automated marking, to deal with greater numbers of students, has I suspect allowed a lot of undetected collusion (and worse). Pandemic-related changes have also increased these tendencies, through greater use of online assessments (open-book exams are an especially terrible idea) and through general depersonalisation. I hypothesise that the tendency to depersonalisation—particularly strong at King's since in Informatics we no longer give in-person lectures—has led some students to fall more into a “system-gaming” mindset, rather than seeing their degree as the matter of personal development that it is supposed to be.

Institutional resistance to academic-as-admin

Beyond student casework, various non-academic tasks can be intentionally pressed upon academics. These include submitting attendance data, mechanically diddling marksheets, physically sorting exam papers, signing paperwork for this or that, and so on. How much this happens is a cultural variable of the institution. During my time at Kent the institution attached considerable virtue to form-filling and other paperwork, to the extent that a colleague was rebuked for suggesting (albeit tactlessly) that academics might not be the right people to carry out sorting of exam scripts. At King's, processes and admin work are much less feted, but there is a flip side: greater prevalence of cases where “nobody knows what to do” where obtaining a sensible outcome can be a stressful process requiring much ad-hoc effort, possibly dragging out over a long period.

Churn in lecturing assignments

It's well-known that teaching something for the first time is hugely more work than subsequent iterations. Therefore, churn in teaching assignments is a hidden source of load. In my three academic years at Kent I lectured on five distinct modules. In each of those years I was lecturing something for the first time. In fairness, if I'd stuck around one more year I probably would have kept a stable assignment at last. But some churn was built in: the department had a very fine-grained workload allocation model which attempted to treat fairly a large variation in research-activeness in the academic body. Although most lecturers were nominally research-active, practically there were very differing extents of that, so if you got a strong run of papers or a grant success, you would expect to shed a chunk of teaching in the next year, to be picked up by someone else. At King's, there is no such fine granularity. Instead there is a vaguely standard load and two exceptions: “workload reduction” for those with major buy-out (50% or over), and a separate, explicitly teaching-focused career path for some academic staff. Although the Kent system is “fairer”, the King's system seems to generate a lot less churn.

Vulnerability to market conditions

In the UK's current marketised approach to higher education, there is much pressure to constantly revise the teaching offering, for example to chase new segments of the student market. I've written before about how foolish and counterproductive a trend-chasing approach can be. But this pressure is greater in some institutions than others, usually because of market position and a lack of security therein.

When in 2020, calamitous university finances at Kent led to a purge of modules, the knee-jerk measure was to discard “unpopular” modules. Of course this inevitably also discards student opportunity and staff investment; I found this exasperating, because in my view duplication of teaching was a far bigger problem, yet consolidation of duplicates was not being considered. The C++ module I taught was cancelled even though clearly there was enough demand for C++—just not among the needlessly narrow sliver of the student body to which the course was being offered. (In fact I had been asked to teach extra C++ content to a different cohort of students in another department. I said no, because there was no way of getting any workload credit for doing it, and because my teaching load was already too high.) The module went away and, of course, I was given something new to lecture the next year.

While it's true that smaller-cohort modules are higher-cost, they often provide higher value for the students that choose them. The compilers module that I taught during 2019–21 was one which, both years, some students declared as the highlight of their degree. But in 2020, this module would also by rights have been cancelled if the standard numbers test had been applied; rumour was that it was “safe” only because cancelling it would have caused embarrassment in a department where “Programming Languages and Systems” was arguably the most successful research group.

Duplication of teaching effort

Although I just covered the “what” of this, it's worth considering how a department can end up with internal duplication in its teaching offering. One ingredient is what we just discussed: courting many market segments, leading to many formally distinct programmes, each with formally distinct modules. The mistake is letting formal distinctions become actual distinctions. It takes wise management to stop this from happening. Sadly the unwisdom tends to be on the academics' part in this case.

At Kent, as a department we used to teach introductory computer science four times over: to Bachelor's students of “Computer Science” at Canterbury, to Bachelor's students of “Computing” at Medway, to “conversion Master's” students and to “Year in Computing” students (Bachelor's students studying an extra year in an otherwise non-CS degree). This could have been fine, but there was some but very little sharing between the four programmes. Sometimes there were good reasons for distinct provision, but at other times there definitely was not. (Once I was told we can't merge two near-identical courses “because that module code starts with a 3 and that with an 8”.)

Sometimes the reasons were more logistical, such as the programmes' teaching being timetabled in incompatible ways. However, even timetabling barriers don't excuse fully separate offering: the same people delivering the same course in two places at two times is much less work than different people delivering two formally unrelated courses, each with their own materials (and no guarantee of mutual awareness!).

At King's there is less duplication, and proposals to introduce new modules seem to be treated more circumspectly. However, we are just starting up a new “Artificial Intelligence” programme that certainly has the capacity for adding duplication (allow me to report back). I'd expect departments' care in avoiding duplication to be roughly inversely proportional to their teaching-orientedness: if spawning a new course is viewed as an expensive undertaking because of the other things the academics could be doing, more effort is made to avoid it, while if the attitude is “teaching is what we pay academics to do” then of course they can always be made to do more! At Kent, prevalence of the latter attitude was one of the factors pushing me away from the institution.

Over-use of crude metrics like staff/student ratio don't help things, because this phenomenon literally doesn't show up in them: same staff numbers, same student numbers, but many ways in which the teaching work can be structured. Although they measure equal, they are far from equal in reality: with many modules on the go, I found that the “divided attention” factor degraded how effective and conscientious I could be in any one part of my teaching.

Academic atomisation

Continuing the broad issue of duplication, a further variable is the extent to which academics talk to each other. Departments differ a lot on how readily they build community. Even pre-pandemic, King's was reportedly not great on this, because central London is expensive and time-consuming to get to, so working from home is more common. (Also, academics not getting their own offices doesn't incentivise working from the department.) At Kent, we were pretty good at talking to each other, with a supportive day-to-day culture, although I don't know how well this has now been revived since the pandemic.

Academics talking to each other, forming a support network, is an important way in which the academic body finds out the good ways to deal with a task or to minimise the work generated by a given issue. One hazard is that with many meetings now continuing to be virtual, there is less opportunity for informal chat, such as in the spaces before and after physical meetings, yet it's largely in these spaces that support networks are sustained.

Bad technology

Ill-fitting technology pushed down from above can easily make simple teaching-related tasks complex to perform. The converse is also true—good technology does exist—although for some reason this phenomenon mostly seems to be going one way. To pick an example continuing the duplication and atomisation themes: nowadays Moodle and similar systems have brought new levels of secrecy and obscurity to teaching within a department. Whereas not long ago, course materials went on a public or internal web page, or could be picked up in hard copy (gasp!) from an office, nowadays you can't even look at your colleagues' materials without requesting access one-by-one, unless they have had the foresight to make their page public—which is not the default. This inappropriately corporate “need-to-know” access control policy is an imposition from above, though the imposition of poorly chosen technology.

As another example, at Kent we had a very nice shared teaching server into which all students had ssh access. The students quickly learned the basics of how to use this system (it was used by many modules) and it proved a very flexible environment. For example, within it I wrote a gated submission system that could give students pre-submission feedback (helping them) and stop them submitting ill-formed assignments that would foil my automatic marking (helping me). It also ran an “anonymous Q&A” CGI script that had been used by several modules over many years. At King's there is no such server and for modules wishing to have a shared machine, a bespoke set-up needs to be provisioned each time, at much academic cost. It turns out that this can be done, at a cost of explaining the requirement from scratch several times, and continuing to fight for admin attention when (for example) the initial provisioning level is wrong or the well-running of the service is otherwise overlooked. These problems would not happen with a single widely-used service of recognised importance to the department's teaching, as was the case at Kent.

Correspondingly, here at King's even more faith is put in bought-in solutions like Padlet, Gradescope and the like (though these were by no means absent at Kent). My view is that in a computer science department these tend to be relatively high-cost and low-value.

Academic autonomy in respect of teaching

The IT problems I mentioned aren't really problems if IT systems are provided for discretionary use, if plenty of alternatives are available, and if academics are free to choose. It's when their use is mandated that the problems start to bite. Reading David Evans's excellent (but understandably US-centric) “How to live in paradise” left me suspecting that many top-level US public research universities leave quite a lot of control in the hands of academics about how to teach, compared to ostensibly “peer” institutions over here (if one includes King's in that, say; one can certainly question this!). If this is true, perhaps the US tenure system and some kind of fractionally more academic-led culture can explain this. (Or I could be completely wrong; this may be a greener-grass effect.)

My experience of UK departments is that there is more pressure, and sometimes a hard requirement, to teach in centrally agreed ways. Correspondingly, this could perhaps be explained by the tight grip here of UK government policies and their knock-on centralising influence on institutions. For example, at King's we have a default formula for teaching delivery, adopted at department level. One can ask to follow an alternative model, but that's it; the menu has exactly two options. Even worse, the alternative model comes pre-torpedoed for undesirability; it seems to be designed so that nobody will choose it. I'm still not sure why this is considered a good idea.

In fairness, institutions do vary on this even within the UK. To pick my favourite outlier again, in Cambridge a few years back, a colleague decided to teach C programming using a flipped-classroom approach, recording lecture videos and then using the timetabled lecture slots to help students with examples. (This in turn was possible only because the class size was below 100.) No other course in the department was taught this way at the time. I'm not sure whether he did this unilaterally or following some permission-seeking, but I'm fairly sure it wasn't an arduous approval process. This kind of autonomy is rare in the other UK departments I know. Compounding this is how although deviation might (or might not) be tolerated on paper, at many institutions it would likely cause the academic many admin-headaches if even the first round of student feedback was less than glowing.

Attitudes to student feedback

This is my last point and is one that has taken me a bit longer to tune in to than the others. My experience has been that departments vary a lot in their attitudes to student feedback. This can have significant consequences for workload: a department that takes every student comment as a knee-jerk “action point”, which the academic must resolve, will create a lot more work than one that simply uses student feedback as a crude aggregate barometer. This can be thought of as an internal dual to the “external” churn of one's assigned courses or modules; it is churn occurring internally to each of those courses.

Pressure to improve can be helpful, but not all the pressure actually does improve. For example, tacit “disengagement compacts” between students and lecturers (and I thank a nameless colleague for bringing that phrase to my attention recently) create pressure instead to lower standards. I wager that most of my colleagues at King's, at least those who've been around a while, would privately agree that this effect has become embedded in our department. Some have already privately said so; as one put it (I paraphrase slightly), “the approach here is to give them a gold star just for turning up”.

Workload-wise, there are other problems arising from a high internal pressure to change. The over-use of video-based delivery interacts especially badly with it, because a unit change is much more expensive when made to recorded video than to in-the-room delivery. My department currently has a policy I regard as a nonsense, which is that lecturers should “re-do 30% of videos each year”. Aside from the madness of video lecturing, it's a silly notion to enforce yearly incremental change, since changes in one place invariably have ripple effects elsewhere. This is why academics prefer to improve their courses every few years in a “big bang”, rather than try to focus changes to a small portion every year. Changes simply don't focus nicely like that! And video delivery amplifies this effect: something as simple as slide numbers visible in a video can made it a Herculean task to achieve a given change to an acceptable standard.

What that standard is also varies. Video-based delivery tends to coincide with an overemphasis on production value. As someone who was often lectured from hand-drawn OHP slides, my view is that any big emphasis on production values means the time isn't being spent on the important stuff. Yet now, if your materials look a bit raggedy, students are sure to complain, while lecturers are given gold stars for producing something slick but superficial. Different departments can be infected by this madness to greater or lesser extents.

The end

I think that's all I have. I hope it was useful to someone! I've not been particularly guarded about what I see as structural problems in how teaching is done, both in my current department and at others I know. However, no department gets away unscathed... I think that says a lot about the general state of UK higher education right now, and no doubt the same problems exist elsewhere in the world too. Reflecting more widely on “the job” (which isn't the same job everywhere), above I made a reference to David Evans's “How to live in paradise”. This is a piece I recommend and that I intend to revisit more broadly in a future post.

[/highered] permanent link contact

Powered by blosxom

validate this page