What I learned from teaching online

Or: How I learned to stop worrying and just plan my teaching and writing

Much has been written about the downsides of online education in the covid era. This can lead one to the common – but largely false – conclusion that all online education represents a poor substitute for its in-person counterpart. Of course, most of the kinds of pandemic-related online learning that happened in 2020-21 and now will play at least a temporary role in 2022, were *ad hoc* in nature and came about in response to the impracticalities of putting a bunch of people together in a room during the pandemic.

However, by total coincidence, I was slated to develop a course on Spatial Epidemiology for the new-ish online MPH program at Michigan during the Fall of 2020. It didn’t quite go off with a hitch, and the process of putting it together was oftentimes quite torturous for pandemic-related reasons: You can see me below, looking bleary-eyed and exhausted in front of a very messy bookshelf in my home office. This was during what was likely the 5th or 6th attempt at recording a video that was repeatedly interrupted by my dog barking or one of my kids busting through the door behind me.

Me in front of a messy bookshelf while recording lectures in the early days of COVID-19

Backwards Design

But the very structured nature of the online course taught me a few things about not only how to structure my classes – in-person and otherwise – but also about how to think about writing up research papers and other forms of documentation in a more-structured way. Because a well-produced online course is a team effort involving graphic designers and video editors in addition to the instructor, more planning is required than in the one-man-band version of course design I was previously used to.

Notice I said required rather than necessary, because this experience has made me re-consider the ‘forward’ approach to course development and writing I have typically followed. (Full disclosure: My partner – Sarah Zelner – is a pedagogical consultant at UM, so credit for any of my ideas about this also belongs to her.) The forward approach to course development is to come up with a list of specific content you want to cover, string it together into a series of class sessions, craft assignments that sort of align with that content, and go ahead and teach the class.

Many great classes have – and will continue to be – taught this way. I have personally taught what I felt like were effective graduate and undergraduate courses over the years which were developed and taught in this seat-of-the-pants fashion. But in every class, there has been something – or a bunch of things – that felt out of place. Often, it was a book or set of articles that I knew we interesting and important, but that didn’t quite link up with the other content in the class. Or it was an assignment that was engaging and exciting, but didn’t necessarily allow students to extend and demonstrate their competency in things I had been ostensibly teaching them. Those things felt not-right, but my only solution to them was to keep bashing away, iterating readings and assignments one semester at a time, trying with limited success to work out the kinks.

My online experience working with course designers at UM introduced me to the ‘backwards’ approach to course design: This involves starting by defining the overall learning goals of the course, using that to guide the assessments and in-class activities, and then using that to figure out what should happen during the class meetings. This made it easier to see how the pieces should fit together, which content was extraneous, what needed more time, etc. In other words: if you thought about why you were doing what you were doing before you did it, it might make more sense to everyone!

Enter ‘backwards paper writing’

Then, along with my colleagues Ella August and Kelly Broen at UM, I began to wonder if this approach could also help with the process of scientific writing. Ultimately, a paper is a pedagogical document where we explain what we did as clearly as possible, justify why we did what we did, and argue for the importance or relevance of the whole thing. Despite the fact that the sections of a paper have clearly-defined reasons for being that are right there in their names – Introduction, Methods, Data, Discussion – I often find myself writing ‘forwards’ rather than backwards.

Rather than thinking about what I want to accomplish with the overall paper from a scientific or professional perspective, I typically find myself bashing through the sections roughly in order and then editing to make them cohere. But what if some of that ‘backwards’ magic could work in scientific writing?

Well, all of this is a long-winded way of saying we think it can, and the product of it is this new paper in Patterns called “A guide to backwards paper-writing for the data sciences”. In it, we provide a set of ideas about how to make sure that the goals in writing are well-defined in advance so that they percolate through all of the sections of the paper.

It was a nice opportunity to think on the page and shouldn’t be taken as definitive in any way, but instead to act as the beginning of a conversation about how we can make the procerss of paper-writing more focused, more effective, and possibly even somewhat enjoyable. Excited to see what people make of this and whether the ideas in here translate out of our small corner of the world!


6 thoughts on “What I learned from teaching online

  1. Fine idea. It seems to me that Andrew has occasionally posted “Advice on writing research articles” (I found one from 2014), part of which is a recommendation to “write backwards” as you describe. In attempting to follow this advice in my own writing, I find that trying to write backwards is not a bad thing, but the articles that I end up submitting to journals are typically on version 20 or so. At that point, the article has been written backwards and forwards several times, so it does not make an enormous difference whether version 1 was written backwards, forwards, from the middle to the ends, or whatever. Course lectures may different matter because most of us do not have the time to perform 20 iterations.

    • Yeah – this is a good point. Certainly it doesn’t take the place of that revision. I think maybe a more accurate way of putting the ‘backwards’ approach wrt writing papers is that it relates to the individual sections of the paper, and having a concrete idea about why you are writing each one ahead of time is super important. I think this is also v. important when you haven’t written a ton of them and don’t have the gestalt sense of what each section is and isn’t supposed to do, and we’re hoping to just give a better sense of that for folks who are getting started.

  2. This is good advice. I took a similar approach to writing academic talks, it helped to prevent getting bogged down in slides chock full of notation. Elsewhere, I think it would be described as taking a “top-down” approach instead of a “bottom-up” one. After all, the structure of a course already has a good deal of built-in structure, such as class lengths and posting a pre-designated topic list, so you might as well use that as a starting point.

    • Yeah – another way I’ve heard it described more pointedly, at least wrt teaching, is ‘teacher-centric’ vs. ‘student-centric’, which I think is about right.

  3. hmmm…i think some of the papers criticized on here got in trouble by being written backwards: “let’s start with the results we want, and then make up the data and wing the theory later”

    (joking, mostly)

  4. Interesting observation. The approach can be extended to start with the findings and their generalizability, as supported by the study outcomes and it design.

    Eventually the study makes claims. To generalize these one can use alternative representations, some with meaning equivalence and some with surface similarity. This lays out a boundary of meaning (BOM).

    Zelner’s suggestion could be to start wit the BOM. We suggested that the last section in a paper would be Generalizability of findings. No one prevents authors to start by writing this section..
    For more on this and examples see https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3035070

Leave a Reply

Your email address will not be published. Required fields are marked *