Tuesday, July 2, 2024

The Structure of Academic Writing: Lessons from John McPhee




In the world of literary non-fiction, John McPhee is a god. Through his New Yorker essays, and prize-winning books McPhee has mastered the art of narrative non-fiction. In fact, he pretty much invented the genre. He has many fans; many of whom are themselves well-known writers. They gush about his capacity to make the most turgid-sounding topics -- oranges, boats, plate tectonics -- fascinating explorations of people, culture, science and history.

Ironically, I have never warmed to him. I've tried. Honestly, I've tried. I have started reading several of his books, each time hoping I would find the hook that has lured in other readers. It doesn't seem to work for me. I usually give up after a few dozen pages. Perhaps his prose belongs to another era. A more thoughtful, more languid era. Perhaps I lack the patience to 'get it'.

There is, however, one aspect of McPhee's writing that I have warmed to: his writing about writing. In his essay collection, Draft No 4, he shares lessons from a lifetime of writing. Some of these essays contain insightful and useful advice. In this article, I want to reflect on one of his primary lessons: the importance of structure in non-fiction writing. I then want to see how that lesson can be applied to academic writing, using one of my own academic articles as a guinea pig.

As you shall see, McPhee has quite an elaborate and playful way of thinking about the structure of writing. A lot of academic writing is formulaic and routine. Rarely does anything break out of the conservative mould of traditional article structures. I think academics could benefit from adopting McPhee's elaborate and playful approach. If nothing else, they might have fun in the process.


1. McPhee on Structure

McPhee says that he learned about the importance of structure from his high school English teacher, Olive McKee. She taught him in the 1950s. She made him, and the rest of his class, do three writing assignments every week (which, wearing my teaching hat, sounds like an exercise in self-punishment from her perspective). They could write about anything but they had to prepare a structural outline for each and every piece. This encouraged him to foreground structure in his approach to writing. He has since passed the lesson on to his own students, telling them:


You can build a strong, sound, and artful structure. You can build a structure in such a way that it causes people to want to keep turning pages. A compelling structure in nonfiction can have an attracting effect analogous to a storyline in fiction." 
(McPhee 2018, 20)

 

But he adds some important nuance to this. Noting that (a) structure should be relatively invisible to the reader and (b) that structure must serve a purpose. In his own words:


"Readers are not supposed to notice the structure. It is meant to be about as visible as someone's bones...A piece of writing has to start somewhere, go somewhere, and sit down when it gets there." 
(McPhee 2018, 34)

 

Indeed. These aphorisms aside, the most interesting aspect of McPhee's exploration of structure is his attempt to bring the reader under the skin of some of his most famous works, and to explain how he came up with the structure of those pieces. Draft No. 4 is full of examples of this. I'll just summarise a few of them, including along the way some of McPhee's infamously bizarre structural diagrams, to illustrate his process.

I'll start with profile pieces. Early on in Draft No. 4, McPhee notes that most magazine profile pieces have the same basic structure. You interview a person and the people around that person, and you thereby triangulate on a vision or perspective on that person. That gives each profile piece the following basic structure, where the X in the middle represents the person being profiled and the dots around the edge represent the people being interviewed about that person.




After about a decade of professional writing, McPhee grew tired with this structure and wondered if he could try something different. He came up with the idea of a dual-profile piece, in which two connected people could be profiled, incorporating the perspectives somewhat overlapping circles of interviewees. The structure is illustrated below.



Of course, this was a structure in search of a subject. Eventually, McPhee hit upon the pair of profiles he could use to fill out the structure: Arthur Ashe and Clark Graebner. Both of whom were well-known US tennis players in the 1960s. They had very different backgrounds. Graebner came from a privileged background; Ashe, an African-American, did not. Nevertheless, due to their talent, they had known and interacted with one another from an early age. They played each other in the semi-final of the US Open in 1968. Watching the game, McPhee realised they provided the perfect fodder for his experiment in the dual-profile. The resulting book -- Levels of the Game -- is considered a classic in sports' journalism. For added structural nuance, McPhee built the profiles around a description of that semi-final game.

McPhee felt that this experiment in structure was a success: that the dual profile had a depth and range that was lacking in the traditional single profile. This led him to consider further experiments in structure. One such experiment was based on a very simple structural diagram:




Translated into English, the idea was to write a series of three connected dual profile pieces featuring a common protagonist (the common denominator in the diagram). In a sense, the common denominator would be the main character in the work, but the other three characters would be given a good share of the spotlight and through their interactions they would shed a unique light on the main character. McPhee found a suitable subject matter for this piece. The end result was one of his most celebrated pieces of writing Encounters with the Archdruid. This was about Dave Brouwer, a famous American climber, environmentalist and conservationist, who founded Friends of the Earth. The book was based on Brouwer's clashes with people that did not share his worldview. They were, respectively: (i) Charles Park, a mineral engineer and proponent of mining; (ii) Charles Fraser, a property developer; and (iii) Floyd Dominy, a federal bureaucrat and evangelist for hydroelectric power dams.

Admittedly, these two examples are not representative of McPhee's typical approach to structure. Both of these examples involve structures in search of subjects. Most of the time, McPhee had subjects in search of structure. In other words, he had researched a topic, accumulated an abundance of material, and then needed to reduce it all to some manageable, informative and insightful structure. One example of this, which is probably my favourite in his book, is an essay he wrote about a canoeing trip in Alaska. (The essay appears in his collection, about Alaska, entitled Coming into the Country).

The trip took place over 9 days, from the 13th to the 21st of a particular month (I don't know which one since I have not read the essay itself, only descriptions of it). The natural structure and the one that most writers would probably follow, would be to write a chronological story of the trip, starting on day one and ending on day nine. McPhee decided not to do that. Why not? Because it didn't fit with the themes he wanted to explore in his writings about Alaska. Those themes included the hardship of nature, the struggle for existence and, perhaps most importantly, the cycles of time in the natural world (birth, life, death, seasonality etc).

So, instead of adopting a linear chronological structure, he adopted a circular structure. He started the narrative (using the present tense) on day five of the trip (the 17th of the month) and then continued it right to the end of the trip on day nine (the 21st). The narrative, however, didn't end there: there was a flashback to day one of the trip (the 13th) and the story continued, told in the past tense, back to day five (the 17th). In addition to this structure replicating the cyclical nature of time in the natural world, it also created dramatic tension. On day one of the trip, the intrepid river explorers, encountered a grizzly bear. This was, as you might imagine, a tense moment, encouraging some reflections on mortality and the mismatch between humans and bears. But if he had adopted a linear chronological structure, the encounter with the bear would have been near the start of the narrative. By switching to the circular structure, the encounter came just after the half-way point. The diagram below illustrates the circular structure adopted.




There is a lot more in McPhee's essay on structure. Hopefully, this is enough to give you a sense of his method. Two things stand out for me. First, is the care and attention that he pays to the structure of his writing. I probably haven't conveyed this effectively in my summary but it is clear that McPhee agonises over structure, often taking weeks or months to figure out how best to structure his pieces. He uses props and diagrams to help him map them out. The illustrations I provided above were created by McPhee himself to explain his thought process to students. Second, is his willingness to play with and experiment with structures. He doesn't follow structural cliches. He tries to alter structures in order to better serve the purpose of the writing.

Could academic writing benefit from taking a similar approach?


2. Experimenting with the Structure of Academic Writing

I think it can. As I already said, a lot of academic writing is formulaic. Nowhere is this more apparent than in the sciences, where the typical academic journal article, particularly if it is reporting on the results of an experiment, tends to follow the same basic structure: introduction/literature review, methods, results, discussion. This structure is so deeply engrained in the academic culture that people are often penalised (or simply ignored) for deviating from it.

Even in non-science disciplines, articles tend to follow the same routine structures. So, for example, in philosophy writing (which is what I am most familiar with), articles will usually set out some problem or puzzle, then introduce an argument that solves that problem, and then defend that argument from counterattacks. The sub-sections within a philosophy article may not be prescripted (as they are in the sciences) and there may be some more experimenting with form, but usually those elements are there and they occur in the same sequence.

Most of the time it makes sense to follow the norms of your given academic discipline. I say this to students all the time. Go back to McPhee's comments on structure: writing should serve a purpose; it should bring the reader somewhere and then sit down. Most of the time, academic articles serve an argumentative or persuasive purpose. The goal is to convey an argument to a reader. The conclusion is the location to which you bring them. You want them to sit down once the conclusion is reached. If there are tried and tested structures for doing this, then most people, most of the time, should adopt those structures. Otherwise, they risk haphazard, hard to follow and nonfunctional writing. This risk is probably highest for beginning students, who haven't yet grasped the purpose of academic writing. At the very least, they should try to master the basic forms of academic writing before experimenting with structures.

This doesn't however, mean that academic writing must be formulaic and boring. And it doesn't mean that some creativity with structure is off limits. One of the reasons that McPhee's comments on structure appeal to me is that, without realising it, I think I have long adopted a similar approach to how I think about the structure of academic writing. In theory if not in practice.

As I just said, the purpose of most academic writing (and certainly most of the academic writing that I do) is argumentative. I think of arguments as having a natural structure. They are chained sequences of premises, leading to conclusions, responding to or rebutting objections and counterarguments. The many many argument diagrams that I have prepared for this blog, and for my classes with students, illustrate these argumentative structures. Consider the diagram below, taken from a piece I wrote on this blog earlier this year (it was about Anselm's ontological argument).



Anselm's Ontological Argument with Objections


The purpose of writing is to get the structures of those arguments into the reader's head. That doesn't mean, however, that you have to follow the same, predictable course through the argument. You don't have to start at premise 1 and work forward from there. You don't have to follow your own pathway through the argument. Just because you started with premise 1 doesn't mean the reader has to do so too. It is possible to divide it up in different ways. When you translate the multidimensional (or at the very least two-dimensional) structure of an argument into a linear sequence of prose, you can make creative choices that could, if done right, accentuate the argumentative purpose of the piece.


3. A Worked Example: My Article on 'Tragic Choices and Responsibility Gaps"

Sadly, I don't always practice what I preach. When it comes to the majority of my academic writing, I tend not to experiment much with structure. Usually, I come up with an argument, I map it out -- sometimes on paper; sometimes just in my head -- and then I write it up, typically following the journey I took through the argumentative structure myself on the page. I don't agonise and second-guess myself in the same way that McPhee appears to do.

I'm not sure why this is the case. It may be due to my own character: I'm quite impatient and, once I come up with an idea, I like to just write it up and not overthink it. I've said in interviews before that, for me, the first draft is usually the last draft. I don't enjoy rewriting and editing. I am not one of these people that 'finds' their writing in the edit. It may also be due to the pressures of academic work. The publish or perish incentive scheme doesn't lend itself to lengthy meditations on structure. McPhee, for instance, talks about spending weeks lying on his back, pondering the best structure for his writing. Most academics don't have that luxury (or, at least, don't feel like they have that luxury).

But that's a pity. Recently, as life has filled with other, more pleasant, distractions, and I have slowed down my rate of academic productivity, I've been thinking that I should play more with the structure of my academic writing. Rather than think about this in the abstract, I thought it might be fun, and instructive, to do a worked example with a piece of my own writing.

The piece in question is my article "Tragic Choices and the Virtue of Techno-Responsibility Gaps". To be very clear, when I wrote this piece, I did not think much about its structure. I just came up with the idea and I wrote it. Some of the structure was added later in response to critical reviewers of the piece. If I were to go back and reconsider its structure -- play with its form a bit more -- how might I do it?

It helps if I break the article down into its main component parts. I won't provide a detailed argument map. Instead, I will break the article down into a few main structural elements. Overall, the purpose of the article (as I conceived it) was to argue that, contrary to received opinion, the responsibility gaps created by autonomous machines could, in some cases at least, be a good thing. Something to be welcomed, not feared.

To argue for this conclusion, the article had the following structural elements:


The Problem: This is the bit of the article in which I outlined the received wisdom, i.e. the common view that responsibility gaps are a problem and something ought to be done to eliminate or minimise them. This required a review and analysis of the existing literature on the topic. Having stated the problem, I then introduced my alternative view, i.e. the proposition I wished to defend.

 

I then introduced three main claims which, when combined, led to my desired conclusion. These claims were numbered Claim 1, 2 and 3 in the article:


Claim 1: There are such things as tragic choices, i.e. moral decisions in which there is no clear 'right' answer and in which every choice seems to leave behind a moral 'taint' or 'remainder'. These tragic choices pose a problem for responsible moral agents.
Claim 2: There are three different strategies we can use to cope with the problem of tragic choices -- illusionism (i.e. pretend its not a problem), responsibilisation (pretend there is a right answer and we bear responsible for it), and delegation (make it someone else's problem) -- each of which has a unique blend of costs and benefits, none of which is ideal.
Claim 3: Autonomous machines could be a useful tool in addressing tragic choices because they allow for a reduced cost form of delegation, but in order to embrace this possibility we have to welcome, not reject, responsibility gaps.


My view is that claim 1 + claim 2 + claim 3 supports my desired conclusion. But claim 3, in particular, is controversial. So, to bulk up the argument, I responded to four objections to claim 3. Some of these I came up with myself; some of which I added in response to critical feedback. They were:


Objection 1 (O1): The randomisation objection - why not use randomisers, not machine learning devices, to address the problem of tragic choice?
O2: The impossibility objection - you cannot delegate responsibility to a machine (or any other entity) in a way that reduces the moral costs of the delegated decision. This is because, in delegating, you retain moral control.
O3: The agency laundering objection - even if the argument is correct isn't there a danger that unscrupulous actors will use it as an excuse to hide their responsibility (launder their agency) for decisions they have made?
O4: The explicit tradeoff objection - because decisions need to be coded into algorithms, doesn't this make them more explicit, and the moral tradeoffs inherent in them, more salient, not less? In other words, doesn't delegating to machines heighten the moral costs associated with tragic choices, not lessen them?

 

I offered, what I thought were appropriate, responses to each of these objections, thus leading me back to the desired conclusion: we should embrace delegation to machines, and the associated responsibility gaps, at least in the case of some tragic choices.

Obviously, there is a lot more nuance in the original article, which, as always, I encourage you to read. If you wanted to find additional structural elements in it, you could. But this should give a clear enough outline of its main contents.

When I wrote it, I essentially wrote it in the sequence I just outlined to you. I started with the problem, I then introduced and justified the three claims, before responding to the four objections. The diagram below illustrates this structure.


Original structure - Tragic choices with focus on the problem of responsibility gaps


How could I have done it differently? Taking the diagram above as a starting point, it is easy to think about ways in which the material could have been rearranged and presented in a different order. Doing so, the same argument would be conveyed, but the emphasis and focus would vary. Altering the structure in this way could also affect the framing of the argument. In the original version, the article was clearly intended to be a contribution to the debate about responsibility gaps and autonomous AI (this was, to some extent, forced on me since the article was part of special collection of articles on that topic). But shifting the initial entry point into the argument, and the point at which the argument ends, could have helped to reframe the argument as a contribution to a different debate.

Let's consider some alternative structures. I could, for instance, have started the article with Claim 1, the problem of tragic choices itself. I could have said to the reader "hey, there is this problem with moral decision-making and it poses a threat to responsible agency. How can we address this problem?" This would have made tragic choices, not responsibility gaps, the main focus/frame for the paper. I could then have proceeded to consider a potential solution to the problem, namely the randomisation solution (Objection 1 in the original draft). I could have argued that this was a partial solution at best and that a better alternative needed to be found. This could have led, naturally, to a discussion of Claim 2 and the different costs and benefits associated with the different solutions to the problem of tragic choice. From there, I could have introduced my preferred solution -- reduced cost delegation to machines -- which is, of course, Claim 3. This would have led to a discussion of the remaining objections to Claim 3 (O2, O3, O4). Then, by way of a general conclusion, I could have pointed out the ramifications of my argument for the responsibility gap debate. The diagram below illustrates this structure.

Alternative Structure 1 - Tragic choices with focus on tragic choices


Here's another possibility. In the middle of my original discussion of Claim 2, I talked a bit about the phenomenon of delegation and its importance in human social life. I used Joseph Raz's famous (in the legal philosophy world!) service conception of legal authority to illustrate this idea. Roughly, one of Raz's claims is that legal authorities sometimes mediate between humans and their moral reasons for action. When there are disputes between multiple agents about the right or desirable thing to do, a legal authority can resolve that dispute by doing the moral reasoning for us and giving us a decision to implement/enforce. We don't always second-guess or question the reasoning of the legal authority because that would defeat the point of having the authority in the first place. It performs a service for us, obviating the need for certain moral debates and disputes.

Properly expressed, I think this is an interesting idea and does highlight something important about the role of delegation in moral life. I could have started the argument there, saying to the reader "Hey look, delegation is a really important, and perhaps misunderstood, aspect of moral agency: sometimes, as moral agents, we need to delegate to others". But delegation is just one of several strategies we use to address the challenges of moral agency (Claim 2). I could then have started to talk about the particularly acute challenges associated with tragic choices (Claim 1), leading to my proposed solution (Claim 3). This would have entailed a discussion of reduced cost delegation as an important new form of delegation. This would have introduced a 'Delegation: Part 2' into to the article. In other words, there would be a degree of circularity in the structure of the article, akin, perhaps, to McPhee's circular journey down the river. Albeit, in my case, I would need to defend this proposed form of delegation 3 from the four objections, leading finally to some mention, perhaps peripheral, of the responsibility gap literature.


Alternative Structure 2 - Tragic Choices with Focus on Delegation


That's two suggestions. No doubt more could be proposed. Hopefully, even with this limited discussion, you can see how restructuring is possible and the potential benefits of doing so. Each of the proposed restructurings I have presented changes, in important ways, the emphasis and focus of the article. All the elements are still present, but the rearrangement allows those elements to serve a new purpose, and appeal to a different audience.

Since I work in the ethics of technology, it made most sense for me to frame my article as a contribution to the debate about responsibility gaps. But perhaps I should not have been so conservative in my approach. I could have presented the article as a contribution to legal and political theories of legitimate authority and delegation. I could have presented it as a contribution to moral philosophy and the resolution of moral dilemmas.


4. Conclusion

This brings me to the end of this article. To briefly recap, I started out looking at John McPhee's reflections on the importance of structure to non-fiction writing. Structure ensures that the writing serves a purpose - that it brings the reader somewhere and then sits down. McPhee has developed his thoughts about structure to a high level, constantly playing with and experimenting with form in order to improve the quality of his writing.

Structure is important in academic writing too. Most academic writing serves an argumentative purpose - it brings the reader to a conclusion and then sits down. But many academics are quite conservative and rigid in how they structure their writing. What I have tried to suggest in this article is that they should reconsider this conservatism. This doesn't require radical experimentation with form, per se. Once you have decided on the main elements of your argument, you can rearrange them to serve (subtly) different purposes. Doing so might allow you to see potentialities in your own writing that would be missed by following the well-trodden path.







1 comment:

  1. An insightful look at academic writing! Just as XeniaCanary revolutionizes the gaming landscape, John McPhee’s lessons can reshape how we approach clarity and structure in our writing.

    ReplyDelete