Skip to content
 

Walter Benjamin on storytelling

After we discussed my paper with Thomas Basbøll, “When do stories work? Evidence and illustration in the social sciences,” Jager Hartman wrote to me:

Here is a link to the work by Walter Benjamin I think of when I think of storytelling. He uses storytelling throughout his works and critiques done on his works are interesting with regards to story telling to convey a message. However, I find this work really highlights differences between information, storytelling, and messages to be conveyed.

Benjamin’s article is called “The Storyteller: Reflections on the Works of Nikolai Leskov,” and it begins:

Familiar though his name may be to us, the storyteller in his living immediacy is by no means a present force. He has already become something remote from us and something that is getting even more distant. To present someone like Leskov as a storyteller does not mean bringing him closer to us but, rather, increasing our distance from him. Viewed from a certain distance, the great, simple outlines which define the storyteller stand out in him, or rather, they become visible in him, just as in a rock a human head or an animal’s body may appear to an observer at the proper distance and angle of vision. This distance and this angle of vision are prescribed for us by an experience which we may have almost every day. It teaches us that the art of storytelling is coming to an end. Less and less frequently do we encounter people with the ability to tell a tale properly. More and more often there is embarrassment all around when the wish to hear a story is expressed. It is as if something that seemed inalienable to us, the securest among our possessions, were taken from us: the ability to exchange experiences.

I’d heard the name Walter Benjamin but had never read anything by him, so I ran this by Basbøll who replied:

My own view is that stories can be usefully compared to models, i.e. we can think of storytelling as analogous to statistical modeling.

The storyteller is a point of agency, not in the story itself (the storyteller need not be character in the story) but in the communication of the story, the telling of it. The storyteller has authority to decide “what happened” roughly as the modeler has the authority to decide what comparison to run on the data.

We can think of the narrator’s “poetic license” here like the statistician’s “degrees of freedom”. While we allow the narrator to “construct” the narrative, a story is not compelling if you “catch” the storyteller just making things up, without any consideration for how this affects the overall plausibility of the story. Do note that this happens even in fiction. It’s not really about true and false, but about a good or bad story. If it’s just one thing happening after another without rhyme or reason we lose interest.

Likewise, the statistician can’t just run any number of comparisons on the data to find something “significant”. Here, again, it’s not that the model has to be “true”; but it must be good in the sense of providing a useful representation of the probability space. Perhaps in storytelling we could talk of a “plausibility space”—which is actually more usefully thought of as a time dimension. (Anything is possible—but not in any order!) Perhaps that’s why Bakhtin coined the word “chronotope”, a time-space.

Like models, stories can be subjected to criticism. That is, we can question the decisions that were made by the modeler or teller. Often, a story can be entirely misleading even though it recounts only things that actually happened. The deception lies in what is left out.

A story can also be inadequately contextualized, which leads us to make unwarranted moral judgments about the people involved. Sometimes merely adding context, about what came either before or after the main events in the account, completely inverts the distribution of heroes and villains in the narrative. I think the corresponding error of judgment can be found in the way models sometimes lead us to make judgments about causality. A story often assigns praise and blame. A model usually suggests cause and effect.

I wonder: what corresponds to “replication” in storytelling? Model studies can be replicated by gathering fresh data and seeing if it holds on them too. Often the “effect” disappears. Perhaps in storytelling there is a similar quality to be found in retelling it to a new audience. Not contextualization in the sense I just meant, but re-contextualizing the story against a completely different set of background experiences.

This is something Irving Goffman pointed out in his preface to Asylums. As we read his description of life in a closed psychiatric ward, he reminds us that he is seeing things from a middle-class, male perspective. “Perhaps I suffered vicariously,” he says, “about conditions that lower-class patients handled with little pain.” A story makes sense or nonsense (some stories are supposed to shock us with the senselessness of the events; that is their meaning) relative to a particular set of life experiences.

Models, too, derive their meaning from the background experiences of those who apply them to understand what is going on. Kenneth Burke called literature “equipment for living”. We use stories in our lives all the time, understanding our experiences by “fitting” our stories to them. Models too are part of our equipment for getting around. After all, one of the most familiar models is a map. Another is our sense of the changing seasons.

I replied that I want to write (that is, think systematically about) all this sometime. Right now (Oct 2017) I feel too busy to focus on this so I put this post at the end of the queue so as to be reminded next year (that is, now) to think again about statistical modeling, scientific learning, and stories.

18 Comments

  1. Ethan Bolker says:

    This

    If it’s just one thing happening after another without rhyme or reason we lose interest.

    reminds me of the E. M. Forster quote

    “The king died and then the queen died” is a story. But ‘“the king died and then the queen died of grief” is a plot.

    https://www.aerogrammestudio.com/2013/03/04/e-m-forster-the-difference-between-story-and-plot/

    (I don’t know the markup that will make those quotes appear as blockquotes)

  2. A good story must be plausible but also tease you with the implausible. The relationship between plausible and implausible–and their relative proportions–will vary from story to story–but the two have to be there. Gogol’s “Nose” stands out as an egregiously implausible story that, within its strangeness, makes sense. Moby-Dick is a more elongated case in point.

  3. jim says:

    But after you’ve made your model and generated your data, to communicate your results you tell a story. It doesn’t have to be honest (I.e. Fairly weigh all the evidence) to be convincing, it just has to be plausible – and this is important – in the eyes of the audience. So the audiences biases are as important as those of the scientist – hence NPRs embracing unbelievable claims of social science.

  4. Andrew:

    Re: My own view is that stories can be usefully compared to models, i.e. we can think of storytelling as analogous to statistical modeling.

    Painting/drawing could also be analogous to statistical modeling. That is if you are more visually inclined. You are right that there is a story that goes along with modelling.

    I make the point in my int. relations circles that a good story has and can beat good reason. Also implied in Khong’s Analogies at War.

    • Keith O'Rourke says:

      > Painting/drawing could also be analogous to statistical modeling.
      I would think any representation (another word for a model) of something empirical that is attempting an evaluation of that (another word for inference.)

      So not a drawing of a triangle as a mathematical object but a drawing of a physically constructed triangle that tries to discern how close it is to being a triangle.

      • Yes Keith I should have been even more explicit.

        Also; Re: Good story beats good reason

        Revision; Charismatic rationals can beat good reason. By that I meant that stories contained in small sample expert opinion have trumped rigorous empirical justification for foreign policy proposals; Backstory is often discernible in casual conversations. Maybe it can be likened to a marketing strategy also.

        • Keith O'Rourke says:

          Stephen John paper discussed here might be related https://andrewgelman.com/2018/01/23/better-enable-others-avoid-misled-trying-learn-observations-promise-not-transparent-open-sincere-honest/

          What non-experts are taken to learn is not any content but rather the apparent consensus of experts along with an assurance that certain groups of experts’ “epistemic standards” should be taken as convincing. Does telling a good story display expertise and suggest some consensus? To adequately grasp “rigorous empirical justification” may require too much expertise and only hard to trump with other experts.

          • Keith,

            I consider myself a student rather than an expert of statistics. I really enjoyed the content to the link you provided.

            I have had interest in your query, not as it relates to statistical reasoning, since I am novice to subject matter. I have had broader interest in historiography/historical reasoning and its role in foreign policy development. It may have been catapulted by Robert Jervis. His book, Perceptions and Misperceptions in International Relations resonated with some of my observations. In that regards, I am predisposed to learning how int. relations specialists think. And I was reluctant to admit that a good # of them relied on mostly opinions/stories and a small pool of expertise.

            It is intriguing to me how we define ‘non-expert and expert’ for the lines between each category are not as sharp as perhaps they were 50 years ago. In part, online resources are available to pretty much anyone who has an interest in any subject. And I think that a non-expert’s interest a subject may be less beset by career & reputational demands. But more stark is that there is less critical thinking attached to it for my tastes at least. I might also have an edge in that I was exposed to some of expert calculus at a young age. So there is institutional memory also at work. I can usually link an assumption in current scholarship to an academic I’ve read or heard.

            I am sure cognitive biases are work in foreign policy development. But it’s a whole lot more complicated than that b/c today, some int. relations programs are established by special interests and wealthy benefactors.

            Lastly I think the current university structure is not as amenable to the extent of consultations with your colleagues as it was in the 60’s and 70’s. i base this on some perspectives gleaned in the Chronicle of Higher Education.

  5. OP says:

    “The deception lies in what is left out”

    This immediately brought Akutagawa’s wonderful Rashomon to mind for me. So much contrast from so few facts!

  6. Joshua Garoon says:

    As an ethnographer, I’d venture that this begins to capture the (or at least an) ethnographic methodology, at least when compared to the (or at least a) statistical methodology. In terms of replication, the most immediate difference would be that ethnographers only rarely share their data and the way they’ve analyzed it — their “code,” whether or not they’ve coded their data — typically because of IRB constraints and epistemological objections (though there are other reasons). There are exceptions; for older studies, especially, you can find field notes and other data in archives, compare what’s there to what the ethnographer actually wrote in their published work, and compare that to other data you’ve collected, historical and/or contemporary. (This is more or less what Moore and Vaughan did in their restudy of Audrey Richards’ work in northern Rhodesia-now-Zambia.) But even setting that possibility aside, there are a number of ways to think about “replication”: re-study; comparative study; extending the case method (to use Burawoy’s terms); even meta-analytic study. Someone might object at this point that we’ve moved far away from literal storytelling, but we really haven’t: because in the ethnographic mode, you’re always going to be coming back to the narrative arguments you’re putting on the page.

  7. Christian Hennig says:

    Funny that my philosophy of science reading group is just about narratives in science, about which apparently a number of philosophers of science do some work:
    https://www.sciencedirect.com/journal/studies-in-history-and-philosophy-of-science-part-a/vol/62/suppl/C

  8. Steve Sailer says:

    Magic realism, such as Gabriel Garcia Marquez’s “One Hundred Years of Solitude”, has less concern for plausibility. I found “100 Years” delightful, but I stopped reading halfway through because the author allowing himself to make up anything he feels like meant that I didn’t have much puzzle-solving interest in the plot. With most stories, I want to see how the author solves the puzzles he has set forth, but Garcia Marquez was clearly going to do whatever he felt like, so I started applying diminishing returns logic to his book. Similarly, I seldom sit all the way through stand up comedy movies because there’s no plot to keep me hanging around to the end.

Leave a Reply