“When I put Jorge in the library I did not yet know he was the murderer. He acted on his own, so to speak. And it must not be thought that this is an ‘idealistic’ position, as if I were saying that the characters have an autonomous life and the author, in a kind of trance, makes them behave as they themselves direct him. That kind of nonsense belongs in term papers. The fact is that the characters are obliged to act according to the laws of the world in which they live. In other words, the narrator is the prisoner of his own premises.” — Umberto Eco, Postscript to The Name of the Rose (translated by William Weaver)
Perfectly put. As I wrote less poetically a few years ago, the development of a story is a working-out of possibilities, and that’s why it makes sense that authors can be surprised at how their own stories come out. In statistics jargon, the surprise we see in a story is a form of predictive check, a recognition that a scenario, if carried out logically, can lead to unexpected places.
In statistics, one reason we make predictions is to do predictive checks, to elucidate the implications of a model, in particular what it says (probabilistically) regarding observable outcomes, which can then be checked with existing or new data.
To put it in storytelling terms, if you tell a story and it leads to a nonsensical conclusion, this implies there’s something wrong with your narrative logic or with your initial scenario.
Again, I really like how Eco frames the problem, reconciling the agency of the author (who is the one who comes up with premise and the rules of the game and who works out their implications) and the apparent autonomy of the character (which is a consequence of the logic of the story).
This also connects to a discussion we had a year ago about chatbots. As I wrote at the time, a lot of what I do at work—or when blogging!—is a sort of autocomplete, where I start with some idea and work out its implications. Indeed, an important part of the writing process is to get into a flow state where the words, sentences, and paragraphs come out smoothly, and in that sense there’s no other way to do this than with some sort of autocomplete. Autocomplete isn’t everything—sometimes I need to stop and think, make plans, do some math—but it’s a lot.
Different people do autocomplete in different ways. Just restricting ourselves to bloggers here, give the same prompt to Jessica, Bob, Phil, and Lizzie, and you’ll get four much different posts—and, similarly, Umberto Eco’s working out of the logic of a murder in a medieval monastery will come out different from yours. Not even considering the confounding factor that we get to choose the “prompts” for our blog posts, and Eco picked his scenario because he thought it would be fruitful ground for a philosophical novel. Writing has value, in the same way that prior or posterior simulation has value: we don’t know how things will come out any more than we can know the millionth digit of pi without doing the damn calculation.
Lots of interesting implications of the Eco quote. It’s why I think writing (or more generally, exposition) is essential to thinking, at least in science. Language imposes a form of logic that shows you what you can actually say, which might not be what you thought you were going to say when you first sat down to write. One of the things that bugs me about researchers (especially grad students) using LLMs to do a lot of the writing for them is that I don’t think it supports the same kind of conscious recognition of why you have to say things a certain way or can’t make a point the way you originally thought you could.
It could be interesting though to think about what it would take to prompt an LLM to narrate why the passage it outputs makes the argument the way it does and what it learned would not work during the composition process.
Artists talk about similar things, like there’s never some idea that gets realized the way you thought, because the materials impose constraints that shape what comes out.
Yes! I sometimes write to figure out what I think.
John –
I sometimes write to figure out what I think.
In my experience, this is so often true for students who have been given an expository writing assignment. Perhaps particularly true for some non-native English speakers.
Often they think they know what they want to write about, simply because they have been assigned a topic, and think that they should be able to basically start with the first word and run the narrative through until the last word. Except they haven’t really given the topic much thought and don’t yet have much of any idea what they want to say. It’s only at the end of that process that they’ve found their thesis.
The problem is that what they really need to do then is take that thesis they’ve discovered and basically start all over again, organizing their witting with that thesis in mind.
To do that it means that they’d have to view revising as intrinsic to the witting process. But often revising feels boring, or tedious. For many students, revision doesn’t feel intrinsic, or it feels like they would face having made a “mistake” (again, there’s often a cultural overlay that builds in a resistance to addressing “mistakes”). Or maybe they’re waiting towards the last minute on a deadline. So many students really, really don’t want to essentially start over and revise.
I was just thinking about what I said about language imposing “a form of logic that shows you what you can actually say,” but I think that should be revised slightly…. it’s more like written language is a better medium than thought for rendering visible one’s own perceptions of the logic governing what can be said. I imagine this being partly because the act of writing invokes a perception of a reader, which presupposes some constraints on what can be considered plausible.
Jessica –
I imagine this being partly because the act of writing invokes a perception of a reader, which presupposes some constraints on what can be considered plausible.
When I put down something I’ve been writing, and come back to it later, more often than not I want to make significant revisions. I think in a sense it’s because my “reader” and I are more removed from the time away. My perceptions of what I’ve written, in a sense, are more objective or perhaps my perceived reader has come into better focus.
Same, I think its easier to sympathize with the reader when I’m revising rather than trying to simultaneously create the text.
That’s a good way to think of it.
Jessica Hullman wrote, “It could be interesting though to think about what it would take to prompt an LLM to narrate why the passage it outputs makes the argument the way it does and what it learned would not work during the composition process.”
I find it therapeutic to ask ChatGPT to criticize what it just had responded to regarding my (original) request. My impression is that this original AI criticism is quite formulaic. So much so that a retired professor of English informed me that it mimicked quite well his (standard) comments regarding his student assignments. I suppose this is not totally surprising given how the training was structured.
Sometimes, depending on how much free time I am willing to waste, I ask for a comment on the second comment. I assume that this sort of activity is marginally better than wasting time in the pre-AI era. But, I have not asked about this assumption as yet. I encourage others to feel free to do so.
Providing informatics support to biologists was my career. After years of listening to them beg and plead for ill-defined assistance I learned to ask them to write one page describing what they wanted. Not one in ten would return.
Super cool quote! But my goodness, such fertile fields left fallow!
Of course, Umberto is trolling! He’s an extraordinary writer – highly skilled at making fiction appear real – and that’s what he’s doing in this case.
After all, the maze in which his characters find themselves and from which he trolls that their outcomes are pre-ordained by logic (even if they aren’t foreseen) is the maze of his own ideas, social beliefs and perceptions. It’s certainly not a maze of the laws of physics and by no means is it a maze that is independent of Eco himself. In fact, when he wants a character to turn left, he builds a wall to the right. In that sense, the logical outcome of the situations for his characters is still as much his creation as if he directed their every step.
In that sense it’s also clear why different outcomes emerge for social situations when the same prompt is given to different people: each person is an aggregator of whatever data generated by the universe reaches their sensory apparatus; each builds a unique model of the universe from their unique data set; each processes the prompt or premise through their unique model which in turn produces a unique set of logical consequences for characters and ultimately leads to different outcomes in the story. And of course, the individual models people build of the universe lean strongly on their upbringing and education, both of which are explicitly engineered to shape their model. Thus Andrew, Jessica, Bob et al may have unique models, but it’s likely their models would fit into a clearly recognizable group when compared to the models of, say, John D., Sam, Elon et al.
It would be useful for people involved in data analysis of social phenomena to understand the differences between how artificial models perceive the universe and how human models perceive the universe. Each has strengths and weaknesses. Artificial models (up until LLMs) perceive an infinitesimally small slice of the universe that is chosen, and usually engineered, by the modeler. That’s very similar to the characters in Umberto’s stories! OTOH, real humans perceive every feature or property of the local universe as they pass through it. They don’t have to name every feature or property or even know that it exists to feel its effects. Furthermore, humans perceive information, update models and respond to the world in real time, while artificial models can be days, weeks or years behind. Just recently I thought I’d give AI another try, so I asked it a question about the 2024 election. It’s not there yet!! Not far from where I live stands a large petroleum facility. Should I wait for the data on refinery fires and for models of their causes before I decide whether to bike past it on any given day, or should I just look for myself to see if it’s burning? Google? :)
These differences play out in real problems. Lately there has been a controversy over how well people are doing in the current economy: the general population seems to be unhappy with the economy; yet some economists claim that data indicates the economy is great and people are doing great, despite what they say in surveys. Who’s right? Economists, with data for maybe ten variables, most of which are at best approximated models (e.g., what’s the “true” value of inflation?) and which are further distorted by effects of aggregation and low granularity; or the general population, in which each individual perceives every aspect of the economy that impacts them? We have this controversey despite the fact that, as bad as it is, economic data is among the more reliable data in the social sciences.
There are *numerous* examples in social sciences in which a small number of variables, representing an infinitesimally small slice of the reality of daily human experience, are carefully engineered by the modeler’s beliefs and ideas to create a model that produces an outcome that may at first glance seem the outcome of logic, but even a cursory glance into the workings reveals how the walls of reality have been carefully constructed to produce the “logical” outcome for the characters. Not surprisingly, people with similar education and backgrounds may not see this very clearly, but people from outside can detect it with ease. We’re not talking about the movement of billiard balls on a table when struck; we’re talking here about, at best, barely percievable effects.
Happily, despite decades of trying, social science modelers aren’t nearly as talented or convincing as Umberto at faking reality! :)
I’ve always thought it was an amazing insight but it occurred to me today that it shouldn’t be surprising that people often write to know what they think. Can one create one’s dream house without drawing it? The imaginary world depends on the physical world for its existence.
I haven’t yet read The Name of the Rose, although it’s on my shelf. In that spirit, I’m thinking this post could have come with a spoiler or some editing of the posted postscript…