I am fond of saying to myself that science is not like art, it is art. However, upon hearing the late Iris Murdoch say that the aims of science and literature are divergent, in that science seeks to clarify whereas literature seeks to mystify, I have been forced to reevaluate my statement. Apparently, some (much?) of modern prose falls under the category of mystification. To my unsophisticated mind, that is sad.
Anyhow, to be honest, I haven’t read a lot of literature to be in a good position to evaluate Iris statement. The few writers that I have read, I daresay I have read them more completely and devotedly than many people. Chief among them is Leo Tolstoy. Whether or not science is art, or like art, what is true is that Tolstoy was a prime practitioner of the scientific method in art. Let me explain.
The major reason I was so attracted to Tolstoy was his power to explain and clarify, to remove the dust from the canvas so to speak; and lastly to set up experiments, the aims of which appeared meaningful and results of which seemed credible. For me it wasn’t the beauty of Tolstoy’s prose, it wasn’t the breadth of his canvas, none of the other things that he is so famous for. My attraction to him was simply his scientific approach, which comes down to analyzing facts honestly and using imagination to set up meaningful experiments. The reader should take note that whatever Tolstoy himself said on this and related subjects is irrelevant! While this statement might appear to be greatly presumptuous, those who have dealt with scientists and have tried to understand the ‘scientific method’ will do well to remember Einstein’s profound words on this subject: “Don’t listen to their words, fix your attention to their deeds”. Noam Chomsky has also echoed such sentiments. Indeed, Einstein’s words hold true for all self-professed “Popperian” scientists whose scientific method is generally, quite rightly, anything but Popperian (For those interested, see Nigel Gilbert and Michael Mulkay’s illuminating paper, “Putting Philosophy to Work: Karl Popper’s Influence on Scientific Practice”).
Tolstoy’s task was harder than that of a modern scientist. Tolstoy had to do both the theory/model and experiment parts on his own. In this respect he resembles the present day agent based modeler.
Let’s consider the modern day agent-based modeler, (the author himself being an insignificant one), whose task is similar to Tolstoy’s. Many areas for which we make models don’t have any theories. Those that do, have either been debunked (think economics); or they verge on being pseudo-science/common-sense. So while we are making our models and setting them up for computer experiments, we are usually not driven by any particular theories/frameworks (as opposed to say physicists). We base our models on a combination of intuition, experience, common sense, and, if possible, clues from existing theories, subject to their state of ‘well-being’ and maturity. In this respect, agent based modeling is pretty much at the same stage as pre-Galilean physics. At the moment we are struggling, let’s admit it, to set up framework(s) that can allow us to ask meaningful questions. There is no shame in that. The struggle is worthwhile and should not discourage anyone. The only point of caution is to indeed be aware that we are in the pre-Galilean age.
Coming back to Tolstoy, why do I say that he was the supreme agent based modeler. For that we would have to consider what exactly is it that we do in agent based modeling and then compare that to what Tolstoy was doing. Broadly speaking, the following is true of agent based modeling:
a) Our experiments are based on a model. If the state of the science in the target area is mature enough, agent-based modelers should use concepts and ideas from pre-existing theories/frameworks to design their models. The pervasive use of the Prisoner’s Dilemma serves as a good example. Most of the time, however, since we are pre-Galilean, we are working from common sense, intuition and experience.
Tolstoy for his part had his own models, partly informed by history, partly guided by the philosophy of his day, and partly based on his experience and intuition. For example, in War and Peace, Tolstoy’s chief aim was to present a new model of history; a people’s model of history so to speak, as opposed to the history of ‘great’ men. He was railing against Carlyle’s hero, Superman and other such tendencies that glorify ‘great’ men. To Tolstoy, even the most ‘inconsequential’ person was important and everyone made a difference. As the great Urdu poet Muhammad Iqbal said:
“Even a particle in its place is as powerful as the sun”
Thus, Tolstoy’s entire design for War and Peace was guided by his model of ‘people’s history’. Ever wonder why there are so many characters in that book? Certainly, Tolstoy doesn’t depict them in order to show off his powers of observation or his craft. No. It would also be wrong to assume that the length of the novel necessitated the enormity of the number of characters. If this were the case, then “Brothers Karamazov”, the magnum opus of Tolstoy’s famous contemporary, Dostoevsky, would have boasted a comparable number of characters. But it has fewer characters by as much as an order of magnitude. In fact, Tolstoy’s underlying model has dictated the number of characters in War and Peace: to test the importance of each character—to test the model.
b) In agent based modeling, we set up agents with some properties. For choosing these properties, we use abstraction and idealization. For example, in an agent-based model that seeks to model the spread of AIDS in Africa, the color of the pendants that agents wear is abstracted away. Tolstoy did the same with his characters. He was a natural at the Galilean-Newtonian style, and he seldom spent long paragraphs (let alone speeches, like his other illustrious peer Dostoevsky) describing every facet of the character. Often single sentences would be enough; broad strokes, the kind of which are required in science. Scientists are aware that the more detailed one gets, the less one can understand and speak about the matter. If one tries to go into too many details, one ends up being not a scientist, but rather a mere describer of facts. The aim is to understand, not to describe. That is why we abstract; that is why we idealize; and in short that is why we use the Galilean-Newtonian style.
Consider an agent-based modeler describing an agent whose properties can be either loving or spiteful. (Let’s assume that the model deals with agents interacting to form cooperative groups). In an agent based model, we would simply write:
Agent name = Vera
Vera temperament = “spiteful”
Only two lines of computer code (backed up by some model, informal or otherwise, that encompasses such concepts as loving and spiteful) and Vera has easily been designated to be a spiteful girl.
The following is how Tolstoy has done it in War and Peace: “The handsome Vera, who produced such an irritating and unpleasant effect on everyone, smiled and, evidently unmoved by what had been said to her, went to the looking glass and arranged her hair and scarf. Looking at her own handsome face she seemed to become still colder and calmer.”
That’s it. No long sermon about her childhood traumas or indeed nothing of the converse either, i.e., explaining how all children in the family were brought up the same, but since nature trumps nurture, hence Vera was spiteful while Natasha was kind and loving. No, nothing of the sort.
One can only imagine how someone like Dostoevsky might have described Vera. In fact one need not imagine but simply read any of the long sermons and speeches that Dostoevsky makes his characters deliver. Of course, this is not to belittle the great man. It just serves to illustrate that Dostoevsky did not have a scientific approach to his art.
Examples such as the above abound in Tolstoy’s novels and short stories, where he simply uses a few broad strokes to completely describe a character/agent, for the purpose at hand. The mistake that people can make while reading Tolstoy is to think that he describes everything in detail. Not true. He only describes things, whatever they might be: objects in a room, facets of the personality, insects sleeping inside flower buds, etc., that are needed to be described for the purpose at hand. It is a fundamental fallacy to assume that Tolstoy is trying to hold a ‘mirror to reality’, as has so often been claimed. He does nothing of the sort! Instead, he uses abstraction and idealization. Tolstoy had a scientific temperament first and foremost. He understood that holding a mirror to reality is probably impossible and more importantly, even if it were possible, it would be quite fruitless and would not bring him any closer to understanding, his chief aim as the supreme scientific artist.
(Young agent based modelers take note: an agent-based model (more often than not) is not supposed to mirror reality. Always ask yourself, what is it that you want to achieve with your model.)
c) Finally, the third thing that we model is inter agent interaction and interaction between the agents and the environment. We need some ground rules for both kinds of interactions. These depend on the agents’ (and the environment’s) properties and also other explicit rules set by the modeler. As a result of these interactions, agents learn and evolve, or at any rate, they change and do not remain static.
About the evolution of Tolstoy’s ‘agents’, instead of saying anything myself, let me just quote a writer zillion times better than me, the great George Orwell. While discussing Tolstoy’s characters (as he differentiates them from Dickens characters), Orwell says: “…he is writing about characters that are growing. His characters are struggling to meet their souls, whereas Dickens are already finished and perfect”.
Let’s not forget that this is where Tolstoy’s work (or the work of any other writer who follows this method) was far more difficult than the present day agent based modeler. Once we, agent based modelers, describe all the agents and rules of interaction, all that remains for us to do is to spin the wheel and let the computer crunch the results for us. Admittedly, understanding and interpreting the results is an art in itself. But at least the computer produces the results. In Tolstoy’s case, he had two weapons for running his experiments: his imagination and his honesty: to imagine what would be the effect of x on y and to honestly interpret the results (even if the result did not come up to his expectations). It can be safely said that Tolstoy was supreme on both counts.
To conclude, Tolstoy used concepts from, and based experiments on, intuition, experience, imagination, and also existing theories where possible; he used idealization and abstraction; and last but not the least, he dealt with evolving entities. In short, he was an excellent agent based modeler, even without the help of a number crunching device!
Here, I would like to pose an important question to all of us, whether agent based modelers, complex systems theorists or proponents of ‘simulation based science’. The question is: What we do is certainly meaningful, but shall we ever move beyond what Tolstoy was already doing 150 years back, and enter the Galilean age of our science? Is it even possible?
In the meantime, happy modeling!