De Landa – Synthetic Reason and Computational Simulation

DMF recently posted this video to Deterritorial Investigations. In it Manuel De Landa discusses various computational simulation techniques and how they influence scientific thought. Having not read De Landa in a while, it was refreshing to return to his mode of thought, and, in fact, find some parallels with my own research. Specifically, there are some similarities between what he describes as “synthetic reason” and the way that I found researchers to be using computational models in a scientific context.

Part of what De Landa is arguing here is that computer simulations function as a new kind of synthetic reason, and a new way of understanding the world we live in (I gather that this line of thought is developed in his recent book  Philosophy and Simulation: The Emergence of Synthetic Reason, which I have not read). Instead of simply observing behavior in the natural world and drawing assumptions from those observations, simulations allow us to synthesize the conditions of our assumptions and test them in the space of a limited world. This can then lead to new insights and intuitions about the world that we can apply to our observations in order to develop new assumptions. The idea is encapsulated when he says:

“… what virtual environments are making possible is precisely to synthesize more and more ideas in that direction, to allow the machine to be kind of a probing head into a world that we don’t see because we have these blinders on.”

This synthetic approach was similar to the kind of science I encountered in my own research on the Chesapeake Bay watershed. I’ve talked about it in another post, but essentially, researchers were using models as tools for exploring aspects of the watershed that they can’t observe directly for various reasons, and for examining the limits of their understanding (i.e. where the models are wrong). The result is a continual feedback between simulation and empirical data that accelerates and intensifies an engagement with the natural world. When models are part of this kind of feedback, I think they can be very productive tools for appreciating the complexity of ecological systems and recognizing the limits of our ability to understand them. This is important to understand because computer modeling is often criticized for being too abstract, technological, and reductionist.

On the other hand, models can also reinforce our predefined assumptions, particularly when they are displaced from the simulation-data feedback. This was the case with the application of computational models to environmental management. That’s not to say that the models used were inaccurate or not scientific, but that the management context is displaced from the feedback system in which the models are constructed. In those cases, the models are often modified to make them more useful for the management staff – incorporating things like costs and benefits of various management approaches, and simplifying the model to make them easier to run quickly and easier to understand for non-modelers. As a result, certain assumptions get built into the models that simply reinforce the assumptions of the broader management context. In the Chesapeake Bay watershed, this meant that the models reinforced the neoliberal approach structured by the Total Maximum Daily Load (TMDL) method (costs and benefits, best management practices, etc.). As a result, alternative forms of engagement with the environment and non-quantitative ways of valuing environmental benefits are ignored. This leads into a whole other critique of the neoliberal management approach, which I won’t address here.

So essentially, my findings support De Landa’s synthetic reason argument to an extent. I think it’s important to recognize that context matters. Simulations can be really valuable tools for understanding the natural world and for breaking through our ingrained assumptions, but only when they are part of a broader context (e.g. science) in which that kind of work takes place. When removed from that feedback process (e.g. management), models can often serve to reinforce and replicate our ingrained assumptions, further preventing us from seeing alternatives and recognizing our limitations. These are things we need to be aware of and on the lookout for as modeling becomes increasingly prevalent in environmental science and management.

2 thoughts on “De Landa – Synthetic Reason and Computational Simulation”

  1. there are a lot of critiques out now about biases in algorithms that could be helpful to some degree with yer project but what they tend to ignore is how keep management and employees from falling into the tyranny of the means, how to institutionalize reflexivity in ways which fit in with other goals of productivity and the like, that would take a lot of field work and experimentation which academics only generally can access at their own universities but by and large have not. Would be interesting to do some studies for how anthro depts are run and how they could be otherwise but I don’t see faculty getting on board with such adjustments in their working lives.

Leave a Reply

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