Presentation short abstract

We studied the computer models underlying Dutch flood risk management and observed that the growing complexity of the modelling ecosystem leads to fragmented expertise, reduced agency, and questions about (democratic) accountability for decisions based on computer models.

Presentation long abstract

In the Netherlands, a low-lying delta country, models underpin legally binding flood safety strategies and are, in this way, structurally involved in anticipatory governance. Over time, the ecosystem of models employed to manage flood safety has grown both in number and in detail. Based on interviews with modellers and decision makers involved with flood risk modelling for the Netherlands, it appears there is little overview of the whole chain of models necessary both to determine dike safety standards as well as models necessary to redesign dikes to meet these standards. We move from Millgram’s Endarkenment thesis, which argues that Enlightenment’s push toward reason and specialization has backfired, leaving individuals overwhelmed by fragmented expertise. In contexts of fragmented expertise, trust becomes indispensable: if one cannot judge a model’s appropriateness, one must defer to another’s judgment. Yet this dependence on another’s judgement has direct consequences for agency and accountability, which becomes diluted between those making the decision and those whose expertise must be trusted. We complement this analysis with Illich’s Tools for Conviviality, showing how technology can expropriate agency and eventually becomes manipulative: when tools become too complex to be questioned, they begin to shape decisions more than the people using them. Taken together, we question the democratic basis of model-based governance.