The Eindhoven University of Technology in cooperation with Delft University of Technology organizes a conference on Values and Norms in Modeling (VaNiM 2012) in June 2012 (25-27) in the Netherlands and invited me to talk about values and norms in the modeling of decision sciences.
I will build on the plenary talk "Mesure des Choses, Valeurs des gens" that I gave at the 2nd conference of the French society of the philosophy of science at ENS in Paris in 2009. I also intend to write these ideas in an academic format, now that most of the mathematical results with Bertrand Lemaire are published. It is for me a great opportunity to read the current research in the philosophy of sciences, and in particular about the new approaches to scientific representation, where the debates between constructivism and objectivism and between empiricism and realism are still very much alive. Below is the abstract that I have prepared.
Biased Representation of Subjective Values for Decision Models
To which extent can subjective values be represented in decision models? To address this question, I rely on the representational theory of measurement to clarify how empirical observation of preferences is represented by relations among numbers. I explain how this approach inherits from models in natural sciences and how decision models attempt to integrate subjectivity as properties of the objects of choice. I criticize this approach, claiming that it misrepresents subjective values and ultimately weakens decision models. I introduce an approach based on biased representations, where the structure of classical representations is embedded in a perspective which reflects the way objects are interpreted by the subject who chooses. The subject is treated as a "device" which measures things, and not as a property of the things that are measured. Building on the metaphor of the biased balance, on mathematical results in the theory of biased measurement and on examples about ethical values in business decision-making, I argue that biased representations can better model the specific articulation of objective and subjective values in decision sciences.
Click here the access the abstract of the VANIM 2012 conference