Modeling R&D Strategy as a Network Search Problem
The search procedures themselves offer considerable scope for the development of models of R&D teams and how they function within organizations. Our own interests lie in the direction of designing logical formalisms to represent learning and emergent behaviour by R&D teams. Economists are more likely to be interested in the development of path algebras to optimize the search for technological developments. The choice of approach depends on how the analyst conceptualizes R&D. If the scope for technological developments is well understood in advance so that (for example) some kind of stochastic search algorithm is appropriate, then the economist's approach is appropriate. In cases where the universe of technological possibilities is not already known to the R&D team so that the information which could be found exceeds the information-processing and computational capacities of the agents, then the agents have to make use of heuristics. It is intended that the agents be given a language in which to express, learn and refine such heuristic search strategies. The difference here is the same as that between substantive and procedural rationality.
Our own modeling methodology relies on the declarative modelling techniques. A model is declarative if the current state of the world determines the actions of agents and the ways in which that state will be changed. In other words, declarative models specify reactions to (usually generally specified) states. Economic models, by contrast are typically imperative. They specify a subsequent state on the basis of a given state. In other words, imperative models specify a state to be in rather than an action to take or process to be undergone. Thus, dynamic imperative models yield sequences of states without specifying how they are achieved whereas declarative models specify the actions taken and changes made and it is these which result in subsequent states. In economics, each state is usually an equilibrium whereas in declarative models equilibrium states are the least interesting and are not in any case very likely since they leave no scope for the dynamics inherent in the model specification.
A further advantage of logic-based, declarative modeling techniques is that they enable the modeler to include qualitative considerations without loss of rigour. This will often be important in modeling R&D where the teams will often consider some potential avenues of development to be "promising" or "difficult" without having any numerical measure to represent these views. It is our practice to develop models which rely on qualitative judgements by agents in such a way that simulation outputs include numerical series which can meaningfully be compared with some statistical data series.*1
Generated with CERN WebMaker