Enhancing decision analysis models with web-agents

Decision analysis (DA) can be enhanced by taking advantage of vast, real-time data available from the World Wide Web (the Web). Human intensive DA models such as influence diagrams may be linked with electronic agents that actively utilize the Web to generate data. At the time of modeling, results of agent actions are treated as stochastic events. Probability distributions are assessed conditioned on the range of outcomes for these events. When the DA model is evaluated the agent performs actions defined by the model in terms of the state of nature. Structuring links for these models presents technical challenges including programming and assessment. Agents can operate in the Internet as probes, sensors, monitors, beacons or in other roles, sometimes making information acquisition decisions. Certain managerial decision classes are especially well-suited to this approach.