Controlling Stable and Unstable Dynamic Decision Making Environments

Controlling Stable and Unstable Dynamic Decision Making Environments Magda Osman (m.osman@qmul.ac.uk) Centre for Experimental and Biological Psychology, Queen Mary University of London, London, E1 4NS UK Maarten Speekenbrink (m.speekenbrink@ucl.ac.uk) Cognitive, Perceptual and Brain Sciences, University College London London, WCIE 6BT, UK using a laboratory simulated complex dynamic task environment. Uncertain Dynamic Environments: Often when determining the outcome in complex dynamic environments a series of inter-related decisions are made (Brehmer, 1992). That is, a future decision builds on the outcome of a previous decision and so on in order to work towards a goal. For instance, if we decide to take a couple of aspirin when we have a headache, we know that there is a variable delay in taking effect, and that the intensity of headaches changes over time. If after some period the headache persists, we may decide to take more aspirin, but without being sure that it will take effect, and if so, when it will do so. In this case, our decision making requires a series of choices to act towards achieving a specific goal (alleviating the headache), but there is uncertainty attached to our choice of actions (when to take aspirin, and what dosage), because we cannot be sure we will reliably produce the desired effect. Typically, people are required to interact with an environment by deciding from various cues (e.g., Drug A, Drug B, Drug C) actions that are relevant (e.g., selecting Drug A at dosage X) to changing the outcome (e.g., reduce the spread of disease). To introduce complexity into the task environment, the cue-outcome associations are probabilistic, and the environment dynamic, which ensures that from trial to trial the effects on the outcome will change. Moreover, it encourages people to adapt their decision making in order to Many have used complex dynamic control tasks (CDC) as a way to examine the effects of varying the specificity of the goal under which the individual is instructed to learn about a complex environment (Burns & Vollmeyer, 2002; Geddes & Stevenson, 1997; Miller, Lehman & Koedinger, 1999; Osman, 2008; Vollmeyer, Burns & Holyoak, 1996). In this way, it is possible to examine the best conditions under which to learn to control an uncertain dynamic environment. Much of the evidence suggests that people are able to control complex systems successfully after sufficient opportunity to explore the environment first. However, training them to learn to control the system to a specific criterion can impair their ability to successfully develop flexible knowledge of the system that they can transfer to a different goal structures (Burns & Vollmeyer, 2002; Osman, 2008; Osman, 2010a). Abstract In the present study we ask: Are people sensitive to the stability of a dynamic environment under short exposure to it? To examine this we investigate people’s cue manipulation and strategy application when instructed to learn to control an outcome in a dynamic system by intervening on three cues. The system was designed in such a ways that in the Stable condition participants controlled an outcome that fluctuated steadily overall 40 trials, and in the Unstable condition the outcome fluctuated erratically over 40 trials. In the present study we show that people tended to intervene more frequently on all three cues when the system was Unstable compared to the when the system was Stable. Overall, the evidence from this study supports the general prediction made from the Monitoring and Control framework (Osman, 2010a, 2010b). It claims that people are sensitive to the underlying stability of dynamic environments in which they are required to control the outcome, but are insensitive to autonomous characteristics of the system. Keywords: Dynamic; Control; Prediction; Decision making Introduction Complex dynamic environments come in many flavors (Cohen, Freeman, & Wolf, 1996; Klein, 1997; Lipshitz, Klein, Orasanu, & Salas, 2001; Lipshitz & Strauss, 1997), such as economic (e.g. stock exchange), industrial (e.g., chemical waste disposal), critical safety (e.g., automated- pilot systems) and biological (e.g., eco-systems). These situations differ from each other for a host of reasons, but crucially they share two fundamental features: they are dynamic and they are autonomous. That is, the outcome (e.g., state of the environment) fluctuates over time, whether rapidly (e.g., a sudden down pour of rain in an otherwise sunny day) or relatively slowly (e.g., steady increase in temperature over the spring months). Additionally, in both cases, changes in the outcome can occur independently of direct interventions made by decision makers. Given the probabilistic properties of these environments, an action may not reliably produce the same outcome each time, which raises the question: What are the differences in learning behaviors when attempting to control a highly noisy environment as compared with attempting to control a less noisy one? The aim of this study is to address this question in detail by examining control-based behaviors