Effects of Feedback Complexity on Dynamic Decision Making Export

@article{citeulike:3918128, abstract = {Prior research shows people suffer from misperceptions of feedback, generating systematic dysfunctional behavior in the presence of dynamic complexity - settings with multiple feedback loops, time delays, and nonlinearities. However, prior work has not adequately mapped the effect of these elements of complexity on performance. We report an experiment where subjects managed an inventory in the face of stochastic sales, a classic dynamic decision task. We vary the time delays and strength of the feedback loops to explore the impact of these elements of dynamic complexity on behavior. Subjects faced financial incentives and had opportunities to learn. Yet performance was significantly worse than optimal across all conditions. Subjects outperformed a naive "do-nothing" rule in the simple conditions, but performance deteriorated dramatically with increasing time delays and feedback effects, and most were outperformed by the do-nothing rule in the complex conditions. Regression analysis of subjects′ decisions showed most ignored the supply line of pending production and undercontrolled the system. Undercontrol increased significantly with growing time delays and feedback strength, showing subjects were insufficiently adaptive despite perfect knowledge of system structure and parameters. Subjects′ understanding of complex feedback settings declines as delays between cause and effect increase and as actions have stronger side effects. Few indications were found of active experimentation or learning: the need to control seemed to override the ability to learn.}, author = {Diehl, E.}, citeulike-article-id = {3918128}, citeulike-linkout-0 = {http://dx.doi.org/10.1006/obhd.1995.1043}, citeulike-linkout-1 = {http://linkinghub.elsevier.com/retrieve/pii/S0749-5978(85)71043-6}, comment = {Discusses an experiment that was conducted where students where asked to play with a generalised variable control model and try to maintain a desired production value in the face of a random walk demand with feedback loops to production as well as a time delay between actions and effects. The end results of this was that subjects, in the worst case scenario with high feedback and high time delay, on average preformed worse then a "no decision" baseline, tho it should be noted that some did up to 4 times better but still far from optimal. In conclusion the authors note that people's mental models of control tasks are highly simplified and tending to exclude side effects, feedback processes and time delays. Even when they are shown that these things are part of the model they are still not correctly absorbed into the model. The authors state that the first is a misconception that can be improved over time, but that the second is a fundamental bound on human rationality, however there is room for computer assistance in this area. }, doi = {10.1006/obhd.1995.1043}, issn = {07495978}, journal = {Organizational Behavior and Human Decision Processes}, keywords = {cognition, complex\_decision\_making, feedback, feedback\_delay}, month = {May}, number = {2}, pages = {198--215}, posted-at = {2009-01-21 00:12:40}, priority = {2}, title = {Effects of Feedback Complexity on Dynamic Decision Making}, url = {http://dx.doi.org/10.1006/obhd.1995.1043}, volume = {62}, year = {1995} }

TY - JOUR ID - citeulike:3918128 L3 - citeulike-article-id:3918128 N2 - Prior research shows people suffer from misperceptions of feedback, generating systematic dysfunctional behavior in the presence of dynamic complexity - settings with multiple feedback loops, time delays, and nonlinearities. However, prior work has not adequately mapped the effect of these elements of complexity on performance. We report an experiment where subjects managed an inventory in the face of stochastic sales, a classic dynamic decision task. We vary the time delays and strength of the feedback loops to explore the impact of these elements of dynamic complexity on behavior. Subjects faced financial incentives and had opportunities to learn. Yet performance was significantly worse than optimal across all conditions. Subjects outperformed a naive "do-nothing" rule in the simple conditions, but performance deteriorated dramatically with increasing time delays and feedback effects, and most were outperformed by the do-nothing rule in the complex conditions. Regression analysis of subjects′ decisions showed most ignored the supply line of pending production and undercontrolled the system. Undercontrol increased significantly with growing time delays and feedback strength, showing subjects were insufficiently adaptive despite perfect knowledge of system structure and parameters. Subjects′ understanding of complex feedback settings declines as delays between cause and effect increase and as actions have stronger side effects. Few indications were found of active experimentation or learning: the need to control seemed to override the ability to learn. IS - 2 JF - Organizational Behavior and Human Decision Processes SN - 07495978 EP - 215 TI - Effects of Feedback Complexity on Dynamic Decision Making VL - 62 SP - 198 KW - cognition KW - complex_decision_making KW - feedback KW - feedback_delay N1 - Discusses an experiment that was conducted where students where asked to play with a generalised variable control model and try to maintain a desired production value in the face of a random walk demand with feedback loops to production as well as a time delay between actions and effects. The end results of this was that subjects, in the worst case scenario with high feedback and high time delay, on average preformed worse then a "no decision" baseline, tho it should be noted that some did up to 4 times better but still far from optimal. In conclusion the authors note that people's mental models of control tasks are highly simplified and tending to exclude side effects, feedback processes and time delays. Even when they are shown that these things are part of the model they are still not correctly absorbed into the model. The authors state that the first is a misconception that can be improved over time, but that the second is a fundamental bound on human rationality, however there is room for computer assistance in this area. AU - Diehl, E PY - 1995/05// UR - http://dx.doi.org/10.1006/obhd.1995.1043 ER -