Fast and Frugal Operators Sub-Optimally Adapt To Machine Failure

Fast and Frugal Operators Sub-Optimally Adapt To Machine Failure Robert J. Youmans (ryouma1@uic.edu) Stellan Ohlsson (stellan@uic.edu) Department of Psychology, 1007 W. Harrison St. University of Illinois, Chicago Chicago, IL 60607 USA Abstract One safeguard against instrument malfunction is to provide backup instruments for machine operators. In previous studies, prior training caused operators of a simulated machine to adapt to instrument malfunction by adopting a suboptimal decision rule rather than by reallocating attention to backup instruments. One hypothesis for these findings is that operators do not notice when their main instruments malfunction. Here we examine warning systems that force operators to notice instrument problems. Our results indicate that warnings did not help operators to reallocate attention to backup instruments. Instead, operators fail the simulation and make sub-optimal adaptations afterward that lead to further failures. A Simulated Machine Interface In our simulation, participants assume the role of the operator of a juice factory. Two Holding Tanks, tank A and tank B, were shown on the upper left side of a computer screen, connected with pipes to a Mixing Tank shown to the right; see Figure 1. On the lower half of the screen was the gauge equivalent of the color information. Here, three realistic looking temperature gauges representing tanks A, B, and the Mixing Tank were displayed; see Figure 1. Introduction The operation of complicated machines often requires a flexible human operator that can adjust the operations of a machine to meet task demands. This is especially so because of the inherent fallibility of complicated machines; people must react to normal feedback from a machine in order to operate them as intended, and must react to abnormal feedback from machines when a corrective action is required. In prior work, we documented how operator performance suffered when instruments that an operator had been trained to use suddenly begin to provide inaccurate information, even when a second, valid instrument was available that could correct the error (Youmans & Ohlsson, 2005). The finding suggested that machine operators have difficulty switching from their usual instruments to a secondary or backup source of information. Why do operators fail to utilize seemingly obvious secondary instrumentation when the primary instruments that they have been using malfunction? One explanation is that training produces biases and automaticity that might interfere with rapid adaptation to changing task demands. Although quickly switching from one task set to another might subjectively seem to progress smoothly and effortlessly, evidence strongly suggests that switching between even simple task sets can be quite difficult (e.g., Allport, A., Styles, E. A., & Hsieh, S., 1994). To what extent are people limited by prior experience when faced with the need to adapt to changing task conditions? We investigate these questions with the help of a simulated human-machine interface in which the degrading of one set of instruments poses a need to re-allocate attention. Figure 1: Example of factory interface. Note. Factory interface was in color. Each Holding Tank contains liquid at a certain temperature. The factory is operated by adding some amount of liquid from tank A and some amount from tank B into the Mixing Tank. The amount and temperature of the juice is determined by the amount and temperature of the previous content of the Mixing Tank, the added input from Tank A and the added input from Tank B. Once a participant entered these amounts, the simulation was animated; the colored liquid was shown flowing through pipes into the Mixing Tank, and the Mixing Tank’s color and gauge responded appropriately to the new input. Once the two inputs were added, the resulting state of the Mixing Tank was computed and displayed 1 , and the operator could make the next decision about how much liquid to add from either Holding Tank. The task of the operator was to maximize the production of juice without overheating the facility, a type of trade-off situation. As shown in Figure 1, the display was divided into two sections by a thick gray bar. Above the bar is the section that TEMPcurrent = [(14 * TEMPprior.) + (7 * TEMPa) + (7 * TEMPb)] / 3, rounded to the closest whole value.