Daniel P. McDonald, Richard D. Gilson, Mustapha Mouloua, Jeremy Dorman, and Patricia Fouts University of Central Florida
Trained operators and the general public alike, are increasingly faced with alarms in their environment. Alarms must be salient enough to capture attention with the intent to redirect it toward desired cues in the environment. Second, alarm should be easily understood. The alarm should be designed to clearly communicate the intended information to the operator. The question is how do we facilitate this communication.
Alarms are often intended to indicate conditions not immediately detectable by the senses. In a sense, alarm systems provide replacement stimuli, designed to represent to us an undetected condition of interest. For example, a fire alarm informs people of the presence of a fire, before detection by human senses. This is desirable to promote safe exit from a building or home. Alarms often take advantage of cue salience to alert people. Highly salient auditory or visual cues can increase arousal, calling for immediate action. Many of these cues may share a natural association with potential danger for the observer. An alarm's association with urgency can be useful to motivate evasive behaviors. However, excessive salience can also prove to be detrimental. A response in haste can prove to be fatal. Some alarm, such as in complex systems, may call for consideration of several situational factors before making appropriate responses.
In addition to capturing attention, alarms are also used to convey information regarding the situation. For example, on a fire alarm, a flashing display indicating "fire" can help indicate the cause of the alarm. Previous experience with fire alarms may be sufficient for identifying the alarm, through association. Similarly, experience may also prescribe appropriate responses such as "exit the building". Alarms convey their messages through these associations made in the past, which may be accurate or inaccurate, depending upon many factors, some related to alarm context. In order to effectively communicate messages through alarms, we must understand better, how natural and learned associations, as well as contextual factors play a role in determining behaviors associated with alarms.
One characteristic associated with man-made alarms, dramatically affecting the message, is the potential for false alarms to occur. False alarms have been a concern in a number of applied contexts. For example Tyler, Shilling, and Gilson ( 1995) found that 5% (1450) of all aircraft hazard reports submitted to the Naval Safety Center from 1984 through 1994 indicated that false or erratic indications were at least partially responsible for a mishap. Anecdotally, experienced pilots may learn to ignore or even disengage an alarm which tends to be false or is perceived as irrelevant within the context ( Gonos, Shilling, Deaton, Gilson, & Mouloua, 1996). The plausibility of a false alarm reduces both the confidence in and responses to that alarm ( Bliss, Gilson & Deaton, 1995; McDonald, Gilson, Deaton, & Mouloua, 1995). This phenomenon, sometimes referred to as the "cry-wolf" or false alarm effect, may exert a powerful influence on how system anomalies are handled in the present and in future situations. Breznitz ( 1984) maintains that a result of false alarms is the loss of credibility of the warning system. Bliss, Gilson and Deaton ( 1995) found that an alarm's expected reliability influences the likelihood of a response to that alarm. Specifically, most people tended to probability match their responses to the expected probability of an alarm being true. These results appear similar to other matching behavior found with animals, where responses are made in direct proportion to the frequency of the reinforcement (see Herrnstein, 1970). Notably, probability matching is not optimal behavior from the standpoint of individual success.
During high workload, when faced with time pressure, presented with insufficient or ambiguous