DOES A HEAD-MOUNTED DISPLAY WORSEN INATTENTIONAL BLINDNESS?

PROCEEDINGS of the HUMAN FACTORS AND ERGONOMICS SOCIETY 50th ANNUAL MEETING—2006 1638 DOES A HEAD-MOUNTED DISPLAY WORSEN INATTENTIONAL BLINDNESS? St...
1 downloads 1 Views 137KB Size
PROCEEDINGS of the HUMAN FACTORS AND ERGONOMICS SOCIETY 50th ANNUAL MEETING—2006

1638

DOES A HEAD-MOUNTED DISPLAY WORSEN INATTENTIONAL BLINDNESS? Stas Krupenia ARC Key Centre for Human Factors, The University of Queensland, St Lucia, Queensland, Australia Penelope M Sanderson ARC Key Centre for Human Factors, The University of Queensland, St Lucia, Queensland, Australia Head mounted displays (HMDs) can present visual information to operators at times when this would otherwise be difficult or impossible using standard visual displays. HMDs have been shown to benefit anesthetists in simulated medical environments. Operators, however, may have trouble extracting information from the HMD in dynamic environments. Operators may also fail to consciously perceive visual events that are important, meaningful, or bizarre if attending to other aspects of the visual scene. We investigated how attention manipulations (Focused, Divided, Just Watch) interact with display type (HMD, Standard Display) and found that participants were less likely to detect unexpected events using the HMD. We also found that unexpected event detection decreased from the Just Watch to Divided to Focused attention conditions. We suggest further testing be taken to ensure that HMDs to not result in failures to detect unexpected events in anesthesia monitoring.

INTRODUCTION In safety-critical sociotechnical systems such as anesthesia or aviation, operators must integrate information from a variety of sources. Sometimes operators divide their attention between tasks and sometimes they focus attention on a specific task. Because of the spatial separation between visual displays and working positions, monitoring system status can be difficult. An alternative way of providing visual information to operators is through the use of Head Mounted Displays (HMDs), which are small transparent or opaque lightweight visual displays worn over one or both eyes. HMDs appear to be a good way of providing continuous visual information to operators when monitoring is difficult with standard visual displays. Although HMDs are beneficial to operators in some domains, a similar technology, Head Up Displays (HUDs) have been associated with failures to detect centrally occurring but unexpected events. This is a potential danger in safety-critical systems. In this study we used an inattentional blindness procedure to examine the relative influence of an HMD and attentional manipulations on participants’ ability to detect unexpected events. The study is the first in a series examining the use of HMDs for anesthesia monitoring. HMD advantages HMDs appear to be beneficial because two sources of information—the display and the real world—can be viewed simultaneously. Performance advantages with HMDs over standard visual monitors have been demonstrated in studies with

anesthetists in high-fidelity patient simulators (Ormerod, Ross, & Naluai-Cecchini, 2002; Platt, 2004; Ross, Ormerod, Hyde, & Fine, 2002; Via et al., 2003). The HMD increased the time anesthetists spent looking at the patient, decreased the number of times they switched attention to the standard visual monitor, and decreased the amount of time they spent handling critical patient events (Ormerod et al., 2002). Via et al. (2003) found that the HMD decreased the time anesthetists took to detect certain critical patient events compared to the standard visual monitor. Anaesthetists also report that HMDs are useful (Via, Kyle, Geiger, & Mongan, 2002). The results of these studies are encouraging because they suggest that HMDs can assist anesthetists in operating room environments and are valued by them. HMD and HUD disadvantages Despite the above encouraging results, other laboratory and applied studies have demonstrated that operators may have trouble extracting information from the HMD. Laramee and Ware (2002) found that users took longer to obtain specific data from an HMD when viewing the HMD image in front of a dynamic background image. They suggest that the HMD improves performance only with very stable background conditions that are rarely found outside a laboratory. A similar technology to HMDs is Head Up Displays (HUDs). HUDs project semi-transparent visual information over the users’ normal field of view. In an early study on the use of HUDs by commercial pilots performing simulated

PROCEEDINGS of the HUMAN FACTORS AND ERGONOMICS SOCIETY 50th ANNUAL MEETING—2006

1639

landings, Weintraub, Haines and Randle (1985) found that only two of eight participants using the HUD detected unexpected runway incursions. Similar failures in detection have been reported in more recent studies. Wickens and Long (1995) found that pilots using HUDs took longer to respond to unexpected real world events and Hofer, Braune, Boucek and Pfaff (2001) found that even when pilots knew that an event was to occur in the external environment, 36.5% of runway incursions went unreported.

effect on detection, so that participants in a Focused Attention condition will be least likely to detect the unexpected event, followed by participants in a Divided Attention condition and then by participants in a Just Watch condition. Finally we predicted that the combined effect of Focused Attention and an HMD might be hyper-additive, leading to a statistical interaction.

Controlled examination of possible HMD disadvantages

Participants

If the above results generalize to HMDs, then there may be cause for concern with using HMDs for anesthesia monitoring. Unexpected events in the OR environment that have an impact on patient safety (e.g. a patient movement, a wrong drip fluid set up on a pump, a drug swap over, etc) may not be noticed if the anesthetist is wearing an HMD. We have embarked on a series of laboratory-based experiments to see if we can replicate with HMDs the relative disadvantages seen with HUDs. Our initial investigations are being performed with a version of an inattentional blindness procedure developed by Simons and Chabris (1999). Simons and Chabris showed participants a videotaped scene in which two teams of three people each pass a ball between them. In a representative manipulation, participants are asked to count the number of ball passes between members of the team wearing white clothes, ignoring the team wearing black clothes. At 45s into the videotape, a man in a black gorilla suit walks through the teams for about 5s. Only 42% of participants notice the gorilla and report it. Thus if attention is directed to one aspect of a visual scene, then observers may fail to consciously perceive other visual events that are important (Weintraub, Haines, & Randle, 1985), meaningful (Mack & Rock, 1998) or bizarre (Simons & Chabris, 1999). Given these results, and further results by Mack and Rock (1998) using a different experimental procedure, we know that the proportion of observers seeing an unexpected event is sensitive to manipulations of (1) how the event is presented (2) physical properties of the event stimulus (e.g. color, etc) and (3) where attention is directed. Accordingly, in the experiment reported here we investigate whether unexpected events are missed more often when part of the display is viewed through an HMD, and how strong this manipulation is compared with manipulations of the observer’s attention.

This study was conducted with 155 volunteers from the general public (55 men and 100 women). Most participants were students at The University of Queensland and all volunteers received either a chocolate or coffee in return for their time. Participants were between 17 and 59 years of age (mean age was 23.8 years) and all provided written informed consent.

Hypotheses Based on results from Laramee and Ware (2003) and the aviation studies of Weintraub et al. (1985) and Wickens and Long (1995) we predicted that fewer people will detect the unexpected event in the HMD condition than in the standard display condition. Additionally, based on results from Mack and Rock (1998) we predicted that attention would have a strong

METHOD

Stimuli and Apparatus Participants completed a monitoring task where three red squares, three black squares, a red ball and a black ball were presented using differing display formats. The shapes subtended a visual angle of 2 degrees and were presented against a background scene of an elevator lobby. Squares were 100 x 100 pixels in size and the circles had a diameter of 100 pixels. The shapes moved around the screen and bounced off invisible boundaries. Additionally, the red ball bounced only off the red squares and the black ball bounced only off the black squares. Red and black shapes passed through each other with the red shapes appearing in front of the black shapes. For 300 ms following a ball bounce with a wall and for 750 ms following a ball bounce with a square, the ball would not bounce off any other object except another wall. The monitoring task lasted for 75 seconds and the correct number of ball bounces was 22. Participants completed the monitoring task at different levels of attention (Focused Attention, Divided Attention, Just Watch) using a combination of display types (standard display, standard display and HMD). The HMD used in this experiment was Microvision’s NOMAD 2100ND, a 32 shade monochrome red display with 600 x 800 SVGA resolution. For the standard display condition, all the shapes were presented on a standard computer monitor (Acer Aspire 1700, and screen resolution of 600 x 800). For the HMD condition, the red shapes were presented on the HMD and the black shapes with the background were presented on the computer monitor (see Figure 1).

PROCEEDINGS of the HUMAN FACTORS AND ERGONOMICS SOCIETY 50th ANNUAL MEETING—2006

1640

Procedure

Figure 1. A schematic representation of the display manipulations. Left panel shows red and black shapes on the computer screen. Right panel shows black shapes on computer screen and red shapes on the HMD. Instructions were as follows: • Participants in the Focused Attention condition were told “Keep a silent mental count of the number of times the red ball bounces off a red square.” • Participants in the Divided Attention condition were told “Keep a silent mental count of the number of times the red ball bounces off a red square. At the end of the trial, you will also be required to report any other event that might occur during the trial.” • Participants in the Just Watch condition were told “Simply watch the shapes moving on the screen.” For all conditions, the same unexpected event occurred. A square segment of a gorilla’s face (100 x 100 pixels), including its eyes, nose and part of the mouth appeared on the standard display 45 seconds after starting the monitoring task and took five seconds to move from the right to the left hand side of the screen and then disappear (See Figure 2).

Upon arrival, participants were given information about the experiment and written informed consent was obtained. Participants were asked for some background information and a short test was given to assess eye dominance. Then the experimenter read the instructions to the participant. For HMD trials, participants were given time to become familiar with the display. Participants could adjust the HMD so that it was comfortable, and focus the display so that they could look from the standard display to the HMD without having to reaccommodate visually. All participants used a chin rest for the experiment and were asked to remain still and to keep both eyes open for the trial (see Figure 3).

Figure 3. Photo of participant completing HMD condition. Following the monitoring task, participants were asked a series of questions adapted from Simons and Chabris (1999): 1.While you were doing the counting, did you notice anything unusual on the screen(s)? 2.Did you notice anything other than eight shapes? 3.Did you see anything else besides the six squares and two balls appear onto the screen? 4.Did you see a gorilla’s face move across the screen? If participants answered ‘yes’ to questions one, two or three, the remaining questions were skipped. Finally, all participants were also asked if they had ever previously participated in an experiment similar to the current one, if they had ever heard of such an experiment, and if they were familiar with the phenomenon of inattentional blindness. RESULTS

Figure 2. Screen shot showing moving shapes and unexpected event (at bottom of screen towards centre).

Of the 155 participants tested, the data from 15 participants were excluded from the analysis either for failure to correctly understand the instructions (n=7), prior familiarity with the experiment (n=6), or for other technical failures (such as a

PROCEEDINGS of the HUMAN FACTORS AND ERGONOMICS SOCIETY 50th ANNUAL MEETING—2006

wireless HMD signal loss; n=2). The final sample size analyzed was 140. A breakdown of participants per condition and detection rates is presented in Table 1. Table 1. Number of participants who saw or did not see the unexpected event for the six conditions tested. Focused Watch Divided SD HMD SD HMD SD HMD See 10 6 23 19 10 10 Not see 13 18 1 5 12 13 Total 23 24 24 24 22 23 The mean number of ball bounces counted (with standard deviations in parentheses) for the Focused Attention and Divided Attention conditions was 20.98 (2.10) and 20.74 (2.12) respectively. When participants in the Just Watch condition were asked to guess the number of ball bounces after viewing, the mean response was 19.71 (19.20). Across all conditions, 78 participants reported seeing the unexpected event and all except three participants reported seeing the event in response to the first question. Two participants responded ‘no’ to the first question and ‘yes’ to the second question and one participant responded ‘yes’ to the third question. A backwards elimination log linear analysis of n-way effects was conducted to examine main effects and the interaction between display type and attention. Simultaneous tests for the nway effects of zero revealed no significant two-way interaction, Χlr2 (2) = 2.225, p = .3288, and at least one significant main effect, Χlr2 (7) = 39.093, p < .0001. Follow-up tests for partial association revealed a marginally significant main effect of Display Type, Χlr2 (1) = 2.885, p=.0894, and a significant main effect of Attention, Χlr2 (2) = 34.440, p