علامت جهره :: تشخیص، حواس پرتی و وظیفه

Abstract In this paper, we describe an empirical investigation of the utility of several perceptual properties of motion in information-dense displays applied to notification. Notification relates to awareness and how dynamic information is communicated from the system to the user. Key to a notification technique is how easily the notification is detected and identified. Our initial studies show that icons with simple motions, termed moticons, are effective coding techniques for notification and in fact are often better detected and identified than colour and shape codes, especially in the periphery. A subsequent experiment compared the detection and distraction effects of different motion types in several task conditions. Our results reveal how different attributes of motion contribute to detection, identification and distraction and provide initial guidelines on how motion codes can be designed to support awareness in information-rich interfaces while minimizing unwanted side effects of distraction and irritation.

Introduction Users typically function in multi-task environments in which information is distributed across windows and applications and is not necessarily exclusive to the task at hand. For example, a financial analyst may be monitoring stock market quotes while reviewing a client's portfolio and evaluating performance patterns over time. Simultaneously, she may be cued as messages to do with office administration arrive from her colleagues. Alternately, a telecommunications manager may be planning scenarios for equalizing phone traffic across variably loaded channels, while sporadic alarms indicate overloads on current routes. In both cases, these users are being made aware of dynamic information outside the specific scope of the data they are using for their current tasks. In some cases, certain types of dynamic information are contained in a dedicated display which the user must constantly monitor, such as a stock ticker or message flag. These displays are typically located on the periphery of a screen (Czerwinski et al., 2000; Maglio and Campbell, 2000; McCrickard, 2000). In other cases the changing information can be located anywhere in the visual field, such as mode information or element state directly tied to data objects in the displays (Sarter and Woods, 1995; Mitchell and Sundstrom, 1997). Monitoring dynamic information can be a cognitively strenuous task which requires the user to examine the currently displayed information and decide whether it has changed, so it is preferable to explicitly alert the user to a change by a signal. Signals are graphical events which indicate to the user that something has happened in some area in the display. Replacing an “empty mailbox” icon with a “full mailbox” to show e-mail status or animating the transformation of old text into new in a peripheral display ( McCrickard, 2000) are examples of signals. Signals are incorporated into peripheral awareness tools ( Maglio and Campbell, 2000; McCrickard, 2000), messaging ( Parsowith et al., 1998; Cutrell et al., 2000), state changes, system events or alarms ( Adams et al., 1995). They can also be used as navigation markers or guides to dynamically emphasize relevant points in a display. Current information visualization interfaces rely heavily on graphical coding devices (also termed display dimensions) such as shape, colour, size, texture, orientation and position ( Ware, 2000). These schemes can be very effective in enabling information analysis because they are mentally economical ( Woods, 1991; Healey et al., 1995): rapidly and efficiently processed by the preattentive visual system rather than attentive effort. However, only a small amount of information can be encoded in each visual dimension. For example, a typical recommendation is that no more than eight colours be used to define information categories ( Shneiderman, 1986; Gilmore et al., 1989). For this reason there is a shortage of perceptually efficient codes than can be used in information-rich user interfaces. One promising way of visually coding information is to use simple motion. Motion has a unique ability to attract attention over a large visual field and offers a rich graphical vocabulary. Its use has only recently become feasible due to the advent of fast graphics processors and supporting software technologies. However, compared with the use of colour coding, which is supported by a large literature of design guidelines based on decades of experimental studies, there has been little research relating to the effective design of motion codes. Such work is urgently needed because available technologies such as Javascript and image animation have led to a riot of moving and jiggling icons that compete for our attention. The notification studies described in this paper investigated the effects of moving icons, which we call moticons, as alerting mechanisms in situations where the user is engaged in a primary task and needs to be made aware when an event occurs outside the task area.

Conclusions and discussion These findings have a number of implications for the use of moticons as signals in human–computer interfaces. They suggest that motion has several advantages as a notification mechanism. It is significantly better than the traditional static codes of colour and shape in designing icons used to attract a user's attention, especially in the periphery. Our results showed that the percentage of undetected targets increased dramatically from 6% to 25% with the peripheral colour targets, whereas the failure-to-detect with motion was less than 2% in both the near and far field. Identification of the moving target was also substantially better with the moving cues than with the static cues across the entire visual field. These results once more emphasize the effect that location has on static and moving cues. Motion seems to be equally efficiently “remembered” in both near and far conditions, while colour and shape are less well noticed and less accurately tracked. The poor performance of the colour and shape cues even in the very sparse experimental displays is particularly noteworthy, as these cues are the most commonly used signals in current computing environments. In complex visualization and control systems where notification is of great importance, the poor detectability and identifiability of these cues argues for much more perceptually distinct signals. Even small, slow linear oscillations of the types reported here are excellent candidates for such notifiers. They have the advantage of being computationally cheap and consume little spatial and temporal resources. Perhaps the greatest advantage of using motion-based signals, however, is that they do not seem to interfere with existing colour and form coding, allowing extra information to be communicated through an object without changing its original codes which represent other variables. The high rate of detection even in the more engaging tasks suggests that motion is effective over a wide range of locations, types and amplitudes. Even the least efficiently detected cue, the slow blink, had a worst-case mean response time of less than 3 s and was detected 89% of the time within the 10-s window, indicating good accessibility even when the primary task is demanding. As one reviewer stressed, however, participants were expecting cues, and thus the detectability of these cues may vary (although we believe the relative performance will remain the same). However, we did find differences in the ease with which people could be distracted from the primary task, although the results were not exactly what we anticipated. Although we had expected that Tetris would be the task most engaging of attention and our participants anecdotally said that it was, the objective measures did not support this. Indeed the measured response times were longest for Solitaire, suggesting that this demanded the most resources. However, it must be emphasized that we did not measure primary task performance but instead signal recognition. In any case, our results show that even with highly demanding tasks motion can be readily used as an alert. They further suggest that such motion cues may be used to gauge task engagement. Finally, while all the tested motions are effective as signalling mechanisms, some are clearly more distracting to the user. Travelling motions which involve both detection and tracking are substantially more distracting than anchored motions. The zoom motions are also (although less) distracting, probably because they elicit sudden perceptual onset (Hillstrom and Yantis, 1994). These findings confirm our experience that animated banners and popping images are not comfortable visual elements on a screen where one is trying to work but are effective if one, in fact, wants to dominate the user's attention. Overall the slow linear motion would appear to be a good compromise. It was rated among the least irritating and distracting, but it elicited good response times and detection rates. We note that it is not confirmed whether these effects are due to the low-level event of a user's attention being grabbed by onset or motion, or the result of a slightly higher-level visual search strategy in which the user polled the display every few seconds. Users are very good at learning the distribution around cues in experimental conditions and adjusting polling strategies accordingly.2 Research leads us to believe in the former. There is substantial evidence to indicate that motion does indeed grab attention at a low level (Pylyshyn et al., 1994; Hillstrom and Yantis, 1994) and across the entire visual field (Sekuler et al., 1981). However, at a functional level, we are less concerned with whether detection was due to polling or low-level attentional grab, since we expect the user deployed the same strategies regardless of cue type. We believe that these operational differences between the various cue types are still striking enough to warrant considering moticons as efficient alerting mechanisms, although future research should indeed look at studies with sparser and more managed cue distribution to ascertain whether these results would hold up in vigilance applications where cues arrive unpredictably and the cost of polling is unacceptable. 7.1. Guidelines for motion-based techniques A review of the research and these results suggests the following preliminary recommendations and guidelines for adding motion-based signalling and awareness integration techniques to an interface. G1: Motion does not seem to interfere with existing colour and form coding, allowing extra information to be communicated through an object without changing its original codes which represent other variables. G2: Small periodic motions are generally better signals than colour or shape cues across the entire visual field. Such motions are more effectively detected and more accurately identified, especially in the periphery, and in displays where colour and shape are already used for other coding. Neither detection nor identification are affected by location, where colour and shape are poorly seen and identified outside the central area of view. G3: Motion amplitudes can be small—approximately 1° of visual angle is highly detectable even in pronounced peripheral vision conditions. While amplitude does not affect accuracy in detection, smaller amplitudes than 1° may have slower response times. G4: Even relatively slow frequencies are effective. The motions tested had periodic frequencies between 1 and 3 Hz. There is some research to suggest that frequency affects the perception of urgency ( Ware et al., 1992), but there was no perceptible effect in the range of frequencies we tested. G5: Motion continuity, or smoothness, appears to have little effect on detection and identification of signals. G6: Motion cue detection time, and to a lesser degree accuracy, is affected by the level of task engagement. This effect does not suggest motion signals are inappropriate in highly engaging tasks, as the detection results of all motion types tested were good even in the most attentionally demanding tasks. However, it does suggest that signals can be “tuned” according to task if immediate response time is required. G7: Motion can contribute strongly to distraction and irritation. In particular, travelling motions are significantly more distracting and irritating than anchored motions. Popping motions (where the object zooms in and out along the depth axis) were also considered distracting. G8: The slow linear oscillation is a good overall signal. It is accurately detected, elicits good response times, and is not considered intrusive or distracting. G9: Motion signals are easily computed. Refresh rates between 20 and 30 frames/s and durations of a few seconds are sufficient to elicit the perception of a single continuous motion event.