Objetivos secundarios
Capítulo 3. Material y Métodos
3.5.2 Objetivos clínicos
Thought there a variety of different models that could have been used to conduct this analysis, the one conducted in this chapter supports our claims that gestures are less disruptive in multitasking situations. Though the power ratings for our results were not high, the trends and observations do support this result. However, we note that conducting laboratory studies on users does not represent the ideal scenario for deter- mining real usability results for any interaction technique. A more telling approach would be to provide the opportunity to use gestures in a real-world scenario, for a va- riety of users, over an extended period of time. Results from these empirically based lab studies do however provide evidence to suggest that gestures are a less distracting technique, however real-world, ethnographic or qualitative approaches would provide a much more complete picture that would enable us to make more decisive conclusions about the effectiveness of different interaction techniques. This is an approach that will be explored in future studies.
3.4
Summary
While Wexelblat and others have criticised the use of semaphoric gestures as an inter- action technique, results of an experiment suggest that there are significant benefits to semaphroic gestures for secondary task interaction. We also argue that there are benefits in terms the consistent number of steps required to gesture at a camera, whereas the use of direct input device controllers can introduce increased delays in completing the sec- ondary task and potential degrade recovery of primary task focus. Task recovery time is of critical interest in notification systems research conducted byCzerwinski et al.(2004) and our results suggest that significant improvements are possible when using gesture over function keys for reducing task recovery time. Our results also suggest that inter- action mode is a significant factor for assessing interaction performance with ambient or secondary task systems. In the next chapter, we further explore the domain of gestures in an empirical study designed to provide us with a more detailed understanding of the different characteristics of multitasking while investigating user tolerance for errors in gesture recognition.
Performance Issues
”Each problem that I solved became a rule, which served afterwards to solve other problems.”
Rene Descartes
4.1
Introduction
In our previous chapter, results from an experiment suggested that there was indeed a functional utility for semaphoric gestures. In that experiment, error rate was held constant, where we did not simulate any recognition errors. We next embarked on a study to determine the level of recognition error that a user can tolerate. Since current state of the art vision technology cannot yet achieve 100% accurate recognition, we set out to determine what level of recognition error users would tolerate. To design an appropriate scenario for this experiment, we conducted a participant observation study to explore tasks, and interaction scenarios that could inform our scenario design. That study is presented in detail in Appendix C, and the applicable results which informed the scenario we used in the current study are discussed in Section 4.3 of this chapter. As in the previous experiment, presented in Chapter 3, the study described in the current chapter considers gestures from the interaction perspective, and supports the human perspective of interaction research. However, in this work, we began to apply our knowledge gained from previous experiments to guide this research.
For the error tolerance experiment discussed in this chapter, we extend the work pre- sented byBeaudouin-Lafon (2004) and propose an interaction model to create a frame- work for guiding our evaluation of gestures. The model represents lessons learnt and experience gained from our previous work with gestures in Chapter 3. We could now understand gestures in terms of their interaction context, system performance measures and the goals of the users. We apply our experience and define interaction context in terms of the physical layout of the interaction space, system performance in terms of the accuracy rate of the gesture recognition system and user goals in terms of task char- acteristics. With this structure in place, we began our investigation to determine what level of accuracy is required for a gesture detection system to be both tolerated and experienced as useful, and in what contexts might gestures be more appropriate over alternative, physical input mechanisms?
We used the Wizard of Oz (WoZ) methodology for this experiment, where we explore user tolerance for errors in gesture recognition systems, described by the interaction model. We also demonstrated how researchers and designers can apply these results to assist in determining if gestures will enhance an interaction scenario. We continue our investigate of semaphoric gestures in this work, where hands are used to sign or signal commands to the computer and discuss how our proposed interaction model can be extended to inform future evaluations. In the next section, we present related work that explores user tolerance for computer interactions, followed by the details, results, and conclusions of our experiment. A short version of this study appeared in the Conference on Advanced Visual Interfaces, 2006 Karam & schraefel(2006).
4.2
Interaction Model
We discuss the three elements proposed for our interaction model for investigating ges- tures, and their role in influencing user tolerance for recognition errors. We refer to
Beaudouin-Lafon (2004) definition of an interaction model for this research, as having the following function:
The purpose of an interaction model is to provide a framework for guiding designers, developers and even users (in the context of participatory design) to create interactive systems. An interaction model is thus more operational than an interaction paradigm and can be used directly by designers.
We present three main elements of our proposed interaction model, and their role in providing a framework from which we can design our interactions to determine user tolerance for gesture system recognition errors.
for controlling the secondary task. We measure user’s tolerance as the number of times they choose to use the keyboard instead of the gestures. In the desktop scenario, the keyboard is located in front of the participant and the monitor used in the experiment (see Figure4.7). In the ubiquitous scenario, we physically extend the desktop metaphor so that the keyboard is located away from the participant, thus simulating the distance style interaction of a ubiquitous computing scenario.
System performance. To investigate system performance and the effects on usabil- ity, we considered the sensitivity level of the recognition system —highly sensitive leads to false positive errors, while low sensitivity leads to false negative errors —as well as the time taken to process and respond to a gesture command. In this experiment, we concentrate on error rates as the key variable for measuring user tolerance.
Users’ Goals For this experiment, we investigated a multitasking situation, where the participants work on non-computer primary tasks, while controlling computer-based secondary tasks. While there are many different task characteristics that can define or describe user goals, we refer to those determined in our ethnographic study (see section
4.3 later in this chapter, including the cognitive and physical nature or complexity of the tasks, the relationship between the primary and secondary tasks, and the criticality of the secondary task. To create criticality, we imposed a timing constraint for the critical tasks, so that the users’ goal was to complete the primary tasks as quickly as possible while controlling the display (secondary tasks). In the non-critical condition, users were to complete the primary tasks without timing constraints while controlling the visual display. We also considered related tasks, where the primary task was dependent the secondary task, and a single-decision task, consisting of a single unit or gesture interaction to complete.