The compact control panel interface 

Contexts in source publication

Context 1

... measure how the system performs while being controlled by experimenter. Therefore how participant reacts to the system can affect experimenter control panel operations. There are two ways of reflecting experimenter operations, one is how ‘system mistakes’ are handled and the other is how real-time responses are presented. Fraser and Gilbert [2] suggested that the wizard should take account of making some mistakes to keep faithfulness to technologies, and they also suggested a 5 percent mistaken occasions. To fulfil this target the experimenter needs to make some mistakes on purpose. For instance, taking the speech ‘thunder’ as ‘honda’ [23] and then presenting a learning progress to convince participants that the system is automatically learning on its own. Some speech misunderstandings may also contribute to mistaken occasions since the experimenter cannot be able to comprehensively understand all speeches which are close to personal lives. By analysing participant reactions toward these mistakes, we can measure how flexible the experimenter-mediated system handles mistakes through control panels. Responding in real-time is the other important criteria which evaluates how fast the system can respond to participant. In this study these responses are all generated via control panels, thus measuring how fast the system interacts with participants can reflect the extent of operation effectiveness. The response durations can be gained from recorded videos according to time stamps. We designed the application system in the domain of domestic communication due to the home is an ideal site for emerging interactive technologies [24-26]. The three system applications were proposed based on daily routines in the home [27, 28], and these were integrated with mundane rhythms such like organising daily appointments and managing multimedia. The system was programmed with C++ (MFC) in Visual Studio, therefore it had normal window elements. The system first used intranet-based communication for remote manipulation and surveillance. After several trial tests the system integrated control panels and applications in one computer due to the instable network could not afford simultaneous manipulation and video surveillance. We proposed two versions of control panels. One was designed with united layouts which contain all modules in one panel (see Fig. 1), and the other design split the control panel into a group of sub-panels (see Fig. 2). This division was made according to system applications that each application had one separate control panel. Both control panels had the same application interfaces. The application interfaces were the same due to this makes participant easy to compare system performance changes (see Fig. 3 and Fig. 4). Domestic interactions require experiment spaces which can provide home-like environments. We set up a scenario in laboratory as a domestic communication scene. It consisted of a set of sofa and a coffee table as well as multiple experimental devices such like projectors, webcams and microphones. The site separated some invisible space for experimenter to facilitate the system. Separated by big screens the experimenter could not be seen by participants, while the experimenter could still observe participant via surveillance video. The experimenter manipulated the host computer and monitored application running. The experimenter was located with the computer while application interfaces (Fig.3 and Fig. 4) were separately distributed in front of participant. A volunteered participant was employed in our study. She had little knowledge of WoZ system but with strong interests in experiencing novel interaction styles. This is due to that experiments may be severely affected if participant becomes aware of experimenter existing. Furthermore the participant had good experiences of mundane affaires. With these conditions the participant was told that she was interacting with an intelligent computer system and her speeches could be recognised and learnt by the system. The participant needed to go through two experiments and she might be asked to compare the system performance changes in interviews. The experimenter, who designed whole system applications as well as control panels, was playing the role of system facilitator. One advantage of this is that the experimenter does not require extra training to manipulate control panels. Meanwhile the experimenter can handle unexpected errors carefully based on system familiarity. The most important reason is that the experimenter can have first-hand experiences of control panel. The invisible experimenter could not directly associate with participant. To address that a new role was introduced which was called ‘instructor’. Her main responsibility was to deal with the participant face-to-face as an experiment moderator. The instructor could deliver some indications about system applications. Each experiment adopted a different control panel as described in Fig. 1 and Fig. 2. The first experiment started from system introduction by the instructor. After that the instructor gave participant a sheet in which command examples and tasks were listed. The instructor then allowed some time for participant to learn system functionalities. This was to make sure that the task completion was based on skilled manipulations, which might low the risks of dealing unfamiliar functionalities and wasting unnecessary time. Once the learning was done, participant was allowed to start tasks. All tasks should be accomplished and these included a) basic calendar operations (viewing / adding / deleting appointments), b) communicating with the system through the dialogue and c) using media manager to play videos. The media manager works with a coloured cube. System responses were facilitated by the experimenter via control panels. When received incorrect speeches experimenter could use the dialogue to display alert messages and thus communicated with participant. While using the cube the system launched a video and played it on the coffee table. The experimenter also needed to sense the cube movements which were assigned with different operations. After all tasks the participant was invited to a semi- formal interview which encouraged the participant to express experiences and thoughts about the system. All comments made by participant were logged in videos. These videos were manually transcribed into scripts with time stamps, therefore operation durations could be calculated and analysed. The participant was told about the simulation system after last experiment’s interview had accomplished. Then participant was encouraged to express the thoughts about experimenter facilitation. And finally participant’s consents were required for further data analysis. IV. E XPERIMENT A NALYSIS The experimental materials were collected from two aspects: observations and video analysis. These materials provided quantitative data in terms of response durations and mistake numbers, and also provided qualitative data which reflects how facilitation interacted with participant. Videos were transcribed into texts by the experimenter who could recall the motivations of facilitation. These were complementary to understand participant reactions. Below is an example of scripts that demonstrates how the participant used calendar. [00:05] Subject: today [ Speaking without hesitation and starting to wait for system responses ] System: showing today’s events [00:08] Subject: create an event. [ Looking at the table and giving the speech when saw today’s events ] [00:10] System: popping up an input window for event contents [ a short pause due to the participant was thinking about an appointment to add ] [00:19] Subject: event for tomorrow and 2 o’clock [ speaking naturally, still looking at the table and waiting for next response ] [ A short pause due to the participant forgot to use the ‘confirm’ to finish the input ...

Context 2

... Gould’s (1983) listening typewriter had only one textbox for speech transcription, and Höysniemi’s (1989) gesture recognition system had one simple interface for direction controls via keyboard. To date control panels are designed with multiple components such as buttons, textboxes and other graphical elements. These components have provided basic accesses to fast responses, although the delay that caused by experimenter is still noticeable as described in [7]. The control panel designs aim to associate the multifaceted and situated relations between system and experimenter [21]. Multiple roles of experimenter exist in WoZ studies, such like controller, moderator and supervisor [22]. These dynamics need to be addressed in control panel designs due to different experimenter roles require different control panels. For example, a supervision control panel has more surveillance functions yet with less control elements and a control panel has more manipulative components. Therefore in our study the control panels are designed on controlling purposes with multiple manipulative elements. III. M ETHODOLOGY Our methodology was conceived as a practical, cyclical progress in which variables were controlled and compared throughout two experiments. Each experiment went through a set of tasks involving three system applications: a domestic calendar, a communication dialogue and a cube-based media manager. The domestic calendar was a speech-recognition application designed for family arrangements planning. The communication dialogue was designed for natural language dialogues and the media manager was a simple multimedia manager. A full experiment cycle contained three action stages planning, acting and reflecting. The planning stage related to setting up experiments, the acting stage to conducting experiments procedures and the reflecting stage to data analysis. All experiments were video recorded by a webcam which was set beside the experiment site. Data was manually transcribed into scripts for further analysis which, was the one of most significant part in our study. A semi-formal post-experiment interview was planned. Through interviews we aimed to extract user thoughts about system performances. In this WoZ study participant could not be told the truth until all experiments were completed. However such low-level ‘deception’ still needed to be explained to participant for data use consents. Criteria are important for control panel design assessments. Unlike normal usability evaluations, control panels in WoZ studies are invisible to participants. To evaluate these control panels we need to measure how the system performs while being controlled by experimenter. Therefore how participant reacts to the system can affect experimenter control panel operations. There are two ways of reflecting experimenter operations, one is how ‘system mistakes’ are handled and the other is how real-time responses are presented. Fraser and Gilbert [2] suggested that the wizard should take account of making some mistakes to keep faithfulness to technologies, and they also suggested a 5 percent mistaken occasions. To fulfil this target the experimenter needs to make some mistakes on purpose. For instance, taking the speech ‘thunder’ as ‘honda’ [23] and then presenting a learning progress to convince participants that the system is automatically learning on its own. Some speech misunderstandings may also contribute to mistaken occasions since the experimenter cannot be able to comprehensively understand all speeches which are close to personal lives. By analysing participant reactions toward these mistakes, we can measure how flexible the experimenter-mediated system handles mistakes through control panels. Responding in real-time is the other important criteria which evaluates how fast the system can respond to participant. In this study these responses are all generated via control panels, thus measuring how fast the system interacts with participants can reflect the extent of operation effectiveness. The response durations can be gained from recorded videos according to time stamps. We designed the application system in the domain of domestic communication due to the home is an ideal site for emerging interactive technologies [24-26]. The three system applications were proposed based on daily routines in the home [27, 28], and these were integrated with mundane rhythms such like organising daily appointments and managing multimedia. The system was programmed with C++ (MFC) in Visual Studio, therefore it had normal window elements. The system first used intranet-based communication for remote manipulation and surveillance. After several trial tests the system integrated control panels and applications in one computer due to the instable network could not afford simultaneous manipulation and video surveillance. We proposed two versions of control panels. One was designed with united layouts which contain all modules in one panel (see Fig. 1), and the other design split the control panel into a group of sub-panels (see Fig. 2). This division was made according to system applications that each application had one separate control panel. Both control panels had the same application interfaces. The application interfaces were the same due to this makes participant easy to compare system performance changes (see Fig. 3 and Fig. 4). Domestic interactions require experiment spaces which can provide home-like environments. We set up a scenario in laboratory as a domestic communication scene. It consisted of a set of sofa and a coffee table as well as multiple experimental devices such like projectors, webcams and microphones. The site separated some invisible space for experimenter to facilitate the system. Separated by big screens the experimenter could not be seen by participants, while the experimenter could still observe participant via surveillance video. The experimenter manipulated the host computer and monitored application running. The experimenter was located with the computer while application interfaces (Fig.3 and Fig. 4) were separately distributed in front of participant. A volunteered participant was employed in our study. She had little knowledge of WoZ system but with strong interests in experiencing novel interaction styles. This is due to that experiments may be severely affected if participant becomes aware of experimenter existing. Furthermore the participant had good experiences of mundane affaires. With these conditions the participant was told that she was interacting with an intelligent computer system and her speeches could be recognised and learnt by the system. The participant needed to go through two experiments and she might be asked to compare the system performance changes in interviews. The experimenter, who designed whole system applications as well as control panels, was playing the role of system facilitator. One advantage of this is that the experimenter does not require extra training to manipulate control panels. Meanwhile the experimenter can handle unexpected errors carefully based on system familiarity. The most important reason is that the experimenter can have first-hand experiences of control panel. The invisible experimenter could not directly associate with participant. To address that a new role was introduced which was called ‘instructor’. Her main responsibility was to deal with the participant face-to-face as an experiment moderator. The instructor could deliver some indications about system applications. Each experiment adopted a different control panel as described in Fig. 1 and Fig. 2. The first experiment started from system introduction by the instructor. After that the instructor gave participant a sheet in which command examples and tasks were listed. The instructor then allowed some time for participant to learn system functionalities. This was to make sure that the task completion was based on skilled manipulations, which might low the risks of dealing unfamiliar functionalities and wasting unnecessary time. Once the learning was done, participant was allowed to start tasks. All tasks should be accomplished and these included a) basic calendar operations (viewing / adding / deleting appointments), b) communicating with the system through the dialogue and c) using media manager to play videos. The media manager works with a coloured cube. System responses were facilitated by the experimenter via control panels. When received incorrect speeches experimenter could use the dialogue to display alert messages and thus communicated with participant. While using the cube the system launched a video and played it on the coffee table. The experimenter also needed to sense the cube movements which were assigned with different operations. After all tasks the participant was invited to a semi- formal interview which encouraged the participant to express experiences and thoughts about the system. All comments made by participant were logged in videos. These videos were manually transcribed into scripts with time stamps, therefore operation durations could be calculated and analysed. The participant was told about the simulation system after last experiment’s interview had accomplished. Then participant was encouraged to express the thoughts about experimenter facilitation. And finally participant’s consents were required for further data analysis. IV. E XPERIMENT A NALYSIS The experimental materials were collected from two aspects: observations and video analysis. These materials provided quantitative data in terms of response durations and mistake numbers, and also provided qualitative data which reflects how facilitation interacted with participant. Videos were transcribed into texts by the experimenter who could recall the motivations of facilitation. These were complementary to understand participant reactions. Below is an example of scripts that demonstrates how the participant used ...

Context 3

... proposed two versions of control panels. One was designed with united layouts which contain all modules in one panel (see Fig. 1), and the other design split the control panel into a group of sub-panels (see Fig. 2). This division was made according to system applications that each application had one separate control panel. Both control panels had the same application interfaces. The application interfaces were the same due to this makes participant easy to ...

Context 4

... directly associate with participant. To address that a new role was introduced which was called 'instructor'. Her main responsibility was to deal with the participant face-to-face as an experiment moderator. The instructor could deliver some indications about system applications. Each experiment adopted a different control panel as described in Fig. 1 and Fig. 2. The first experiment started from system introduction by the instructor. After that the instructor gave participant a sheet in which command examples and tasks were listed. The instructor then allowed some time for participant to learn system functionalities. This was to make sure that the task completion was based on skilled ...