There is a substantial theoretical and empirical interaction design literature that can be used to both explain the role and utility of Attribute Gates, and configure their spatial layout and dimensions. Attribute Gates were designed to both optimise the targeting and steering effort in combination with the operationalization of actions proposed by activity theory
3.3.1 Activity theory and chunking
Activity theory (Kuutti, 1995; Bødker, 1989; Kaptelinin, 1995; Fjeld et al., 2002; Halverson, 2002; Nardi, 1995) is a useful framework for analysing physical and co-located collaborative interaction settings. The definitions of activity, action, and operation, the task levels they represent, and the transition between them, forms the basis of the following analysis (Figure 3.3). Activityis defined as the minimal meaningful context that is directed to an object in order to transform it into an outcome. It is the basic unit of analysis, driven by a motive, and is carried out by a series of actions. An action is a conscious act with a direct defined goal that usually consists of a number of operations. An operationis the subconscious act that might once have been an action done consciously, but with practice and repetition became a routine act (was turned into an operation). Thus, when a person with little experience of using a QWERTY keyboard wants to type text, locating each letter is an action in itself. However, when this user gains experience, typing becomes a series of operations performed subconsciously.
Individuals new to a certain activity need to think about every step of the process. Such a process is undertaken through a series of well thought out actions, with clear specific intentions behind each. In such cases, no operations are involved and focus shifts from the high-level task to trivial low-level tasks. After practice and repetition, appropriate actions are performed subcon- sciously, and this transformation to operations (in users’ heads) allows them to focus on higher level tasks. As more actions are turned into operations, it becomes easier to stop worrying about the details and concentrate on the desired outcome as the individual will subconsciously trigger the appropriate sequence of operations depending on the conditions at hand. Kaptelinin (1995) observed that by looking at whether a subject’s behaviour, in a specific situation, is oriented to- ward a motive, a goal, or is in response to a specific condition, one can better understand and predict the subject’s behaviour. Moreover, activity theory brings into consideration such issues as whether the user’s focus is on the tool or the goal (the tools must keep its user’s focus on the goal and not draw attention to itself). So if a user wants to move an object between two territories and change its attributes, the tool to change these attributes should not capture the user’s attention (as is the case with having to select six command/subcommands in contextual menus).
In simple terms, activity theory tells us that good user interface components should move actions into operations, or at least allow for this move as experience increases. Buxton’s (1995) work on chunking is strikingly similar to activity theory. In discussing the differences between
Figure 3.3: Activity levels: a good user interface element helps shift more actions into operations, which helps the user to focus on the higher level activity.
the levels of detail that novices and experts attend to, he describes how for novices, finding a character on the keyboard or remembering the name of a command, requires valuable cognitive resources (which can be performed by experts automatically). Where activity theory describes the progression from novice to expert in terms of carrying out more actions as operations, Buxton’s notion of chunking refers to the amount of a problem that can be performed automatically (i.e. as an operation) and thus he proposes gluing a number of subtasks into one task (chunk). According to Buxton, the three subtasks required to select a command from a contextual menu (right-click to show the contextual menu, moving the pointer to the required command, then clicking on the command) can be glued together if a simple modification is made where the user presses and holds the right button, moves to the desired command, then releases the mouse button. In this case the muscular tension of pressing the mouse button is theglue. Such gluing brings subtasks into one chunk that corresponds more closely to the user’s model of the task.
3.3.2 Crossing-based interfaces
Accot and Zhai (2002) proposed crossing as an alternative to point-and-click interfaces, espe- cially for pen-based interaction. Crossing allows the initiation of a command by simply crossing a specific target (i.e. an object at a spatial location) without the need to point or click on interface components. Accot and Zhai found that target crossing could be more efficient, or at least as effi- cient, as pointing. They provided a number of guidelines for designing crossing-based interfaces and recommended that whenever possible, the target to be crossed should be orthogonal to the direction of movement. Selecting a command using the crossing technique, unlike point and click,
is one indivisible task. By appropriately positioning commands, it is possible to issue (cross) multiple commands in one operation, thus satisfying the design recommendations suggested by activity theory (e.g. Kuutti (1995))
Although crossing-based interfaces were originally proposed for pen-input, they are similarly appropriate for finger-based interaction. This is particularly true for finger-based interaction with large surfaces, where the size of the interactive surface can comfortably tolerate the size of finger- tips (as compared to pen-input).
3.3.3 Targeting and steering
Attribute Gates use the crossing principle. Setting attributes involves steering between elements and crossing others. The layout of these elements may be optimised (to increase ease of use and efficiency) by the application of targeting and steering laws derived from Fitts’ law (Fitts, 1954). A number of variations of Fitts’ original law have been introduced to address different aspects of user interface design. Accot and Zhai (1997) extended Fitts’ law through the introduction of an equation to calculate the time required for steering inside a path and validated the accuracy of the equation experimentally. Assuming a path of fixed widthW and of lengthD, the time (T) required to move inside that path is
T =a+b D W (3.1)
In equation (3.1),aandbare empirically determined constants that are characteristic of a user, and the ratio D/W is the index of difficulty. Steering time has a linear relation with the ratio D/W, unlike Fitts’ original formulation, in which the targeting time has a logarithmic relation to D/W. Accot and Zhai have demonstrated how this law can be used to estimate the time required to select and navigate through commands in a multi-level menu structure. Although an earlier investigation of steering tasks had been conducted by Drury (1971), it was for the study of vehicle guidance tasks in linear and circular paths, and consequently had to include such extra parameters as the risk factor. The function proposed by Accot and Zhai is simpler and more relevant to the task at hand.