This section addresses the functionality that was considered core for our tabletop interface. An exhaustive set of implemented functionality and features is left for the following chapter. Here, the emphasis is on core facilities that we may build upon to address our design drivers. A summary is provided in Table3.1.
3.3.1 Select
Select is accomplished by physically touching an object on the display (G1a). Visual feedback is given that indicates which object is selected (G1e) and, when user identification is available, a colour corresponding to the user who selected it (G2e). Objects with hotspots – special areas that trigger actions other than move – also have their hotspots highlighted when selected. For example, selecting a photo will show the areas that can be used to rotate and resize it. All objects on the display are selectable (G1c), unless they are currently being manipulated by another user (G2e) – items cannot be stolen.
An explicit deselect is not normally required (G1b) – selecting a new object triggers an implicit deselect of selected objects. Touching the background (i.e. selecting nothing) also deselects. Maintaining only a single selection for each user also assists the goal of simplicity (G1b), but note that we do need a selection model that keeps track of each user’s activity (G2e). Grouping techniques (§3.3.5) can provide methods to manipulate multiple objects,
that multiple selection would provide.
Using hotspots also means that actions such as move, rotate, resize, etc. do not need to be preceded by an explicit select, and do not require modes (G1f). That is, the touch and drag primitives may be combined where user memory of the hotspot locations allows the select step to be skipped. Hotspots also allow the number of interface elements to be reduced – toolbars, icons and/or menus are not required (G1b) to change the mode. 3.3.2 Move
Moving objects is accomplished by a drag operation – the selected object, or objects, are moved the same displacement as the touch point (G1a), whilst giving continuous feedback
3.3. Key Design Elements CHAPTER 3. DESIGN OVERVIEW
(G1e). All selectable objects on the display can be moved (G1c).
It is sometimes not possible to move a particular object to the new location, for example, if it would move out of the display area. In these cases, each object is moved to the closest permissible location. This is pragmatic – we do not want users to lose objects by inadvertently moving them so that they fall off the table. However, the same logic can be used to enforce private areas, only accessible by particular users (G2e).
Flicking objects, an action motivated in some of the earliest touch screen research
[Minsky, 1984], involves releasing an object whilst it is being moved quickly. The object initially retains its velocity, but is subjected to a frictional force that slows it down, similar to flicking a physical coin across a regular table (G1a). For tabletops, flicking allows objects to be moved to distant parts of the display (G2a, G2d), pass objects to other users who may be out of reach (G2e), and potentially to transmit objects to other displays in the environment, such as a wall display (G2f).
3.3.3 Rotate & Resize
This is a composite function, using the drag operation and hotspots, that we created as a way to achieve simple interaction (G1b). To activate, a user touches an object in one of its corners and drags.
By combining rotate and resize in one action – rosize – it is possible to maintain the appearance of a single point on a photo being stuck to the participant’s touch point (G1a), as it is for move. The hotspot area is shown when an object is selected, and all four corners behave alike (G1c). This design symmetry is also important as it is not always possible to tell which way an object is oriented, and objects may be viewed and manipulated by users seated at all angles around the tabletop as the social nature (G2e) means that seating arrangements might not be fixed1.
During a rosize, the object is rotated and resized concurrently to keep the location on the object that was touched continually beneath the touch point (G1e). Note that it would not be possible to maintain this stuck effect if we performed only a rotation or resize without requiring the user to restrict their movement to a purely radial or circular motion. The stuck effect is also what we would expect when manipulating physical objects (G1a).
Rosize also fulfils some new requirements of the tabletop interface, namely the provision
of an easy resize to make elements larger (G2b), and a rotate so objects may be viewed at arbitrary angles (G2c). In addition, an easy way to resize also lets objects be made
smaller, e.g. to manage clutter (G2d).
3.3.4 Copy
Copy is also considered a core function to provide for the tabletop. The collaborative
nature of tabletop interaction (G2e) means that we can expect multiple users to need to speak about, look at and act on the same object. With a large table (G2a), this is infeasible or uncomfortable with a single object. Even a single user may want multiple copies of an object; for example, where they want to create several collages which each include the same image, or where they want to keep a copy of an original image before performing actions to transform it. In addition, file copy is a fundamental aspect of operating systems, and needed to support current user behaviours with digital cameras (i.e. to backup digital photographs).
The need for a copy is also indicated by studies where people were observed to steal objects from others [Morris et al., 2004b]. This can be avoided by making it easy for people
1In particular, for some types of social interaction (e.g. storytelling) users may have a preference for a
to copy an object so that any or all people at the table may have their own copy, oriented as they please and at the size they choose.
Clearly copying and resizing are not feasible with most physical objects but can be supported in tabletop interfaces.
3.3.5 Grouping
Grouping applies to creating areas of the tabletop to designate as personal space (G2e), creating objects to serve as temporary storage, or creating associations between items. A temporary storage helps people to manage clutter (G2d) and making the storage movable can assist in moving groups of objects around the table (G2a). Creating associations allows a hierarchy of object relationships to be formed, can assist in tagging, and may serve as a means to place new, dependent objects in a meaningful place, via attachment.
Other important forms of grouping can be based on the notions of personal spaces (objects or screen elements that can be made private), storage bins (movable and resizable containers, with behaviour like regular photos – G1c), and attachment. Personal spaces enforced by the interface can help avoid stealing behaviour [Morris et al., 2004b], and Scott’s [2005] work suggests the need to explore personal spaces as well as personal storage [Scott et al., 2005].
Attachment could be activated by flipping an object over like a postcard (G1a), with items able to be attached to the reverse side (and automatically laid out) simply by moving them over a flipped object and releasing. Browsing techniques also need to present an initial grouping for objects when they are newly loaded onto the interface, for example to group the contents of folders on disk.
3.3.6 Delete
Unlike a physical table, the virtual objects on a tabletop display are created easily. In order to minimise clutter (G2d), and to support the user in task management, a delete operation is important. In designing the delete operation, we wanted a reversible action that could be executed simply (G1b) and with continuous feedback (G1e), so that the action could be cancelled early if was started unintentionally – an image should not simply disappear, based on some action that may be triggered accidentally. A confirmation dialogue asking, “Do you really want to delete?” was dismissed as it could be irritating, it would detract from social aspects like storytelling (G2e), we would not know where on the table to show it (G2a) or which way to orient it (G2c), and the text could be hard to read (G2b).
We also did not wish the user to navigate a command hierarchy, toolbars or text as these create additional interface elements to learn and remember (G1b). In addition, menus and text typically require an orientation which might not be determinable (G2c). This is compounded by the typically large dot pitch of the tabletop display (G2a), which can make text hard to read and icons hard to distinguish. Another important concern when interacting with such objects on touch interfaces is the fat finger syndrome whereby small objects are hard to select (G2b), and the display is often occluded by the user’s hand whilst trying to find the item they wish to select – there is no mouse cursor (2f).
Our solution to this was the Black Hole and is somewhat analogous to the astronomical phenomenon (G1a, G1c). It is discussed in detail in Section 4.3.1.
3.3.7 Capture
The desire for a capture functionality initially arose from a cultural probe study [Risborg and Quigley, 2003] which highlighted the importance of scrapbooking and postcards in existing non-technological photo sharing habits. Capturing a highlighted area of the display allows a collage of images to be loaded as a new object, similar to a physical postcard G1a.