by Juan C. Dürsteler
[message nº 128]
|Interaction is a key element in learning and acquiring information. It is intrinsically dependent on time and on control. In the previous article we focused on time. In this one we explore the techniques for interaction control within visualisations
What can be influenced in order to change the state of the visualisation we are examining? In Information Visualisation, interaction affects the phases where we transform data into information and this into [graphic] representations.
Data is transformed into information by incorporating metadata, data explaining the data, to it and grouping it into data tables that are coherent with its meaning. On the other hand, information becomes graphic representation by establishing a mapping between data and the elements of the visual representation and applying geometric transformation like zoom or translation to them.
For example, the selection of particular sets of rows or columns gives rise to different views of the database and, hence, to different representations. In the same way we can change the type of representation and/or zoom or rotate the graphic components. This means that the user can perform changes that allow him/her to modify the parameters governing these transformations.
|Controlling interaction: The user can interact with the parameters that govern the transformation of data into information and from this into visual representation. This way the user can potentially modify the structure and values of the data to be presented, the visual mapping employed to depict it and the view with which to see it. Graphic by the author. |
These parameters can have influence on, at least, three types of interaction
- Interaction with data transformations
- Querying the database. Depending on the subset we want to show, different queries can extract different subsets of data.
- Statistical transformations of and among the stored data. We could wish to visualize derived magnitudes, like differences, averages or other statistical magnitudes.
- Structural changes in the database. In the same way we derive new magnitudes from existing data we can change or combine the structure of the database in order to find new patterns.
- Interaction with the type of graphic representation, the correspondence between data and graphic structures, i.e. the visual mapping and/or the type of correspondence between what is shown and what is meant.
- Interaction with the geometric transformations that affect the rendering of the representation, like rotations, translations and projections. You can change the perspective, the angle of rotation, the pan or zoom…
According to these three types of interaction Card et al. in “Readings of Information Visualisation” distinguish several techniques for performing interaction.
- Interaction with data transformations.
- Dynamic queries. In this technique sliders and buttons in their diverse forms are used to select and set boundaries to the data to be visualised. The rest of the data remains hidden or is partially visualised (for example using transparency). We have an example of this in CityO’Scope. In it a system of parallel coordinates represents data of more than 50 cities distributed in 48 axes, one per variable, provided with sliders that allow the user to set upper and lower limits to each and every variable and select that way the cities whose variables are within the ranges so defined.
- Direct walk. This technique concatenates the different data of interest. A browser like Explorer or Navigator is an example of this type of interaction. Jumping from link to link we trigger new visualisations each one associated to the particular link.
- Details on Demand. It allows the user to expand the details of a small set of objects. The user can make the object (a case of the data table) to show more variables than the displayed because there is room left to do so.
- Attribute Walk. The user selects some data and then the system searches data with similar attributes, showing a visualisation of the same.
- Brushing. This is a powerful technique that presents the user different and simultaneous visualisations of the same data set seen from different perspective and/or showing different attributes of the data set (each one is called a view of the data set). When you select a particular case of the data table in one of the visualisations its corresponding item is highlighted in the other visualisations. A good example of this technique is again CityO’Scope, that provides three different views of the citiy dataset. In one you have a mapamundi with a dot for each city in its geographic location. Another is a similarity plot where the dots that depict cities are separated by a distance proportional to the statistical similarity between them. The third is the parallel coordinates plot showing all the variables for all the cities. Selecting a particular city in any of them highlights the corresponding symbol of the city (dot or line) in the remaining two.
- Direct Manipulation. Dialogue elements embedded in the visualisation allow the user to directly manipulate the data transformations.
- Interaction with the correspondence between data and visual form
- Dataflow. This type of interaction has been used in visual programming. It uses an explicit representation such as node-link diagrams to show the flow of data instead of showing the flow of logic (the the sequence of programming sentences) so that the user can change it by manipulating the visual mapping. AVS (Advanced Visual System and other tools like LabView use these dataflow diagrams to control the flux of data within programs in visual form.
- Pivot tables. This is a technique found in spreadsheets that lets the user easily manipulate the mapping of data to rows and columns. (See an example of Pivot tables in Excel)
- Interaction with geometric transformations
- Direct selection. Deals with the schemes that allow selecting and highlighting individual graphic objects or groups of them
- Camera movement. Accounts for the change in position of the center of projection. This changes the perspective, the point of view of the observer avoiding in certain applications the occlusion that occurs between objects, specially in 3D images. This technique also allows us to see parts of the visualisation that can’t be seen from another given perspective.
- Magic lens. This technique selects objects according to the X, Y of their marks and then combines with further selection techniques like dynamic queries. They can combine also with other techniques transforming the view or the data. See for example the Magic Lens project
- Overview + Detail. It consists of two or more linked visualisations. One contains a detailed view of the visualisation whereas the other(s) show a general view that spans all the objects contained in the visualisation. The detailed view is marked as a region in the general view that can be moved to show other parts and objects in detail. Examples of this technique are the “radar” of videogames that shows you in a small window which part the general layout of the game is showing on the main screen. In a castle, for example the radar depict the whole plant of the castle with a rectangle showing which room the main, detailed, screen is displaying .
- Zoom. Reduces the number of visible objects but increasing the level of detail, possibly by increasing the number of variables of each object that are shown. Conversely it can be used to increase the number of objects decreasing the level of detail. Zooming can be
- Physical, just by multiplying by a factor every graphic coordinate scale, thus increasing or decreasing the size of the graphic objects and clipping the resulting plot to the limits of the window.
- Conceptual, in which the increment is not [only] the physical size of the objects but the number of their variables shown.
By using judiciously time and control is how the majority of the best interaction systems have been built. This is an important aspect of any system since, in most cases, interaction is the key to productivity.
Links of this issue: