Historical Considerations

4. Interfacing Humans with Computers

2. Move L1 between leafb andc.

3. Align L1 with index finger such that leaf (b) now lies aligned to leafa.

4. Move L2 such that leafaand b lie together and are hold between L2 and L3.

The complete gesture takes about 2 seconds and is often repeated to scan through the pile, sometimes in the following variants.

At 63.5s: Flip gesture with leaf. Goldsworthy picks a leaf and stores it between L2 and L3.

Variants

This alters the gesture such that the previously performed action of L2 is now executed by the combination of L2 and L3, whereas the L4 is now responsible for the movement previously performed by L3.

At 65.0s: Flip gesture with changed hand roles. Goldsworthy alters his flip gesture such that the right hand (R) now has the part of the left (L) and vice versa.

At 77.5s: Goldsworthy puts R2 between the leafs and flips it repetitively from right to left,

Index finger flip

each time flipping one or more leafs from one side to the other. Each iteration takes about half a second. This gesture really looks like it is easy to perform. It is supported by the relatively big gap between the single leafs that is caused by their waviness.

At 73.5s: Goldsworthy takes up a leaf with his right hand (R) while slightly rubbing it

Selection and texture

test between R1 and R2. He may feels its structure/texture and then decides to keep it.

At 79.5s: Take and select the leaf next to L2. It is not visible (also not to Goldsworthy).

He therefore uses only his tactile sense for the selection process.

Based on this analysis, I argue that Goldsworthy used his experience on the specific structure

Remarks

of leafs, not only for his sculptures, but also for their formation and building process.

Action and inter-action in reality, possibly incorporating other people or objects is often a

Observations

complex attempt. Goldsworthy’s use of his fingers to sort, identify and select leafs gives us a hint of this circumstances. The complexity of his movements and their variety, though, are fundamentally different from the common manipulations we physically apply to interfaces designed for data processing. In the short example, already four different manipulation gestures can be identified. Sorting the leafs involves many subconsciously performed tasks and analyses of peripheral information. Not only their size or colour are of interest, but also their texture, their material, quality or stiffness. It is difficult to cover this highly direct coupling of sensing and understanding in one word, the German wordbegreifen as a polyseme for to touch and to understand may fit best.

In difference to the observed behaviours and actions of Goldsworthy, today’s typical environments for data processing and exploration use only a rather limited part of our interaction and manipulation skills. In difference to Goldsworthy’s leafs, they often feature a symbolic interface to mediate the users intends and the data representation, which, in addition, is uni-modal most of the time. It is an aim of this thesis to apply techniques to interface design learned from the leafs example.

4.2. Historical Considerations

for centuries. The mechanisation of these working fields then required systems to trigger and control the evolving semiautomatic tasks. Trucks and motorised diggers, but also smaller tools like electric toothbrushes have controller boards and mechanical, pneumatic or electric switches, levers, or hand wheels to make automated action possible and controllable.

The increasing tendency to miniaturisation transformed previously mechanical systems into electric and finally electronic devices. The user interface, however, remained roughly the same, such that today, we operate complex machines with interfacing technology having substantial origins in a long history. All interfacing sensors capture their degrees of freedom, and they feed the resulting information as control parameters into the digital process under control. Physical appearance thus turned into virtual (mostly graphical) symbols. The Graphical User Interface was born.

But there were also systems invented for mechanical data manipulation, long before Analogue Computing

electronically supported computation was invented. Due to their mechanical construction, their form directly corresponded to attributes and characteristics of the incorporated data.

This makes them interesting concerning the design of new Tangible Interfaces for data exploration. In the following, I discuss two such machines; the slide rule and the planimeter.

Before delving into these two examples, a short overview on analogue computing in general is given.²

t s(t)

Analog Representation Digital Sample and Hold Real Signal

Figure 4.4.: Difference of continuous and discrete variables as they appear in analogue and digital systems.

Analogue computing has two distinguish-ing characteristics: parallelism and con-tinuity. Parallelism means that opera-tions can be performed in a truly parallel manner, i.e. many calculations are com-puted at the same time. Therefore, even sequenced modules can calculate their result in true realtime, making it easy to implement features like feedback control of sensor-actor systems. Continuity, in difference, is the use of continuous vari-ables/parameters: Analogue Computing devices change their state not in discrete steps, but in a smooth and continuous manner. In contrast, a digital computer performs operations sequentially and operates on discrete numbers represented in floating point or integer values.

Figure 4.4 exemplifies the difference between analogue and digital systems by means of typical signal representations. Although both types induce representation-specific artefacts, their nature are inherently different.

4.2.1. Slide Rule

The slide rule, invented by William Oughtred and others in the 1600s, is a mechanical analogue computing device that makes use of logarithmic scales to support numerical multiplication (see Figure4.5). It is based on the work on logarithms by Napier [Nap19].

It is elongated and comprises of two main parts: an angular rod and a rail that guides Hardware and functionality

its movement to be in parallel to its longest axis. On both elements, logarithmic scales are printed. Many slide rules additionally feature a glass window with either one or three

2Much of the information on analogue computing originates from the excellent web-siteThe Analogue Computing Museum[Cow00].

4. Interfacing Humans with Computers

hairlines printed perpendicular to the rod’s movement. The window can be moved parallel to the rods movement on the rail. By adjusting the scales to each other, multiplications can be performed. (a) The first multiplication factor (on the rail) has to be lined up with the number 1 on the other scale. Figure 4.5 shows such a configuration for π. (b) The multiplication’s result then can be read off the other scale: it is the number facing the second factor. Note that the slide rule makes it easy to compute all products for one given factor. This is achieved by glancing at the varying factors for each calculation, remaining the other factor fixed according to (a). The mechanical computing process is truly parallel, only the reading is done sequential.

Figure 4.5.: A slide rule. In its current configu-ration it can be used to read off all results forf(x) =πx. The hairline on its sliding window indicates that it is used forx= 1.16.

Calculations using the slide rule are of limited precision due to their analogue inputs and outputs and possible mechan-ical imprecision. Conversely, because of the discrete numerical input and floating point electronic operations, even modest modern calculators have output resolu-tions of at least six significant figures.

However, a slide rule tends to moderate the fallacy of false precision and signifi-cance. The typical precision available to a user of a slide rule is about three places of accuracy. This is in good correspon-dence with most data available for input to engineering formulas. When a modern pocket calculator is used, the precision may be displayed to seven or more decimal places, while in reality the results can never be of greater accuracy than the input data available.

A slide rule features a characteristic tangible experience which heavily depends on the used

Tangible experience

material. Slide rules are build either from wood, metal or plastics. The mechanics – if such a simple mechanism can be named as such – of a well-built slide rule does make the user feel a smooth friction, allowing him to easily adjust the relative positions of the rod and the window on the rail. This is an important feature, which, although not affecting the general functionality, adds a substantial value to its operation because it supports an exact positioning of the parts to each other. Furthermore, slide rules do not depend on electricity and are, due to their mechanical nature, easy to replicate. From a given example of a slide rule, more can be constructed by a competent craftsman from rudimentary materials using non-industrial processes.

4.2.2. Planimeter

A planimeter is an instrument that uses geometrical features to compute the area of graphically represented planar regions by tracing their boundaries. A planimeter has two

Hardware and

functionality arms with a freely moving elbow where one arm is a fixed to an anchor point. A needle traces the boundary of the region to be measured, while moving a wheel in the elbow, whose orientation is perpendicular to the elbow-to-needle arm. The net distance rolled by the

4.3. Research in Human Computer Interaction and Interaction Design

wheel is exactly the same as the contoured area.³ Planimeters exist in various forms, and can be classified into two main groups, (a) purely mechanical and (b) digitally enhanced ones. Their basic computation system is based on Green’s Theorem, which proofs that the tangential line integral of a vector field around a curve equals the double integral of the curl of that vector field. Thus the distance travelled by the rolling wheel equals the double integral over the region of the curl of the relevant vector field.

Figure 4.6.: A mechanical planimeter by the Ge-brüder HAFF GmbH.

See Section1.1 for cinformation.

Digital planimeters use the same tech-nique to measure areas of arbitrary shape, but can also be used to per-form other measurements, e.g. lengths, and volumes using contour lines. Their mechanism and mechanics, however, are exactly the same as used in analogue planimeters: only post-processing, inter-pretation and display of the measured values are done digitally by an integrated circuit board and are displayed by a dig-ital display.⁴

Both planimeter and slide rule are good Conclusion

examples on how physical relations and features of our environment can be utilised for manipulation and computa-tion of abstract data. Thus, one goal of this thesis is the promotion of these techniques to be integrated into Tangible Interfaces, helping to blur the gap between abstract data manipulation and reality with its rich feature-set.

4.3. Research in Human Computer Interaction and