|
Abstract | This article describes an evolution of the presentation of quantitative data. Technology driven presentations (paper, pencil, ruler, typewriter) are replaced by graphics meeting perceptual and cognitive psychological requirements. Discussed are single variable presentations (alphanumerical, analogue scale), multi variable presentations (tabular, Isotype, x-y graphs, Gapminder) and interacting multidimensional presentations.
To top.
| |
This evolution is general and applied to banking, traveling, train driving and population statistics. It suggests next-next generation presentations. The future of information presentation is not ahead but next to you. The future of information presentation is not in a black box but in the psychology of the user.
|
|
1. From alphanumerical to… |
Human visual performance with alphanumerical presentations is low. Small visual differences like between the characters 1 and 7 are difficult to perceive. In nature such small differences are not relevant for survival. Alphanumerical forms provide little information on the content. Visually ‘2345’ has more types of figures and lines than 10000, but its value is far less. Length of numbers and quantity presented might be compatible but can be obscured using leading zero’s *. (See Figure 6C).
| |
Some domains are in the alphanumerical phase and present variables not meeting the perceptual requirements. Not meeting psychological requirements increases reading time, search time and errors (Treurniet, 1980, Creed et al, 1987, Campbell et al., 1981; Beldie et al., 1983). Typical examples are the information banks present their customers. Public transport also is in the alphanumerical phase but the characters used to present information meet perceptual requirements much more. |
| |
* Leading zeros are zeros some numbers start with. James Bond (secret agent 007) has two leading zeros in his code name.
| | |
To top. |
|
2 …analogue |
The visual focus area on the retina has an arc of approximately 5 degrees. The circular form is compatible with circular scales like clocks. The hand on a clock is a straight line of which small changes in angle are easier to perceive than the details of an alphanumerical character. The traditional ergonomic rule is that digits are to be used when accuracy of perception is needed and an analogue presentation when speed of perception is needed. This rule of the thumb does not account for the use of scales having more than one hand and modern scale technologies that can combine the two (See Figure 3B).
To top.
| |
In addition, scales can present more than the value of a variable. They can show more parameters of the value such as the speed of change, the distance of the current value to the minimum and to the maximum value. All these parameters can be perceived at one glance. Visually and in practice, still more parameters of the value can be added such as a target, point zero, interpretation, reliability, units of the user, and trend. An analogue presentation can be more compatible with the content to be presented and that is what we need (Spivey, 2007) (See Figure 1).
|
|
2.1 Present user’s target |
Humans can be seen as black boxes reacting in some incomprehensible way to stimuli. Humans also can be seen as intrinsically goal-directed systems (Duncan, 1990). Making the goal parameter explicit is the first step in understanding behavior, especially today when the number of goals increases (economics, safety, comfort and all kinds of friendlynesses). Educational psychology developed a target-based methodology, called formative evaluation (Bloom et al. 1971). Domains dealing with quick responding, dangerous processes obeying the laws of Mother Nature, taught that planes crash and nuclear power plants explode when no target parameter is presented to the operator (See Figure 2, 3 and 5A). In non-physical and non-technical domains, targets might be less explicit (See Figure 4) and hidden for marketing or political reasons. However, when no targets are set, human cognition cannot control a process and solve a problem. Targets are landmarks for cognitive activities.
| |
In 1992 development started of the train driver interface for the European Train Control System (ETCS-dmi) *
The need to test the benefits of the presentation of a target speed (maximum speed permitted) was considered evident by the European train control experts involved and not tested. Recently this was confirmed by Fénix (2005) who presented schedule and energy targets to drivers and observed a 7% reduction of driving time and a 14% reduction of energy.
A bank balance without a target means the customer cannot answer basic questions such as: “Do I have to increase income/reduce costs?” A percentage showing the change of your balance would be more helpful than the balance of your account itself. A next step might be to present on entering your pin code for an expensive purchase: “This purchase will delay your target. Your 14 feet yacht will be launched four months later.”
The OECD** Quarterly growth table in Figure 4 is a typical descriptive target-less presentation. If no target is available the first quarter of the year or some mean could be used as a target. See Figure 10C. Content related targets in this domain might be Millennium Development Goals or the number of votes associated with a value.
| | |
* European Train Control System driver machine interface (ETCS-dmi), a European safety system to prevent high speed trains driving too fast. Forty European train safety experts and 101 European train drivers participated in this project. This mmi is now being implemented. www.humanefficiency.nl/etcs/etcs_dmi.shtml. |
| |
** OECD (Organisation for economic co-operation and development. OECD brings together the governments of countries committed to democracy and the market economy from around the world to support economic growth. www.oecd.org
| | |
To top.
|
|
2.2 Present user’s point zero |
Technical scales have a minimum, a maximum, and a current value somewhere in between. This three point scale model is not compatible with the visual system. When the left eye focuses left on zero, the right eye should focus right on maximum. That is impossible. In addition, there is no eye left for the information in between. Of course, we can switch to peripheral perception but in that case, we will miss details. The three points have to be inspected individually and the information has to be stored and interpreted in short term memory.
To top.
| |
In addition the cognitive focus of the user is on the difference between target value and current value (See Figure 5A). This difference should be the focus of the scale as well. The rest of the scale could be deleted.
|
|
Brase and Stelzer (2007) found that relative information is perceived as clearer than absolute. Some public transport companies apply this knowledge and prefer presenting user’s point zero (See Figure 6B and Verhoef 2008b), others do not (See Figure 6A). Keuning and Roding, (2008) compared these two conditions.
a) In the ‘no user’s point zero’ condition ‘time of departure’ is shown. The passenger has to search, read and remember current time and time of departure. Then calculate the time left and decide what to do: get coffee, walk, or run to the train.
b) In the second condition the system counts down and the ‘time left’ is shown. There is less searching, reading and calculating for the passenger.
To top.
| |
The investigators conclude that ‘time of departure’ is better than ‘time left’. Only empirical results and no perceptual and cognitive analysis are presented. A cognitive analysis (Verhoef, 2008b) suggests presenting ‘time left’ (point zero) to be better. The explanation for this interesting discrepancy is in the methodology of the experiment of Keuning and Roding (2008). The experiment does not meet basic requirements, each of which individually has to be met for a valid conclusion (Verhoef, 2008c). There were several differences in the treatments at the same time in the two conditions. These kinds of discrepancy in conclusions are typical in a field where design, marketing, cognitive psychology and research meet. Experiments that do not meet cognitive psychological methodological requirements undermine the value of empirical validation in the debate.
|
|
Conversation provides insight in this debate on the point zero parameter of a variable.
When we are free, we use a relative indication using our position as point zero. We usually do not say June 20 2009 but yesterday, today or tomorrow.
Computers and bureaucrats interpret ‘last year’ as 2008 because on paper it is not practical to print timetables and priceslists every day again. Whereas from a cognitive point of view it is the previous 365 days. That is what dynamic presentations might show.
The difference between ‘31-12-2009, 23:59’ and ‘1-1-2010, 0:01’ is large in a calendar clock system but these two seconds are irrelevant in most processes on earth. Categorizing data in months gives a delay up to 30 days and the year or month mean obscures the moment and the cause of changes.
When the information is presented on screens then computers and bureaucrats should not capture users in their calender-year-technical model. Point zero is ‘now’ and ‘here’ and that is what they should present.
To top.
| |
The unit in the OECD population statistics in figure 4 is a quarter. The units of a content based time scale might be a change of a law, presidency or a national disaster. Categorizing the time variable in quarters makes the task for the OECD designers and data collectors easy and the interpretation task for the users more difficult. A technical problem, of course, is that the resolution of the database and the underlying system is year, quarter or month based. Once this should be at day-level. An in between option is to set periods selected by users.
|
|
2.3 Present interpretation |
When there is a target, then there is an interpretation parameter: on target or off target. This parameter saves the user searching for actual value, target value, norms, performing calculations and interpretations. The interpretation depends on the task performed. The task might be performing a fixed (algorithmic) or undefined (heuristic) sequence of steps. There are visual and cognitive search tasks. For a monitoring task, the interpretation parameter has four levels:
a) No action needed
b) Action needed shortly.
c) Action needed now!
d) Action needed is not performed, game over, disaster occurred, target will never be reached.
These levels are general and can be applied to cognitive tasks like teaching, train driving, traveling and piloting to the planet Mars. The levels are elaborated in Verhoef (2007) and briefly shown in figure 5B.
To top.
| |
In population statistics setting norms for interpretation of data is not common practice because of the objective observing Popperian tradition (see section 2.1). This leads to tables that are a flat landscape without landmarks to navigate (See Figure 4B and 15). This impairs the exploration of data and the fine tuning of laws and rules. For instance, there is an unnecessary load to find out which countries have a too high birth/death ratio and need health support. Using these kinds of norms, the interface should interpret the value of the variable, e.g. coloring the values. The result might be a table in which the eye only can find the cells of the table that show where the problems are (See Figure 10).
|
|
2.4 Present trend | Anticipation is essential for survival in nature, so that is what human cognition does. A human action is performed to have an effect in future, to arrive at a target. A trend and history shows what might happen in the future. Today’s technology can provide the trend parameter of a variable.
To top.
| |
In tables and graphs, one axis can be used for time and trend. Unfortunately, that axis cannot be used to present an other variable. Gapminder solves this problem using the play button to visualize changes over time (See Figure 7). In quick responding processes the speed and the direction of the change show a short term trend parameter. A car driver can see that his speed is increasing or decreasing. The planning area of ETCS provides a long-term speed profile ahead (See Figure 8A). More informative than your bank balance might be the trend of your balance. Balance decreasing means: cut expenses, balance up means: buy more.
|
|
2.5 Present reliability |
The trend parameter might be more speculative than the current value of a variable and increase the need for the presentation of the reliability parameter. Recently the importance of controlling and presenting reliability of loans, savings and mortgages became clear rather dramatically. Also in public transport information it is common practice to hide reliability of a service.
There are several causes of unreliability.
Technical characteristics such as rounding values and discontinuous technology cause unreliability. When the resolution of a time of departure is one minute, the reading error can be 59 seconds when the information is perceived one second before the change of the minute. When the train safety system of a train is discontinuous having blocks of 1500 meter the inaccuracy of a train driving at 60 km/h can be 1.5 minutes. For a train traffic controller 2 x 1,5 minutes inaccuracy of two trains is a lot when he has to decide which of those two train will pass a crossing first.
External disturbances such as traffic jams or technical failures increase unreliability.
To top.
| |
The stronger the demand for 100% reliability of a public transport service, the larger the margins the planner will use. When maximum train speed would be the basis for scheduled speed, travel time is lowest but there always will be a delay. Reliability should not be avoided and hidden but should be used and presented (See Figure 9, 10A and 14).
Population statistics can be unreliable because national procedures differing from the international standard. The OECD statistics have an unreliability declaration. This, however, is more a legal statement (“Don’t blame us”) than an aid to draw the right conclusion. Figure 10A shows a population statistics table enabling the eye only to deal with unreliable data.
|
|
2.6 Present units of the user |
There is a tradition of presenting physical, easy to measure and to count units such as degrees centigrade, km/h, Euros and months. Cognitive task analysis unveils that units for targets and problem solving are different. Consequently, the units of a value should be transformed to a cognitive user’s unit (See Figure 5B).
| |
After entering your pin code, you confirm payment. The system should also ask you to confirm the consequences for your financial target: “OK, this shop will be paid € 500 but no restaurant dinners this month anymore. Is that what you really want? Press OK or cancel” The train indicator should know the distance from itself to the train indicated and present the walking speed needed to catch that train. When train speed is close to maximum speed, ETCS does present current speed and maximum permitted speed, of course, but time to intervention* becomes more prominent when the driver is close to the maximum permitted speed (See Figure 11).
| | |
*
Time to intervention: time from now to the moment the train safety control system will start braking because the driver is driving too fast, seems not to be able of braking anymore and the train will pass a signal at danger.
| | |
To top.
|
|
|
Neurath and Bertin carefully designed descriptive statistics in such a way that the visual units, such as surface, are compatible with the proportions of the content. Statistics should not lie. In target-based presentations, the visual units seen by the users should be compatible with visual and cognitive thresholds. In that case the visual quantities experienced e.g. the surface of the value, does not represent the measured quantity (e.g. the speed or 10%) but e.g. time left for user actions. Figure 8B presents a small square to the train driver, meaning much time to intervention by the safety system. Figure 11 present a large square, meaning little time to intervention.
To top.
| |
Presenting a departure time enables passengers, managers and politicians to notice delays. A walking speed is far more difficult to check. The train traffic controller could then present a higher walking speed to an already delayed departing train than the walking speed really needed. In doing so, he might discourage passengers from an arriving train trying to catch the departing train. This trying would delay the departing train even more and might have more serious unwanted consequences elsewhere in the network. Cognitive analyses show that for the benefit of the users there should be more lies in information, (Verhoef 2009).
|
|
2.7 Graphical figures |
| Alphanumerical forms and line charts are easy to draw with a pencil, ruler and typewriter. The fovea contains many thin cones, tightly packed together and not line-wise organized. Visual acuity and hue discrimination is best in the fovea and becomes poorer more eccentrically (Krech et al.1969). The shape of the fovea is a 5 degrees circle. The retina is not a relay station for isolated photosensitive visual elements; it is capable of extensive and complex interactions and integration (Krech et al. 1969 et al.). These characteristics of the visual system have direct consequences for the design of information.
To top.
| |
It is supposed that 5 degrees filled figures projected in the fovea take more advantage of these foveal characteristics than line figures and spreaded out over the retina. From a graphical point of view, it is easier to design a presentation for a value and its six parameters using a filled figure than using a line figure. The hypothesis is that presentations as in figure 12 allow faster psychological processing of the parameters of a value than alphanumerical and line presentation not fitting in a 5-degree circle. |
|
|
Written text, for instance, consists of line shaped forms (letters) and line shaped sentences. Incompatibility between the form of text and the fovea makes that while reading we perfectly perceive text of the previous line, which we have read, and text of the next line, which we should not read yet. Most of the text we can see we should not read at the moment we see it. That is not efficient.
To top.
| |
Characteristics of the cognitive activities of users also have consequences for the presentation of parameters of values. So far, it was suggested to present a figure for a variable. We can go one step further. In many cases, the values of two variables of a system interact and the potential conflict is the focus of the user. Consequently, it might be better not to present the two conflicting values independently but to integrate the space between them. Calculation and presentation of the space between variables reduces psychological load substantially. The conclusion can be in the fovea from a large distance (See Figure 8B, 13 and 14).
|
|
Another cognitive advantage of presenting space between variables and not separated is that the effect of simultaneous changes of two values is integrated as well. When two changes each make the situation worse, the graphical change is double. When two changes compensate one another there is no graphical change.
To top. |
|
When a train starts going uphill and the train slows down while the driver increases traction that speeds up the train, the time to intervention square (See Figure 11) will not change. When a train departure delay increases because the signal does not turn to green and the passenger decreases walking speed, the train indicators he sees while walking to his train will not change the walking speed the indicators present needed to catch the train.
|
|
3 From formal tables to…
|
|
So far, we discussed how to present six parameters of the value of one variable. Real life is a synthesis and much more complex. For performing today’s tasks, more than one variable is relevant. A common option to present more variables is a table. Tables presented on screen often meet restrictions of old technology more than they use the options of new screen presentation technology to meet psychological requirements.
Tables with borders for cells are evident with old mechanical type writing technology (See Figure 15). The lines are a visual scaffolding to position the numbers aligned. Once set, erasing the lines was not possible.
To top.
| |
The same applies for headings of tables. The design team and their computers need headings to discuss and know where to put the values in the table. When the meaning of the value of the cell is obvious, there is no need for an explanation in the heading for the user. It is unlikely that a passenger will read that his train departs at 30B hours from platform 15:15. In case of confusion about the units of the values presented in a table (km/h, Euros, degrees) should not be presented in a heading but at cell level to reduce short term memory load. In Figure 6D the heading ‘minutes to departure’ causes its column to occupy half the space of the indicator whereas for the minutes itself four positions would be sufficient. In addition there is no need for an explanation.
Printing headings on the frame of a passenger information screen, fixes the columns on the screen to a general maximum width (See Figure 6). Selecting a column width appropriate to the space the values on the screen need, is not possible.
The restrictions of old static paper technology are transposed to dynamic screen technology. The options new technology offers are not used. Using new options it is possible to present more information or to use smaller, cheaper and easier to position displays.
|
|
4 …content tables |
With new technologies, it is possible to bring more graphics into tables such as colors and bold type for text, background and borders. This can be used to mimic old technology, for aesthetic styling (See Figure 4A) or to improve readability.
To top.
| |
New technology should be used to present content. Without reading the numbers, the eye can understand the parameters of the values (See Figure 10).
|
|
5. From x-y forms to…
5.1 Visual perception |
Line charts (See Figure 16) are not compatible with the characteristics of human perception and filled figures are to be preferred (See section 2.7). The range between current value and target value of a variable in most cases is the interesting part of a scale.
To top.
| |
When the scales have a large range and the units on the scale are equal, then the resolution of these most interesting parts might be not legible (e.g. where two lines cross). Some kind of zooming is needed.
| |
|
|
5.2 Visual routing |
For reading text in the West there is an eye routing convention: from top left to bottom right. Line charts spread information out over an area larger than the fovea. Nevertheless there are no conventions for eye routing reading graphs. The icons of Isotype* are attractive and eye catching (See Figure 17). At the time of their invention attractiveness was an important requirement. However, the icons attract attention whereas attention should be at the position just after the last icon. That position indicates the value presented.
| |
Gapminder (See Figure 7) takes advantages of modern technology in an intelligent way. The basis is a traditional xy line graph. Gapminder controls eye routing using a man on the stage** telling where to look. Another solution Gapminder uses is to project the graph on a casino table, again adding a flesh and blood man as a croupier and giving the users tokens to play with (Rosling, 2009). As with the Isotype graphs, attractiveness is beyond doubt. There is a danger that making statistics attractive distracts designers from the problem: how to make statistics understandable.
A compromise might be story telling. Story telling has exactly what graphs are missing: a strong linear structure. The route in a graph could be presented using numbered text balloons in graphs (Birt, 2009). The text of the man on the stage could be in the balloons.
| | |
* Isotype is a system of pictograms designed by the Austrian educator and philosopher Otto Neurath, to communicate quantitative information with social consequences in a simple way.
| | |
** For a demo, search google video with keywords: ‘gapminder hans rosling’. Gapminder founder Hans Rosling controls attention of the graph readers using very attractive gestures and spoken text.
| |
To top. |
|
5.3 Short term memory |
Humans read graphs by keeping parts of the graph in short term memory. In figure 16 short term memory is loaded with the meaning for four lines and the positions of these lines in the graph from one to another. Even for this simple graph several seconds are needed to inspect and to retain all values. This might not be a problem for experienced professional graph readers interested in the data. In graphs for public information several seconds might be too long. Not understanding short term memory by the designer of the graph might be explained as users not understanding the graph.
To top.
| |
The effect of spreading out on understanding can be so strong that impossible figures become possible (See Figure 18). Research indicates that cues for the correct interpretations of these kinds of figures only have effect when presented near the attended area (Peterson & Gibson, 1991). When an impossible figure is positioned in the fovea, the eye only can see the impossibilities (See Figure 19).
|
|
The Isotype icons provide a perfect solution for the short term memory problem. Where ever the focus is, the icons show the meaning of the content. There might be some memory load to learn the meaning of the icons, but being very self-evident and concrete, once read, the meaning will not impose a load on memory. Short term memory load is far less than commonly used arbitrary coding with line types (See figure 16). This Isotype advantage is applied in figure 20.
Gapminder does not have ‘aide de memoires’ for the values presented. However, the user selects the x-y variables himself and that reduces the short term memory load. The other variables (time, continent, and quantity) always are presented in the same way.
To top.
| |
Traditionally axes are presented as straight lines. Short term memory load could be reduced presenting the values of x and y where needed. Using floating axes alphanumerical labels of the values can follow the data. The effect should be that on interesting spots in the graph all values and their labels are present and within the fovea (x value, y value, variable 1 score, variable 2 score). See Figure 20.
|
|
5.4 Understanding |
Suppose the graph meets all psychological requirements. Can the user understand the information? Isotype makes a value concrete, filling up the bar with icons each representing a proportion of the number to be presented.
Research from developmental cognitive psychologist Jean Piaget clearly indicates that for five year olds there is no need to subdivide a number. For a five year old the question: “What is the height of the pile of pancakes you want?” is no problem at all. He will point with his hands as high as he can to indicate an imaginary pile. There is much empirical data supporting this hypothesis: “+Piaget +conservation” gives today 182.000 hits on Google. Piaget (1896-1980) was a contemporary of Neurath. Young children do not find the spatial demands of graphs at all difficult (Bryant & Somerville, 1986). Using concrete piles, Isotype intended to provide insight for schoolchildren (Burke, 2009 ;) and workers. Maybe Isotype tried to solve an understanding problem that did not exist.
To top. |
|
In the twenties, time of departure and platform number was sufficient for a passenger. Today’s passengers have to be informed in a split second to be able to choose between trips that differ in price, travel time, comfort, environmental load and reliability of these variables. For the train drivers in the twenties, the physical red lights beside the track were the focus of attention. Today’s high speed train drivers have to understand that they drive in three electronic dynamic cocoons at the same time. Each safety, schedule and energy cocoon has its own current and target speeds and its own control of attention curves. It is impossible to present all these values using piles of icons for red lights, green trees and shouting delayed passengers. Using piles of concrete icons does not solve the problem of today’s train driver, of tomorrow’s passenger and of after tomorrow’s bank customer.
The recent Neurath-revival (Burke, 2009; Engelhardt & Zambrano, 2009) is deserved but cannot solve today’s information presentation problems. Going back to finger counting is not the way to go for information design to present multivariate interactions. We have to go forward, find parameters of values (e.g. reliability, trend, point zero, etc.), find the abstract cognitive variables that are relevant for a task, find their interactions and find out how to present all this. What is beyond Isotype?
|
|
Isotype focuses on proportional relations of a few variables. Gapminder focuses on interactions between five variables. That is closer to today’s problems. However, Gapminder might be captured in the traditional x-y model. Using modern technology, Gapminder cleverly added three dimensions to the xy-model: (1) time using a play function, (2) location/continent using color and (3) circle diameter.
To top.
| |
The visual inconsistencies and the inconsistent behavior of the scales of these five variables are the bottleneck for this five dimensional presentation. The inconsistencies increase psychological load and impair the visual identification of specific situations and interactions. This specific combination of five values might not produce an easy to identify and to remember picture. What is beyond Gapminder?
|
|
6. …squares |
The x-y model is perfect for the traditional empirical analytic two variable approach. For cognitive psychologists educated in this tradition it might be difficult to find new synthetic presentations. They might know that there are other traditions, such as the Feyerabend (1984) one, with books titled: ‘Against method’.
To top.
| |
Unfortunately a new more ‘chaotic’ synthetic methodology was never elaborated. This gives cognitive psychologists a blind spot for the synthetic character of reality and design where all variables are active at the same time. Cognitive psychologists working in practice will have to solve that huge methodological problem themselves (Verhoef, 2007).
|
| 6.1 A solution |
One solution is more than one hundred years old. Mendeleyev’s periodic table of the elements for chemistry is a scientific synthesis of a body of knowledge that up to that moment was analytic, chaotic and anarchistic. That sounds familiar. A multidimensional model might solve the problem for today’s chaotic usability ‘science’ and interface design theory (Ruecker & Liepert, 2006; Verhoef, 2007). Multidimensional orthogonal structures might also be applicable to the presentation of interacting multidimensional variables. That is exactly what Bertin (1967) did on paper for the presentation of demographic information.
To top.
| |
How to elaborate this solution for information presentation? The traditional model has two axes, one vertical axis, one horizontal axis, and they meet at the bottom left in an absolute point zero. Why not have more axes, not meeting at point zero? No piles of icons, bars and pies, but positions only. In addition, of course, to reduce short term memory load the presentation should fit in the fovea. The scales should present the six parameters specified in section 2.1-2.6.
|
| |
Finally, the visual result should be something that can easily can be described using words to get a ‘forme mémorisable’ as Bertin (1967) calls it. Figure 21 shows a four variable solution. The scales are positioned in a cross and the user’s point zeros are connected. A perfect square means that the values of all four variables are on target. When the value of variables departs from target, the square becomes distorted. Small deviations from horizontal and vertical are easily noticed by the human eye and indicate the trend (getting too much or getting too little). Differences between large deviations from horizontal are more difficult to evaluate. Manipulating the units of the scale and using colors for attention as suggested in section 2.3 solves this problem. The deviations should form easy identifiable and verbalisable figures, representing a typical status of the process such as a wizard’s pointed head or beard.
To top.
| |
It is frequently commented that such ‘complicated’ presentations require instruction and learning time. Our experience is that novice users of interfaces based on cognitive psychology perform much better than experienced users with the traditional interface. When the four corners of the square stand for safety, schedule, energy and passenger comfort, one can expect that without any explanation within one short trip an experienced train driver will understand the presentation. If learning and understanding is a problem, modern technology offers many solutions. Nevertheless, it should not be a problem. The presentation should be compatible with the mental model of the user.
|
| 6.2 Validation
|
Multidimensional interacting presentations have been investigated during several decades. Initially they were called integrated displays (Beringer, 1987). Brown (1985) plotted 16 toxic substances using human faces to enable quick identification of the severity of poisons swallowed by humans. The face for water would be a normal face and the face for a deadly poison would be as far as possible from normal suggesting emptying the stomach of the patient immediately. Brown did not use the xy-axis model for his multidimensional smileys. Their form is compatible with the fovea. In addition, specific combinations of face variables can result in an easy identifiable visual configuration such as a happy face and a scared face. Figure 22 shows multidimensional smileys. Ware & Beatty (1988) used color and space to present clustered points of a multidimensional data set (See Figure 23).
To top.
| |
Empirical evaluations of these kinds of multidimensional presentations show that human performance is moderately better. This is an interesting conclusion since higher compatibility with human physiology, perception, cognition and the process controlled, should give a much enhanced performance.
|
| |
|
Figure 1. Visual – content compatibility of digital and analogue presentations
Figure 2. Clock scale, implicit presentation of static targets
All dials are positioned having the target value at the top. The effect of this implicit static target is that an off target value is conspicuous.
Figure 3. Clock scale, speedometer for high speed trains with two targets
Figure 3A.
Train speed: 192.
Maximum permitted speed 200.
Next target permitted speed 40.
Figure 3B.
A scale with a hand for fast but inaccurate perception.
A scale with a triangle for slow but accurate perception.
(ETCS, experimental design, approx. 1994).
Figure 4. Table, OECD quarterly growth table
Figure 4A.
The graphics are used for aesthetics only. This form is the OECD house style that also is used for signposting.
Figure 4B.
The table is a flat landscape without cognitive landmarks to navigate.
(http://stats.oecd.org/Index.aspx)
Figure 5. Scale with several cognitive psychological parameters of the value to be presented.
Figure 5A.
User’s point zero (OK) in the middle of the scale.
The distance between the middle of the scale (the target, OK) and the current value (the pointer, not shown here) is the cognitive focus of the user.
Figure 5B.
Interpretation of car speed using colour:
Dangerous, orange: driving too slowly for this (high) way.
OK, white: mean speed for this road (maximum speed in most cases).
Too fast, yellow: too fast but within limits for over speeding.
Too fast, orange: too fast, you might get a fine.
Far too fast, red: You will loose your driving licence when caught.
(Holslag & Verhoef, experimental demo design)
Figure 6. One trains indicator, different point zeros
Figure 6A.
For Netherlands Railways (trains at the top) point zero is midnight.
Figure 6B.
For Amsterdam Public Transport (metros at the bottom) time of departure is point zero. Count down time is presented.
Figure 6C.
Minutes to departure for metros is presented using leading zeros (bottom right).
Figure 6D.
The heading ‘minutes to departure’ is superfluous and causes that half of the dynamic space cannot be used.
(Indicator for trains and busses, Amsterdam Arena, 2008)
Figure 7. Graph, Gapminder
Five dimensions in a xy model. X and y for population characteristics, the area’s (circles) for continent, ‘play’ for time/trend/history and circle size for quantity. (www.gapminder.org)
Figure 8. Trend for train speed
Figure 8A.
At the right of this screen maximum train speed profile ahead.
Lower maximum speed at 1000 m.
Figure 8B.
The small square at the top left shows that there is sufficient time to safety system brake intervention. A large square means little time to intervention. The driver can notice the square even when looking outside. (ETCS)
Figure 9. Scale showing reliability of the value presented
Reliability presented using shading, e.g. less shading less uncertainty, much shading much unreliability. (Holslag & Verhoef, experimental demo design).
Figure 10. Table with graphics for content
Figure 10 A.Reliability
Darker numbers means more reliable. The human eye can conclude that reliability of the values at the bottom is lower.
Figure 10B. Interpretation for control of attention
Column 1: no colour, no problems.
Column 2: alarm colours coming up (2 yellows, 1 orange).
Column 3: alarms whole column, all orange, problems everywhere.
Figure 10C Target
Quarter two is set as target; next quarters are presented in proportion to the target.
(OECD, see Figure 4, adapted by Holslag & Verhoef)
Figure 11. A square, showing time to intervention
Train braking curve is close to safety braking curve. The square top left shows time to intervention of the safety system. Time to intervention depends on maximum permitted speed minus current speed and the traction/brake application. Figure 8 shows a smaller square, meaning more time to intervention. (ETCS)
Figure 12. A triangle, presenting five parameters of a value
One value and its parameters:
Current value: the middle of the triangle, at the OK position.
Target value: thick vertical line, current value is target value.
Trend: slope to the right, value is decreasing.
Interpretation: yellow; attention is needed.
Reliability: shading, current value most probable at target value.
(Holslag & Verhoef, experimental demo design).
Figure 13. Train indicator, space between time variables
Platform train indicator. The triangle presents minutes to departure.
Left long leg: position of the minutes hand at time of departure.
In this case the 31 minutes position.
Right long leg: position of the minutes hand at current time.
In this case the 19 minutes position.
The conclusion is in the fovea, even from a large distance.
(NS, Amersfoort 2008, adapted by Holslag & Verhoef, experimental demo design)
Figure 14. Rectangles presenting reliability between conflicting variables
Each line shows a value left and a value right.
Unreliability indicated by shading.
A: values close separation is sufficient no problem.
B: values distant separation very large no problem at all.
C: values close separation small take care!
D: values distant separation large no problem.
In case of a conflict the control of attention colors can be used (see section 2.3).
(Holslag & Verhoef, experimental demo design)
Figure 15. Table, no graphics
No colours or bold type in this table. Lines are needed to position the data.
(OECD, adapted by Holslag & Verhoef for demo purposes.)
Figure 16. Traditional x–y line chart
With four variables the short term memory load is rather high.
Little space is left for problem solving. (Holslag & Verhoef, experimental demo design)
Figure 17. Isotype
Isotype is a system of pictograms designed by the Austrian educator and philosopher Otto Neurath, to communicate values of variables with social consequences in a simple way. (Holslag, inspired by Neurath).
Figure 18. Impossible figure
Details of the left part and details of the right part cannot be visually inspected simultaneously. There is an overview only in short term memory. There details are lost and details unavailable are reconstructed to form a sensible interpretation. See also figure 19.
Figure 19. Impossible figure made possible
Details of the left part and details of the right part fit in the fovea. There is an overview within the eye. The eye only can see the impossibilities.
Figure 20. Line graph, floating axis
The labels of the axes follow the data.
On the interesting spots all values are available in the fovea.
Fixating on the top right, the details and the conclusion are available at a glance: in the year 2025 the income of man and women is almost equal. (Holslag & Verhoef, experimental demo design)
Figure 21. Squares, presenting interacting variables
Each corner presents a variable.
Left: Perfect square, all four variables are on target (no colours).
Middle: Top right variable is off target, level acceptable (yellow).
Right: Top right variable is far off target, level unacceptable (orange).
(Holslag & Verhoef, 2009)
Figure 22. Faces presenting interacting variables
Four dimensional smileys:
Eye for variable 1: Several diameters.
Nose for variable 2: Several thicknesses.
Mouth for variable 3: Several thicknesses.
Misty for parameter: reliability
Color for parameter: control of attention.
(Holslag & Verhoef, experimental demo design)
Figure 23. Multi variable presentation of a database
Variables are presented using position (x-y), color, luminance and circle size.
The concept as it was applied for the customer is modified for copyright reasons.
Colour is essential but not present in this picture. Holslag & Verhoef.
|
|
References | Barnett, B.J. & Wickens, C.D., (1988). Display Proximity in Multicue Information Integration: The Benefits of Boxes. Human Factors, 30, (1), 15-24.
Beldie, I.P., Pastoor, S., & Schwarz, E., (1983). Fixed versus Variable Letter Width for Televised Text. Human Factors, 25, (3), 273-277.
Beringer, D.B., (1987). Peripheral integrated status display. Displays, 8, (1), 33-36.
Bertin, J., (1967). Semiologie graphique, Les diagrammes - Les réseaux Les cartes. Paris - La Haye: Mouton, Gauther-Villars.
Birt, A., (2009). Background Stories - visual communication for sustainability. Paris: IIID and OECD conference DD4D, Data designed for decisions, Enhancing Social, economic and environmental progress.
Brase, G.L. & Stelzer, H.E., (2007). Education and Persuasion in Extension Forestry: Effects of Different Numerical Information Formats. Journal of Extension, 45, (4). http://www.joe.org/joe/2007august/a1.php.
Brown, R.L., (1985). Methods for the graphic representation of systems simulated data. Ergonomics, 28, (10), 1439-1454.
Bryant, P.E. & Somerville, S.C., (1986). The spatial demands of graphs. British Journal of Psychology, 77, (2), 187-197.
Burke, C., (2009). Isotype - representing social relationships pictorially. Paris: IIID and OECD conference DD4D, Data designed for decisions, Enhancing Social, economic and environmental progress.
Campbell, A.J., Marchetti, F.M., & Mewhort, D.J.K., (1981). Reading speed and text production, A note on right-justification techniques. Ergonomics, 24, (8), 633-640.
Creed, A., Dennis, I. & Newstead, S., (1987). Proof-reading on VDU's. Behaviour and information technology, 6, (1), 3-13.
Duncan, J., (1990). Goal weighting and the choice of behaviour in a complex world. Ergonomics, 33, (10/11), 1265-1279.
Engelhardt, Y. & Zambrano, N., (2009). Engaging Citiziens with Animated Statistics - from Neurath to Gapminder. Paris: IIID and OECD conference DD4D, Data designed for decisions, Enhancing Social, economic and environmental progress.
Fénix, J., Graffagnino, T, Sagot, J.-C. & Valot, C., (2005). User centred design applied to increase timetable stability, Moderne Schienenfahrzeuge, 129, 32-37.
Feyerabend, P., (1984). Against method, Outline of an anarchistic theory of knowledge. New York: Schocken Books.
Keuning, C. & Roding, R., (2008). Lezen en wegwezen. Verkeerskunde, 59, (4), 26-31.
Krech, D., Crutchfield, R.S. & Livson, N., (1969). Elements of psychology. New York: Alfred A. Knopf inc.
Peterson, M.A. & Gibson, B.S., (1991). Directing Spatial Attention Within an Object: Alerting the Functional Equivalence of Shape Description. Journal of Experimental Psychology: Human Perception and Performance, 17, (1), 170-182.
Rosling, H., (2009). How to Increase Innovation in the Use of Statistics. Paris: IIID and OECD conference DD4D, Data designed for decisions, Enhancing Social, economic and environmental progress.
Ruecker, S. & Liepert, S., (2006). Taking Medeleyev's correspondence course: Interface design lessons from the periodic table of the elements. Information Design Journal, 14, (3), 236-245.
Spivey, M.J., (2007). Redesigning our theories of human information processing. Information Design Journal, 15, (3), 261-265.
Treurniet, W.C., (1980). Spacing of characters on a television display. In: Kolers et al. Processing of visible language 2, pp. 365 - 374.
Verhoef, L.W.M., (2007). Why designers can't understand their users, Developing a systematic approach using cognitive psychology. Utrecht: Human Efficiency. www.humanefficiency.nl/designers_understanding.shtml.
Verhoef, L.W.M., (2008a). Presenting numbers to teachers, train drivers and travelers. In: Applications of Information Design 2008IIID, International Institute for Information Design, www.humanefficiency.nl/gui/teachers_drivers_travellers.shtml.
Verhoef, L.W.M., (2008b). Vertrektijd is passé, leve de afteltijd, Vertrektijd laat zien in welke eeuw de bordenmaker leeft. OV-magazine, 3 juli, 26-27. www.humanefficiency.nl/public/vertrektijd_is_passe.pdf.
Verhoef, L.W.M., (2008c). Hoe onderzoek je het denken van reizigers. Verkeerskunde, (7), 63. www.verkeerskunde.nl/moxie/podium/hoe-onderzoek-je-het-denk-3.shtml.
Verhoef, L.W.M., (2009). Why car park signs should lie. Traffic Engineering & Control, 50, (7), 2-3. www.humanefficiency.nl/public/why_car_park_signs_should_lie.shtml.
Ware, C. & Beatty, J.C., (1988). Using Color Dimensions to Display Data Dimensions. Human Factors,30, (2), 127-142. www.ingentaconnect.com/content/hfes/hf/1988/00000030/00000002/art00001?crawler=true.
To top. |
future systems
etcs mmi
