UCS seminar: Thinking spatially...through visualization

Talk given to sixth form geography students as part of a seminar series at University College School, September 2015

Transcript

1. Thinking spatially … through visualization Roger Beecham, gicentre.net/
2. Dr. Roger Beecham, gicentre.net/ giCentre Creative User-Centered Visualization Design for Energy Analysts and Modelers Sarah Goodwin, Jason Dykes, Sara Jones, Iain Dillingham, Graham Dove, Alison Duffy, Alexander Kachkaev, Aidan Slingsby, and Jo Wood, Member, IEEE CHANGE PROPORTION OF APPLIANCE CONSUMPTION SHIFTED FROM ‘SHRINK’ PERIODS TO ‘GROW’ PERIODS TOOLS FOR DATA SCULPTING CONSUMPTION WILL SHRINK DURING THIS PERIOD CONSUMPTION WILL GROW DURING THIS PERIOD CLICK AND DRAG ON TIMELINE TO SELECT PERIODS SHIFT TO HERE ... FROM HERE CLOTHES DRYER Fig. 1. Demand Horizons show modeled weekday energy demand over 24 hours amongst high consumption domestic appliances. Data Sculpting allows us to shift consumption interactively by ‘moulding’ the horizons to explore ‘what if?’ scenarios. For example, here fifty percent of ‘Clothes Dryer’ consumption is shifted from the evening peak to a period when overall demand is lower. Abstract—We enhance a user-centered design process with techniques that deliberately promote creativity to identify opportunities for the visualization of data generated by a major energy supplier. Visualization prototypes developed in this way prove effective in a situation whereby data sets are largely unknown and requirements open – enabling successful exploration of possibilities for visualization in Smart Home data analysis. The process gives rise to novel designs and design metaphors including data sculpting. It suggests: that the deliberate use of creativity techniques with data stakeholders is likely to contribute to successful, novel and effective solutions; that being explicit about creativity may contribute to designers developing creative solutions; that using creativity techniques early in the design process may result in a creative approach persisting throughout the process. The work constitutes the first systematic visualization design for a data rich source that will be increasingly important to energy suppliers and consumers as Smart Meter technology is widely deployed. It is novel in explicitly employing creativity techniques at the requirements stage of visualization design and development, paving the way for further use and study of creativity methods in visualization design. Index Terms—Creativity techniques, user-centered design, data visualization, smart home, energy consumption 1 INTRODUCTION These are exciting times for utility companies and their energy analysts – the energy domain is data rich and globally significant. Energy an- alysts and modelers are now striving to effectively use the volumes of data from emerging Smart Home technologies to understand consumer behavior, conserve energy and manage supply and demand. Data vi- sualization can offer great potential in this domain, but developing ap- propriate solutions presents considerable challenges, since the nature of the data are relatively unknown and the needs of energy data an- alysts and modelers are not yet well understood. The design brief is therefore essentially open-ended. • Sarah Goodwin, Jason Dykes, Iain Dillingham, Alexander Kachkaev, Aidan Slingsby, Jo Wood are with the giCentre, City University London. E-mail: {Sarah.Goodwin.1, J.Dykes, Iain.Dillingham.1, Alexander.Kachkaev.1, A.Slingsby, J.D.Wood}@city.ac.uk. • Sara Jones, Graham Dove and Alison Duffy are with the Centre for Creativity in Professional Practice, City University London. E-mail: {S.V.Jones, Graham.Dove.1}@city.ac.uk, Alison@perspectiv.co.uk. Manuscript received 31 March 2013; accepted 1 August 2013; posted online 13 October 2013; mailed on 4 October 2013. For information on obtaining reprints of this article, please send e-mail to: tvcg@computer.org. Participatory approaches to user-centered design, in which users and other stakeholders are involved in co-creating requirements and designs for interactive systems can lead to solutions that are more use- ful and usable [35]. We have successfully used human-centered ap- proaches in the design of visualization solutions before and have doc- umented these in detail [27]. However, the role of creativity in these approaches has as yet been only implicit. Over the last decade some fields of interactive systems development have increasingly focussed on introducing elements of deliberate creativity into participatory user- centered design processes. The aim here is to enable all participants (users, designers and other stakeholders) to contribute to the explo- ration of new fields and the generation of requirements and design ideas for novel and useful systems [1, 6, 53]. Establishing require- ments can be considered a fundamentally creative process whereby requirements analysts and stakeholders work collaboratively to gener- ate ideas for software systems [29, 30, 32]. Indeed, Robertson [42] regards requirements analysts as inventors who bring about innovative change in designs to establish advantage. Techniques for deliberately introducing creativity into the process of user-centered design can be used effectively in this context. For example, Schmid [46] used cre- ativity triggers [42] to help workshop participants invent requirements, whilst co-creation [45] and creativity workshops [24, 31] have been shown to be effective in generating novel requirements.
3. Waldo Tobler, 1970 Everything is related to everything else, but near things are more related than distant things …[spatial autocorrelation]
4. Waldo Tobler, 1970 Everything is related to everything else, but near things are more related than distant things. [space is a special unit of analysis]
5. Hue OAC category (grey for all) as selected in C Shading 1st-9th declines can be turned off Axes 41 census variables can be reordered Labels Off can be turned on or available on mouseover F Spatial selections List "None! and "all! as default can add more identified in A, illustrated with thumbnail image Selection [Display]: all OAs, affects A and E can be set to be any number of categories [Baseline]: none selected, affects E can be set to be any category Fig. 3. Screenshot of the software that implements our design as described in section 4. Details listed on the right. Here, all OAs are selected (see [display] in F). A is zoomed to the NR postcode (overview in B). Mouse cursor indicates a ‘Countryside’ OA that is also quite similar to two other super-groups (D). Its population profile (41 census variables) is the black line in E shown alongside two national profiles indicated by [display] in C. accessible. This suits demonstration and chauffeuring well because functionality is only revealed when required; a fact positively com- mented upon by one of the participants in our evaluation. There is evidence that cluttered user interfaces can detract from the data being shown [40] and that aesthetics can have a significant effect on user experience [6]. These considerations have strongly influenced our de- sign. 4.1 Colour Perceptually evenly-spaced hues of equal lightness [56, 53] depict super-groups (×7), shown in Fig. 2 and used consistently across views. Groups’ (×21) colours are derived from their parents’ hues and are also perceptually equally spaced. Although too indistinguishable for lookup tasks, they allow heterogeneity in adjacent areas to be detected. Since lightness is held constant, it is available to encode other in- formation. Perceptually-uniform variations in lightness represent the relative similarity to allocated category (section 4.4), such that light- ness can be directly compared across hues [56]. As the similarity de- creases, colours converge to white indicating that the allocated cate- gory is a poor characterisation of the area. Lightness is considered an appropriate visual variable for this kind of information [31]. The result is a categorical sequential bivariate colour scheme [4] showing cate- gory and uncertainty as is appropriate for encoding attribute accuracy for categorical data in areal coverages [47]. 4.2 Geographical distribution Each Output Area (OA) is assigned an OAC category. The dot map in Fig. 4 reveals that most of the land area contains OAs classified as ‘Countryside’, yet Fig. 2 shows this is a relatively small propor- tion of the population. The zoomed-in portion in Fig. 4 (right) reveals huge variations in OA density, illustrating the difficulty of producing national population maps [42]. Density-normalising population cartograms size geographic areas by population. By distorting geographical space, the visual promi- nence of areas becomes more proportional to population. This has Fig. 4. Centroids of OAs coloured by super-group. Labels indicate post- codes. Left: National map shows most of the land area is ‘Countryside’. Right: The zoomed-in portion centred on London shows marked geo- graphical differences in OA density and classification. Fig. 14. Non-sketchy and sketchy visualizations for Scenario 2. Inset on the bottom showing detail of the sketchy rendering. Fig. 15. Non-sketchy and hybrid visualizations for Scenario 3. 6.4 Participants We recruited 52 unpaid participants (40 male, 12 female) via personal e-mails and social networking posts to participate in our Web-based evaluation. Participants were predominantly in the 22–34 years age group (31 participants) and 11 participants were between 35–44 years of age. All but three participants reported to have at least monthly exposure to data charts (19 daily, 15 weekly, 15 monthly). 6.5 Results We analyzed the number of annotations per condition and task as well as the length of annotations as a way to study annotation coverage. We Fig. 16. Non-sketchy and sketchy visualizations for Scenario 4. Fig. 17. Non-sketchy and sketchy visualizations for Scenario 5. also analyzed the types of annotations people made quantitatively and assessed the quality of the critique for Task 5. Overall, we found rela- tively high and equal engagement in both conditions based on the quan- titative measures, with a median of 7 annotations in the sketchy con- ditions and 6 in the normal condition and a slightly longer annotation length in the normal condition. Overall, however, these differences were not statistically significant. We, therefore, focus on reporting our qualitative assessment of the difference between the two conditions. Scenario 1: In this hybrid scenario only part of the data was drawn in a sketchy style for SC. For both conditions participants drew the same type of annotations including trendlines, highlights of data items and ranges, data labels and comments—all only with slightly varying counts between the conditions. 29% of participants in SC used free- hand annotations but also 25% of participants in the normal condition did so. We also analyzed the spatial distribution of comments across the charts that referred to the past (always drawn non-sketchy) and the future (drawn in sketchy style in SC). We observed a slight trend for participants to draw more annotations in SC that referred to the future (53% of annotations) compared to the normal condition where only 47% of annotations referred to this part. Overall, however, the difference between the annotation styles was relatively small. Scenario 2: In this scenario participants annotated a bike map of London. We found that they drew slightly more annotations in SC (median = 10) vs. NC (median = 7). The type of annotations in both conditions were fairly similar and largely concentrated on highlighting data regions with circles. In SC, 46% of participants drew at least one freehand annotation while only 36% did so in NC. Labels were signif- icantly more common in SC (p = .044 using a Mann-Whitney U test). 46% of participants in SC used labels to provide a legend for their an- notations while only 11% of participants in NC did so. Considering the spread of annotations (Fig. 18) we noticed a tendency for partici- pants in NC to highlight the same four regions, while the annotations in SC seemed to be more spread across the map. Scenario 3: This was a hybrid scenario with more personal data than the company stock data of Scenario 1. We saw a stronger differ- ence between SC and NC for this task. The overall annotation count Fig. 4. Sketchy spatial treemaps of London boroughs (left) and London wards (right). Fig. 5. Sketchy Processing primitives. enabling interaction in a sketchy style. The degree of a line’s sketchi- ness is controlled by a roughness parameter r measured in pixel units. r defines the radius within which the two endpoints of the line are ran- domly perturbed (with a uniform random deviate). In addition, approx- imately 75% along the length of the line an additional vertex is added at a random displacement within r, orthogonal to the straight line AB. To avoid aliasing effects when close parallel lines are drawn, the po- sition of the displacement is itself randomly selected with a range of 10% of the distance AB. Finally, to mimic hand-drawn control, the line is ‘bowed’ by adding a midpoint vertex randomly perturbed within 0.5% of AB. The four vertices are joined using a Catmull-Rom spline (implemented in Processing with the curveVertex() command). To create a more sketchy effect, each line is rendered twice with uniform random selection of displacement vertices. Increasing r can thus in- r s o e o Fig. 7. Parameterization of sketchy ellipses. The Handy library provides the option to fill all closed shapes with hachures (see Fig. 5). These are implemented by simply drawing lines in sketchy style within the outline shapes, as inspired by other work [1]. Line intersection between the unsketchy geometry is used to define the endpoints A and B necessary to contain hachures approximately within the shape boundary. The visual appearance of the rendered shapes can be further customised by changing fill colors, line thicknesses, hachure density, and hachure angle. To facilitate production of sketchy graph- [visualization is an approach especially suited to spatial data]
6. Case study 1: urban cycling behaviour in Central London
8. dbase  schema   Stations stationID shortName easting northing capacity INTEGER TEXT INTEGER INTEGER INTEGER 1 River9St\nEC2 531202 182838 19 2 Phillimore9Gdns\nW8 525207 179398 37 3 Christopher9St\nEC2 532984 182007 32 4 St.9Chad's9St\nWC1 530436 182918 23 5 Sedding9St\nSW1 528050 178800 27 . . . . . . . . . . . . . . .
9. dbase  schema   summary  scripts  
21. Case study 2: local elections in Greater London May 2010
22. CHADWELL VOTE FOR NO MORE THAN THREE CANDIDATES Gertrude Chadwell 22 Some St, London N1 2AB UK Independence Party 2 CROUSE Justin Crouse (Address in constituency) The Labour Party Candidate 3 AARON Lawrence Aaron 17 Newington Road, London N1 6FG Liberal Democrats 1 DEBOSE Joanne Debose 16 Acer Avenue, London NW4 8XT Green Party 4 HANDY William Handy (Address in constituency) The Labour Party Candidate 5 HOOPER Malcolm Hooper (Address in constituency) The Conservative Party Candidate 6 NOOR Anjit Noor (Address in constituency) The Labour Party Candidate 8 PFEIFFER Dale Pfeiffer 103 Elephant Way, London NW1 8RH Liberal Democrats 9 KOZLOWSKI Michael Kozlowski (Address in constituency) The Conservative Party Candidate 7 TALLY Deborah Tally (Address in constituency) The Conservative Party Candidate 10 WHITFIELD Sarah Whitfield 45 Kingham Place, London N1 6SL Liberal Democrats 11 YILMAZ Shaquil Yilmaz 4 Pocklington Walk, London N1 5DS Independent Candidate 12 5973 candidates 5025 from the three major parties 1842 elected to office
23. CHADWELL VOTE FOR NO MORE THAN THREE CANDIDATES Gertrude Chadwell 22 Some St, London N1 2AB UK Independence Party 2 CROUSE Justin Crouse (Address in constituency) The Labour Party Candidate 3 AARON Lawrence Aaron 17 Newington Road, London N1 6FG Liberal Democrats 1 DEBOSE Joanne Debose 16 Acer Avenue, London NW4 8XT Green Party 4 HANDY William Handy (Address in constituency) The Labour Party Candidate 5 HOOPER Malcolm Hooper (Address in constituency) The Conservative Party Candidate 6 NOOR Anjit Noor (Address in constituency) The Labour Party Candidate 8 PFEIFFER Dale Pfeiffer 103 Elephant Way, London NW1 8RH Liberal Democrats 9 KOZLOWSKI Michael Kozlowski (Address in constituency) The Conservative Party Candidate 7 TALLY Deborah Tally (Address in constituency) The Conservative Party Candidate 10 WHITFIELD Sarah Whitfield 45 Kingham Place, London N1 6SL Liberal Democrats 11 YILMAZ Shaquil Yilmaz 4 Pocklington Walk, London N1 5DS Independent Candidate 12 To what extent does a candidate’s name influence the number of votes received during an election?
27. random
28. Labour First Third Liberal Democrat Conservative Second First Third Second First Third Second
29. Labour First Third Liberal Democrat Conservative Second First Third Second First Third Second random
30. Case study 3: visual approaches to probabilistic reasoning
31. How risky is cycling in London?
32. 0 100 200 300 400 500 600 700 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Number of incidents Cyclists Killed or Seriously Injured, Greater London Serious Injuries Fatalities
33. “denominator neglect”
34. 0 50 100 150 200 250 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Annual number of journeys (millions) Incidents per million journeys Bicycle journeys and KSI rate, Greater London Serious injuries Fatalities Number of journeys
40. In 2012 there was one serious injury for every 0.3 million bicycle journeys made.
42. What are the chances of being killed if you make a a bicycle journey this afternoon?
47. giCentre Creative User-Centered Visualization Design for Energy Analysts and Modelers Sarah Goodwin, Jason Dykes, Sara Jones, Iain Dillingham, Graham Dove, Alison Duffy, Alexander Kachkaev, Aidan Slingsby, and Jo Wood, Member, IEEE CHANGE PROPORTION OF APPLIANCE CONSUMPTION SHIFTED FROM ‘SHRINK’ PERIODS TO ‘GROW’ PERIODS TOOLS FOR DATA SCULPTING CONSUMPTION WILL SHRINK DURING THIS PERIOD CONSUMPTION WILL GROW DURING THIS PERIOD CLICK AND DRAG ON TIMELINE TO SELECT PERIODS SHIFT TO HERE ... FROM HERE CLOTHES DRYER Fig. 1. Demand Horizons show modeled weekday energy demand over 24 hours amongst high consumption domestic appliances. Data Sculpting allows us to shift consumption interactively by ‘moulding’ the horizons to explore ‘what if?’ scenarios. For example, here fifty percent of ‘Clothes Dryer’ consumption is shifted from the evening peak to a period when overall demand is lower. Abstract—We enhance a user-centered design process with techniques that deliberately promote creativity to identify opportunities for the visualization of data generated by a major energy supplier. Visualization prototypes developed in this way prove effective in a situation whereby data sets are largely unknown and requirements open – enabling successful exploration of possibilities for visualization in Smart Home data analysis. The process gives rise to novel designs and design metaphors including data sculpting. It suggests: that the deliberate use of creativity techniques with data stakeholders is likely to contribute to successful, novel and effective solutions; that being explicit about creativity may contribute to designers developing creative solutions; that using creativity techniques early in the design process may result in a creative approach persisting throughout the process. The work constitutes the first systematic visualization design for a data rich source that will be increasingly important to energy suppliers and consumers as Smart Meter technology is widely deployed. It is novel in explicitly employing creativity techniques at the requirements stage of visualization design and development, paving the way for further use and study of creativity methods in visualization design. Index Terms—Creativity techniques, user-centered design, data visualization, smart home, energy consumption 1 INTRODUCTION These are exciting times for utility companies and their energy analysts – the energy domain is data rich and globally significant. Energy an- alysts and modelers are now striving to effectively use the volumes of data from emerging Smart Home technologies to understand consumer behavior, conserve energy and manage supply and demand. Data vi- sualization can offer great potential in this domain, but developing ap- propriate solutions presents considerable challenges, since the nature of the data are relatively unknown and the needs of energy data an- alysts and modelers are not yet well understood. The design brief is therefore essentially open-ended. • Sarah Goodwin, Jason Dykes, Iain Dillingham, Alexander Kachkaev, Aidan Slingsby, Jo Wood are with the giCentre, City University London. E-mail: {Sarah.Goodwin.1, J.Dykes, Iain.Dillingham.1, Alexander.Kachkaev.1, A.Slingsby, J.D.Wood}@city.ac.uk. • Sara Jones, Graham Dove and Alison Duffy are with the Centre for Creativity in Professional Practice, City University London. E-mail: {S.V.Jones, Graham.Dove.1}@city.ac.uk, Alison@perspectiv.co.uk. Manuscript received 31 March 2013; accepted 1 August 2013; posted online 13 October 2013; mailed on 4 October 2013. For information on obtaining reprints of this article, please send e-mail to: tvcg@computer.org. Participatory approaches to user-centered design, in which users and other stakeholders are involved in co-creating requirements and designs for interactive systems can lead to solutions that are more use- ful and usable [35]. We have successfully used human-centered ap- proaches in the design of visualization solutions before and have doc- umented these in detail [27]. However, the role of creativity in these approaches has as yet been only implicit. Over the last decade some fields of interactive systems development have increasingly focussed on introducing elements of deliberate creativity into participatory user- centered design processes. The aim here is to enable all participants (users, designers and other stakeholders) to contribute to the explo- ration of new fields and the generation of requirements and design ideas for novel and useful systems [1, 6, 53]. Establishing require- ments can be considered a fundamentally creative process whereby requirements analysts and stakeholders work collaboratively to gener- ate ideas for software systems [29, 30, 32]. Indeed, Robertson [42] regards requirements analysts as inventors who bring about innovative change in designs to establish advantage. Techniques for deliberately introducing creativity into the process of user-centered design can be used effectively in this context. For example, Schmid [46] used cre- ativity triggers [42] to help workshop participants invent requirements, whilst co-creation [45] and creativity workshops [24, 31] have been shown to be effective in generating novel requirements. thanks to Jo Wood, Jason Dykes, Aidan Slingsby gicentre.net