Periodic Reporting for period 2 - LostInZoom (Not Getting Lost in Multi-Scale Maps: Exploring the Anchor Theory in Cartographic Zoom Interactions in the Context of Crisis Management)
Reporting period: 2022-10-01 to 2024-03-31
Multi-scale interactive maps such as Google Maps, Bing Maps or OpenStreetMap have replaced paper topographic maps for most professional and daily uses. People use them regularly on their smartphones, on tablets, or on computer screens. Past research told us how to design paper topographic maps at a given scale to make them readable and understandable by human users. But these rules or guidelines do not apply anymore : it is not necessary to put as much information as possible in the map with such applications, because you can always zoom in to see more details or zoom out to get a broader view of the area. As a consequence, map designers lack guidelines to make pan-scalar maps (i.e. multi-scale zoomable interactive maps) that are smooth to explore through scales, and it is common for a pan-scalar map user to feel lost for a few seconds after a zooming interaction. The user often needs to zoom in and out, back and forth, to localise themselves. Even though these map applications are now ubiquitous, their use is far from optimal, and both the general public and the professional users of topographic maps could benefit from maps that are designed based on principles that encompass their pan-scalar nature. The LostInZoom project (Not Getting Lost in Multi-Scale Maps: Exploring the Anchor Theory in Cartographic Zoom Interactions in the Context of Crisis Management) seeks to establish a new zooming paradigm for multi-scale maps, based on multi-scale visual landmarks that act as anchors during the zoom, which could solve the disorientation problems. To achieve this novel zooming paradigm, we need new grounding knowledge on how people perceive and understand pan-scalar maps, to design maps and interactions that make multi-scale explorations smoother. The LostInZoom project will be based on three main pillars. In the first one, we will explore the cognition of pan-scalar maps with an experimentatal approach, to identify the anchors or landmarks that help the map readers locate themselves when zooming. In the second pillar, we will design new cartographic techniques to derive, as automatically as possible, pan-scalar maps that magnify the landmarks that are important for multi-scale exploration. Finally, the third pillar will be dedicated to the design of new zooming interactions that focus on these landmarks to smooth even more the zooming exploration: to put it simple, in this last pilar, we want to change how the map reacts when the user pinches the map, or uses the mouse wheel to zoom.
The first achievement of the LostInZoom is the definition of our research subject, i.e. the interactive multi-scale maps now used by the general public and by professional users on their phones or computers (e.g. Google Maps, Bing Maps, OpenStreetMap, or authoritative counterparts, such as the Plan IGN in France). We called these maps “pan-scalar maps”, because the integrated map holds more information for the user than the sum of what can be seen at each scale independently. For instance, in a zoomed view inside Paris, you do not need to display the name Paris but the user knows as they arrived here by zooming in from a view that displayed the name, or by entering an address located in Paris. Then, we collected user experiences on cartographic disorientation, i.e. disorientation periods while using a pan-scalar map, to better characterise this phenomenon. We also conducted experiments that showed that disorientation was more frequent when map users were zooming in strongly. By contrast, zooming out, even strongly, does not generate much disorientation, as it is an action that brings more context to a self-localisation cognitive process. We also conducted experiments with users drawing on maps, to capture what they consider as the main landmark in a topographic map: for instance, the Seine River is considered as a major landmark or anchor in maps around Paris, France. We coupled these experiments with eye-tracking measures that showed that people use these anchors with frequent gazes to localise themselves during a zoom. We also started the design of machine learning systems able to automatically recognise these anchors in any image of a map. Regarding the design of novel maps, more adapted to the interactive use, i.e. the second pillar of the project, we developed an open Python library for cartographic generalisation, called CartAGen. Cartographic generalisation is a complex process of abstraction and simplification of the geographic data to obtain legible small-scale maps. For instance, cartographic generalisation chooses the best roads and rivers to show at small scales. Finally, regarding the professional use of pan-scalar maps, we conducted a joint design process with firemen in charge of post-earthquake crisis management, and proposed a prototype of pan-scalar map to use in a control room during crisis management.
Arrived at the midterm of the project, we expect more advances in the coming years, in particular in the last two pillars of the project. First, we studied anchors and disorientation in independent user studies, and we now plan to conduct studies to better understand the role of anchors in disorientation and self-localisation problems. In particular, we are interested in the role of anchors to favour the memorisation of mental representations of the map. In the second pillar focusing in novel pan-scalar designs for the maps, we are working on the concept of progressivity, as a way to make the transition between two scales more fluid and immersive. We take inspiration in the theories on cinema, music and comic books, where smooth transitions are important to keep the audience immersed in the artist work. We also plan to experimentally demonstrate the importance of cartographic generalisation for the users of the pan-scalar maps, in terms of reduction of disorientation, cognitive load, or just by making the map more practical to use (less zooms are required to solve a task). Finally, in the last pillar, we plan to propose new zooming interactions, which means that the applications will not react similarly as today when the user pinches the screen, or rolls the mouse wheel. These novel zooming interactions will be based on two principles: (1) a slower zoom to let time for the brain to perceive and process the change of scale; (2) a zoom assisted by prior knowledge or predictions of what the user is searching. In addition to these two principles, the role of will be central, as a slow zoom will let time for the user to look at the anchors to self-localise, while the assisted-zoom will use the anchors to make sure there are always enough anchors visible in the screen during the zoom. These novel zooming interactions will be assessed by users performing several usual tasks with maps, to verify that they really improve the usability, and reduce the cartographic disorientation.