Urban Survival Kit
KITSUN, an Example of Extended Relativity applied to Urban Navigation
Prof. Maurice Benayoun1, Prof. Maurice Benayoun2 and James Edward Imi3
1 City University of Hong Kong, School of Creative Media, Hong Kong, firstname.lastname@example.org
2 Universités Paris VI-VII, Laboratoire de Physique Nucléaire et des Hautes Énergies, Paris, email@example.com,
3 City University of Hong Kong, School of Creative Media, Hong Kong, firstname.lastname@example.org
Article published with all relevant illustrations, in Architecture, City & Information Design (Proceedings of EuropIA.14: 14th International conference on the Advances in Design Sciences and Technology). Edited by: Khaldoun Zreik, 2014, ISBN 979-10-90094-18-5
In 2010, we proposed a model, based on the rules of physics, ‘for urban and information navigation’. This theory, named ‘Extended Relativity’ examined the nature of a subjective individual’s movements through an informational urban space, and suggested ways that an individual’s movements through a navigable environment could be influenced by a coalescence of both individual subjectivity and the distribution of a qualified ‘information mass’, namely the accumulation of qualitative information or data concerning the uses of spaces in a given city. In our original piece on the subject, we explored the possible similarities between particle trajectory in Einstein’s theories of space and time, and the individual’s progression through the physical and informational environment. Such an individual could be a flâneur exploring an unknown environment, or could otherwise be an individual navigating a familiar space on a quotidian basis. We theorized that despite some etymological and semantic links between conceptions of the ‘individual’ and the ‘atom’, glaring differences would inevitably arise in any relative comparison of their movement through a given space. As we explained in an earlier piece on the subject:
The semantic and etymological link between individual and atom should not lead to a purely physical vision of human behavior. On the contrary, it is by taking into consideration the autonomy of the former that we manage to differentiate between these two kinds of objects, “individual” and “atom” frequently considered to be indivisible. Individual: from Latin individuum “an indivisible thing”, neuter of individuus (“indivisible, undivided”) and “atom” from Ancient Greek ἄτομος (atomos, “indivisible”). Although “indivisible”, both, individual and particle, do not have the same level of determination and autonomy. Intentionally or not, the “individual” determines himself the majority of the interactions between himself and his close environment, whereas the particle simply undergoes these interactions. (Benayoun & Benayoun, 2011)
We can presume that an individual’s movements in any direction will result from a personal inclination, and thus will always bear a degree of unpredictability in comparison to laws of physics. However, by taking into account the impacts of personal subjectivity on an individual’s movements through a space, and thereby extending relativity to incorporate an individual’s attraction to varying factors, it can be possible to establish an abstract correlation between the rules of physics and the trajectory of the individual. For such a correlation to be feasible however, urban spaces should be reconfigured as featuring various information masses. Such masses represent the uses and symbolic values of places in the city to an individual. Therefore, ‘mass’ in this instance refers to the perceived degree of ‘attractiveness’ of the space for an individual’s tastes– the greater the density of places in an urban environment appealing to an individual’s tastes, the greater the mass. By viewing information regarding the attractiveness of parts of the city as mass, and by assuming that the individual’s direction of movement (or curvature from a direct path) is a reflection of general inclination, personal taste and susceptibility to outward stimuli, the movements of the individual can be charted in a mode that is analogous to that of an atom in space; mass (in this case, a build up of relevant information) gives structure to a sensitive space in which the individual navigates. Such masses of information will thereby affect the surrounding environment and exude a pull onto individuals; the greater the accumulation of information that is relevant to an individual’s tastes, the greater the subjective distortion of the urban context and the greater the ‘curvature’ of the individual’s path. As an individual moves through a space, the energy fields resulting from information masses determine a non-linear route, and in the process, the dynamics of the space adjust for each individual.
Figure 1: Travelling in a city, exploring a database, is moving between Points of Interest (POI).
The resulting trajectory is determined by force fields for which the intensity depends on the “interest” of the POI in relation with the visitor’s motivation.
These described parallels between general relativity and the nature of space for an individual should be treated as an analogy rather than a true scientific thesis. Yet, the impacts of such an analogy can be significant in terms of rethinking our notions of urban space and our movements. A natural corollary for us of such an analogy has always been that conventional systems for the navigation of urban centres require a paradigmatic overhaul in order to reflect the subjective urban experience. It is our hypothesis therefore, that a new method for navigation, based on the notion of ‘extended relativity’ can be produced that takes account of how masses of variously aggregated information can impact the spatial dynamics of an urban environment. Outlined here is a prototype for such a navigation system – ‘KITSUN’.
KITSUN is a non-cartographic method for exploring the informational space of the city. It has been conceived as a ‘guiding light’ for tourists and city wanderers, as a means of accurately gauging user interests in the city, and of giving directional indicators that reflect the pull of information masses. KITSUN deals with the city, not as a collection of roads, buildings and other structures, but as a diverse agglomeration of data pertaining to use, and pertaining to the possible attractiveness or usefulness of places for each individual user. Interest and motivation become the equivalent of masses and force fields weighted by distances. By dynamically integrating choices and motivations, which express the subjectivity of a user, the KITSUN tool enables the definition of susceptible directions to deflect one’s trajectory in the urban environment, just as a distribution of masses contributes to define the trajectory of a test particle.
KITSUN is designed as an application, based on Extended Relativity, for use on smartphones and Internet connected tablet computers. It is a means of navigating through a city that is viewed as a layered informational space. KITSUN has been conceived as a new method for urban navigation that understands the relationships between space, the build-up relevant information (here perceived as mass), and the subjective individual. In configuring the spatial dynamics of the city around these variables, KITSUN addresses many of the shortcomings of existing technologies outlined in the following points:
Current recommendation software and systems are map based – maps do not represent the subjective dynamics of spaces. While maps that were drawn several hundred years ago were representations of spatial subjectivity (for instance, maps of coastal defenses exaggerated the size weak spots), maps currently in common use are the product of enlightenment developments in the rationalization of space through ‘objective’ scientific measurement. While it can easily be said that maps have never truly been objective, the appearance of objectivity has always been upheld through the notion that space and form are wholly quantifiable (Harvey, 2001). However, while an ‘objective’ representation of urban space has its definite uses, we would argue that spatial understanding within the urban environment results equally from the perception of the individual. As has been suggested in previous studies on the subject, space is at once ‘morphological, perceptual, social, visual, functional and temporal’ (Koseoglu & Erinsel, p.1191). , its meaning as much a result of the (often changeable) characteristics of the user as the characteristics of the space itself. The city can be conceived as a complex mix of semiological components that remain unfathomable to users when looking at the interfaces of current navigation tools: the map, while representing the whole city, in fact lacks the representative capabilities to reflect the informational nature of space, and different spaces’ legibility to each person. Maps used today reflect a long standing wish to rationalize a living environment, a wish resulting from many obvious practical reasons that we are not attempting to understate. However, as a tool for pedestrian navigation for urban explorers, maps do not represent the true subjective nature of the city, and the means by which the city’s character can be formed through an extended dialogue between the different spaces and the user, or the measure by which space is constructed socially (Lefebvre, 1991). Current navigation and recommendation interface tools such as Aroundme and Foursquare, use the map as the sole means of communicating location. Points of interest are dots on maps. The city between these points is often a blank grid or maze of roads. In contrast, we are proposing a system and method that can measure not only the nature of space, but also the nature of space for each individual, and the possible level of attraction that a user will feel for manifold points of interest in an urban environment. In other words, we are proposing technology that can act as an ‘informational skin’ (Benayoun, 2010) – the protective intermediary between user interiority (the information unique to each user i.e. expressed taste, behavioral patterns, general inclination, or even emotional sensitivity to stimuli etc.), and the much wider exteriority (the sheer volume of information that exists outside the individual). And while in any navigation technology, a 2D map will generally have to be incorporated into the system as a measure of potential routes, and precaution against various obstacles and hazards (so that a user is not directed through buildings or across 8 lane highways etc.), the data pertaining to these maps can remain invisible to the user; the interface for navigation can move from being the distancing birds eye rationalization of city infrastructure, towards being a reflection of the experiential nature of the city for each user.
Some people cannot read maps. This includes people with types of dyslexia, such as dyscalculia, people with visuospatial issues (an inability to relate pictorial representations of space to actual space), people with Alzheimer’s disease (of which difficulty reading maps is one of the earliest symptoms), people who are partially sighted, or people who simply struggle to relate representations on to their view at street level.
With current software, generalized information is usually presented in lists or in a common manner. Sites such as Trip Advisor or Google will generally aggregate previous customer reviews and then provide an ordered lists based on generalized user feedback, usually as a star rating. A map on the side illustrates where these places are. The information does not reflect the user’s own experiences, and there is no learning algorithm that can assess a user’s temporal inclination, react to their past and current behavior, and cross-compare this with other users’ feedback. When information is presented, there can also be too much information to sieve through when on the move.
KITSUN has been devised as a method, utilizing the theory of extended relativity, as a non-cartographic ‘context aware’ tool that reflects the experiential nature of the city for the urban explorer or flâneur, as a means of reflecting the user’s projected ‘curvature’ in relation to external information masses. Insodoing, KITSUN addresses the aforementioned shortcomings of current technologies. With KITSUN, the user can select themes of interest. The application, through a learning algorithm, can further narrow down a user’s interest to gain a more specific understanding of what he or she wants from the city. Oftentimes the user will not be aware of what they are looking for, in which case KITSUN can learn this from behaviour analysis. KITSUN can establish a correlations between user input factors, subsequent user behavior factors, and the attraction fields generated by various points of interest (POIs) around the city, in the process determining a suggested personal route. KITSUN does not rely on the map as a navigation tool, and instead has its own compass navigator interface system that can dynamically represent the directional pull exerted by various urban information points. The compass navigator then, is not attracted by the Earth’s magnetic fields, but is instead a representation of a kind of toing and froing between the informational fields of assorted POIs. The user need only follow the lines of force from these informational attraction fields to explore an urban environment that is informationally configured, and therefore spatially re-understood around individual concerns.
KITSUN does not attempt to force a user to follow a particular route. It offers a suggested direction of travel based on both user selected themes and through an analysis of user behavior to form a dynamic user profile – if a user deviates from an expected direction of travel, KITSUN will incorporate this deviation, using its cache of information on the nature of places and urban environments to dynamically redraw the profile. As the user profile is altered, so too is KITSUN’s representation of informational pull, cross comparing input data from multiple users. KITSUN is a hybrid of content-based and collaborative filtered methods of recommendation that takes the mobile navigation interface beyond the map.
Glossary of Terms
Point of Interest (POI): the location of an object (or mass of information) that reflects the perceived interest of the individual user.
Theme: themes are common features or genres related to POIs i.e. ‘restaurants’, ‘contemporary art’, ‘antiques’, ‘national heritage’, ‘literature’, ‘architecture’ etc.
In the outlined application of KITSUN, themes are presented on a selection board for user selection.
Tag: A sub category of the themes i.e. if the POI theme is restaurant, then possible tags could be ‘gourmet’, ‘romantic’, ‘trendy’ ‘calm’. The tag is not a priori positive or negative, it is the polarity of the service user that determines its characterization. “Trendy” might be positive for a given individual and a negative tag for another. A series of tags are assigned to a theme. In practice, they are usually eight.
Interest Curve: The set of properties or features of a POI, given a weight in proportion to its importance, constitutes the interest curve of the POI. This curve can evolve in real time according to the tagging declared by the service users.
Motivation curve: Linked to the service user. It encompasses all the tags, from each theme, with their respective weights. It constitutes the service user’s profile. It varies according to time and location. It is also redefined by the successive actions of the service user.
Selection board: the selection board is an interface that allows users to select themes, by the spheres, called satellites, into the ‘my selection’ area of the display. These satellites will then be used with the navigator compass. The selection board enables the regrouping of themes, and to upgrade your themes according to the time of day.
Figures 2-4: Selection board: Each ball (satellite) corresponds to a theme.
The user can drag and drop them into “my selection” area to select the active themes.
The number of possible themes is not limited.
The KITSUN user walks through the city. On some days this user will have a clear travel destination and be under time constraints. In such an instance, the user won’t generally be as open to outward stimuli. However, when the user has time to play with, he or she will be far more receptive to the nature of the urban landscape. In this case, the user will select up to four themes from the selection board. Following the selection, a central navigation compass sphere is displayed surrounded by the selected theme satellites. The satellites around this central planet sphere give an indication of where POIs are located in the vicinity. By following the visual indicator of the direction and intensity of light hitting the central planet sphere, the user can choose to navigate in and around a suggested direction of travel in relation to the displayed theme satellites.
The satellite orientation interface is deigned to give direction without being proscriptive. It is defined according to the position of the user (the planet). The build up of mass from the POIs is what lights the central planet sphere, so that the intensity and direction of the light can emphasize not only the direction in which the greatest element of attraction lies, but also the accumulation of information from various POIs in this given direction. In a more vertical city, the light can give an indication of height, so that light reflecting from the very centre (or top) of the planet sphere can indicate that the user should head upwards. This represents a sea change in the way we can explore and navigate the ‘3-dimensional city’ i.e. the city in which points of interest are located on the higher floors of buildings, or in which the inner-city shopping mall located over a transport hub can stretch to over 15 storeys, or in which there exists a matrix of underpasses and interconnected raised walkways.
Figure 5: Locating a light source (POI) by its impact on a sphere (planet)
These 3 dimensional cities (increasingly prevalent in east Asia) can be near impossible to navigate with a 2 dimensional map. KITSUN addresses this shortcoming, and while general positioning technology is not able to determine the how high a user is at present, indications are that this is the direction in which technology is headed. By judging the positions of the theme satellites and relation to the direction and intensity of light, the user can assess the nature of the surrounding environment, and gain a solid idea of where is open for exploration, be that at street level or anywhere else.
Figure 6: The Navigator Screen. The planet (central sphere) is surrounded by user-selected thematic satellites. Their orientation gives the heading. Touching one of them gives the priority to the related selected theme and the planet matches its color. The light on the planet gives the direction and the property (tagging) of the POI.
Colour to qualify the destination
One application of the KITSUN system allows the user to set colours for different themes and sub-group tags. For instance, if looking for an art museum, the user could follow a blue light in the knowledge that, having set the colour choices personally, this is a contemporary art museum.
Figure 7: The Color Selector: The user turns the color wheel to associate colors to tags (“romantic”).
The impact of light on the planet is colored as selected.
The user’s motivation curve
The KITSUN system allows users to ‘tag’ various places. The tags, as subcategories of the theme of the POI, are generalized user feedback and give a greater indication of the specific nature of an environment. The navigation user’s motivation curve is created according to the correlation between the tagging given to a POI (THEME: restaurant; user tagging: calm) and the time spent in that location (interval between “Tdin” time–distance<Xm and “Tdout” time-distance>Xm), in which “X” is an adjustable parameter in meters (X=10 by default). For instance, if a user decides to stay in an environment tagged as calm for 2 hours, when can regard this as being a positive attribute. Such information is used for the adjustment of the recommendation information via a weighting coefficient.
The weighting coefficient
The “weight coefficient” determines the importance of the most recent messure of user motivation on the overall result. The higher its value, the more the motivation curve reflects the current state of the user; the lower its value, the more it reflects the average profile from the total duration of the observation. This value is expressed as a percentage and is set at 30% by default.
For example: before entering a new restaurant rated as “calm = 3.7” (q1), the service can understand that, by accepting to enter a place with a positive “calm” factor, “calm” can be considered as a positive value for the user. This interpretation is confirmed if the user rates the restaurant as calm = 5 when he leaves it after staying for 2 hours. If he or she would have rated it “calm=5” after 5 minutes in the restaurant; the service would acknowledge that “calm” is not a positive tagging for this user that seems so impatient to leave the place.
The wear of motivation
Within KITSUN, there is a system for deleting already visited POIs, or areas in which the user is simply not interested. This is because a tourist generally would not want to make repeat visits to the same places if in a city for less than a week. The wear coefficient will delete already visited POIs from the system, or POIs where the user has simply swiped ‘next’. What this reflects is how the parameters of the city change for the user in relation to the amount time spent there, with the user building up information on themes of interest, but diversifying towards different relevant parts of the city the longer they use the app. Such a system can avoid ‘black holes’ in which the amount of information mass in one area of the city is so relevant, and proven to be so popular for an individual that it becomes impossible to escape the pull.
The interest curve of the place (POI)
The principles governing the evolution of the motivation curve apply, largely, to each Theme and to the set of tags of the theme. Thus, they apply to the interest curve of the locations.
The interest curve of the place is composed of all the tags.
Their values determine a profile that the system will bring closer to the user’s profile and the greater the accuracy, the greater the attraction force (as with mass in the gravity algorithm).
Explicit or declarative tagging
The navigation user leaving a location tags it declaratively by using the tagging wheel. He gives the location he has just visited a tag and a color.
The tagging of a place on the KITSUN database is the dynamic synthesis of individual tagging, constantly updated by the feedback from the users.
Figure 8: The Tagging Wheel : The user can « qualify » a POI. Turning the “steel wheel” displays
the appropriate tag for the active theme. Then clicking on the planet sends the tagging to the server.
Other parameters involved
The device allows for the integration of other tags that are not modulated by the users. For example, the “newness” for the interest curve, and “already visited” for the motivation curve. In the first case, this value is not affected by the tagging of the user. This value decreases at a rhythm that can easily be customized.
pVALn = newness coefficient = 5 / (number of weeks)
uVALn = wear coefficient = 1 x (number of ‘Next’ on the POI)
The temporal update depends on the schedules of access to the locations: for restaurants it is the time for placing the last orders, for a museum the time for the last entry.
free not free – = + (0-5)
The user is not asked for his pricing tolerance. It is deduced from his choices for each theme. It is well known that we can accept high pricing for an activity that is dear to us and save money on another that is vital.
Sensitive tagging of location
Though the user curve is a result of his individual, declarative and spontaneous choices, the interest curve results from collective choices. It is the average of the users’ feedback possibly weighted by the average of users with similar profiles. We can image that there are virtually two interest curves for each location: a standard curve (sQc) that can apply to any novice user and a custom curve (pQc) resulting from the tagging of users with a curve similar to that of the user.
Figure 9: The Navigator Screen
It is the proximity between the motivation curve of the navigator user and interest curve of the nearest location that constitutes the basis of KITSUN navigation. It is what we will call “sympathy”.
Sympathy becomes a sensible oscillation between fluctuating force fields depending on the dynamic nature of data, but also on the relative distance separating the navigator user’s POIs. As with Newton’s law, attraction relates to the respective masses and to distance.
Relative mass POI/user
In the implementation of the gravity model, the “mass” of each element should be considered. It should be compared, for each entity, to the value by theme (a location that may be concerned by several themes) and by tag in the theme.
Direction and trajectory
What the navigator translates is less the forces than the direction that results from the POI’s attraction potential. Through his choices, trajectory, stops, avoidance, the user modifies his undeclared demand and the navigator redefines the heading. The heading may, at a distance, correspond to a group of POIs (a neighborhood), upon approaching it, it can point to a street and even closer to a street address or a specific destination. The heading is deduced from all the forces acting in his position. The forces are calculated according to their adequacy between user and POI. A high-value tag that matches the user corresponds to a high mass. Other parameters may be considered (e.g. the number of declared tagging by the users) probably related to the percentage of visitors to the location. Mass is time-dependent and a location loses all its mass after closing time.
The basic principles of General Relativity provide a framework and a language that appropriately describes the specificity of urban informational navigation (the city conceived as a data field). Interest and motivation become the equivalent of masses and force fields weighted by distances. By dynamically integrating choices and motivations, which express the subjectivity of a user, the KITSUN tool enables the definition of susceptible directions to deflect one’s trajectory in the urban environment, just as a distribution of masses contributes to define the trajectory of a test particle. It therefore responds to the expectation of an essentially serendipitous tool: creating the conditions of a discovery by automating the formulation of the query. If urban browsing becomes a distinct discipline, KITSUN should bring a natural response, based on taking into account the informational curvature of urban space-time modulated by the subjectivity of the individual experience. As there is only one adequate trajectory in the space curved by subjectivity, there is only one answer to a non-explicit query – a question that has not yet been formulated with words but through individual’s behavior- because it will be felt as such: the right answer.
Adomavicius, T. (2005) Toward the Next Generation of Recommender System, IEEE Transactions on Knowledge and Data Engineering, 17(6).
Anonymous (26th March 2009) Henry VIII’s Coastal Defence Maps < http://www.bl.uk/onlinegallery/onlineex/henryviii/militmap/coastdef/> [accessed 30th June 2013].
Anonymous (no date) Dyscalculia, Dyslexia and Maths. British Dyslexia Association website < http://www.bdadyslexia.org.uk/about-dyslexia/schools-colleges-and-universities/dyscalculia.html> [accessed 30th June 2014].
Benayoun, M. (2010). The Nervous Breakdown of the Global Body : an organic model of the connected world. In Proceedings of Future En Seine 2009 (pp.133-145) Paris : Cap Digital.
Benayoun, M. & Benayoun M. (2011). Extended Relativity: A model from physics for navigation through urban and data information, DRHA Proceedings, University of Nottingham, Ningbo, China <http://www.nottingham.ac.uk/drha/programme/speakers.aspx> [accessed 30th June 2014].
Downs, R. M. & Stea, D. (1973). Cognitive Maps and Spatial Behavior : process and products. In Downs, R. M. & Stea, D. (eds), Image and Environment: cognitive mapping and spatial behavior. Chicago : Aldine Publishing.
Harvey, D. (2001) Spaces of Capital : Towards a critical Geography. Edinburgh : Edinburgh University Press.
Harvey, D. (2009) Cosmopolitanism and the Geographies of Freedom. New York : Columbia University Press
Hillier, B. & Hanson, J. (1984) The Social Logic of Space. Cambridge : Cambridge University Press.
Jacobi, J. A., Benson, E. A. & Lindon, G. D. (published 26th September 2006) US Patent 7113917 B2 Personalized Recommendations of items represented within a database, assigned by Amazon.com Inc. <http://www.google.com/patents/US7113917#v=onepage&q=&f=false> [accessed 30th June 2014].
Kallus, R. (2010) From Abstract to Concrete: Subjective reading of urban space. In Journal of Urban Design 6(2), pp. 129-150.
Koseoglu E. & Erinsel Onder D. (2011). Subjective and Objective Dimensions of Spatial Legibility, Procedia – Social and Behavioral Sciences, 2011 (30), pp. 1191-1195.
Krier, R. (1979) Urban Space. New York : Rizzoli International Publications.
Lefebvre, H. (1991) The Production of Space. London : Blackwell.
Stanek, L. (2011) Henri Lefebvre on space : architecture urban research, and the production of theory. Minneapolis : University of Minnesota Press.
  Benayoun, M. and Benayoun, M., Extended Relativity. A model from physics for navigation through urban and data navigation, DRHA Proceedings, University of Nottingham Ningbo, 2011
 KITSUN (KIT de Survie Urbain Numérique, Digital Urban Survival Kit). The prototype has been developed in 2010-2011, in the frame of the French call Proxima Mobile, launched by the French Ministry of Research. Due to the defection of one of the partners, in charge of the POI database, the operational prototype didn’t reach the industrial stage.