The meeting to report on my first results on tourist activities in Florence provided the opportunity to further plan my year at the MIT SENSEable City Lab. Prior to moving here, I extracted the keywords of my research: feedback loop, manual location disclosure, digital traces, granularity, uncertainty and co-evolution. Instead of finding complete coherence, its seems that now the completion of my thesis could take two separate avenues each related to some of these keywords:

Leveraging digital traces
In the first, I can build further upon the Tracing the Visitor’s Eye project and consider the analysis of digital traces or volunteer generated information to understand how they can be helpful to tourism (or more in general mobility?) and support decision making. It could be about forging new ways to describe tourism with a validation through second order analysis with other dynamic data such as cellphone data (flickr 70% and 30% cellphone data). Analysis could take place in Florence or Rome (better for statistical validity). Part of the analysis would focus on the accuracy of the data at hand and highlight the shortcomings and potentials. It would be about how flickr users (and maybe another dataset) describe the space (semantic analysis of the flickr dataset). The outcome would be a set of interactive tools and visualization to analyze the data and why not a model that could simulate the mobility of tourist from the flickr and cellphone datasets.

Research questions: How digital traces (or in a narrower way “volunteer generated information”) can enhance current tourism (or in a more extended way mobility) observations? With potential sub-questions as follow:

• What new information on mobility and tourism do these data bring? -> traces, scalability, richness of the explicit act of disclosing information, peope-defined area of influence of points of interests, people’s area of attention (digital footprints to improve the virtual representation of the space), geographic relevance
• How can we validate these data? -> use techniques to calibrate the flickr dataset with other mobility databases.
• What are the data quality (accuracy, noise, …) issues in volunteer generated information? This would be about revealing some factors that influence people’s decisions when they georeference information. ->In addition to Flickr data, I could setup a field experiment in Florence or as part of the WikiCity Rome project.
• How does automatic positioning influences location disclosure? Retrieve users who georeference automatically and study the semantic descriptions they use to disclose the information.
• How to visualize uncertain location information? This might involve setting-up an experiment with practitioners in urbanism/tourism or observe their current practices.

The appropriation of location information
The second avenue aims at building a coherence (a story) from the outcomes of CatchBob! and my taxi driver study and the semantic analysis of the flickr dataset. The main theme/question would be to better understand how do people relate to space (and its multiple spaces) through location information with a set of evidences each study would bring. CatchBob! indicated that technologies representation of the physical environment is uneven and fluctuant leading to feelings of uncertainty. Observations of taxi drivers revealed the importance of the prior experience of the space to appropriate a satnav system and the pitfalls of the discrepancies revealed in CatchBob!. In addition, current satnav systems do not fully support the practices of taxi drivers who need to access different levels of granularity of location information during a journey (trunked access to the information as if it was process through a funnel). This is for the reading/accessing part of location information. So what happens when we let people write and describe space. How does that translate to the different levels of granularity of multiple spaces (spatial semantics)? The semantic analysis of the flickr dataset could help understand how people manage multiple space. I could add a field study in Florence to bring another perspective to that question. The outcome of this research avenue could consist in a list of evidences revealing the issues around the granularity of information, a tool to study people-generated content. The sub-questions could be:

• What factors influence uncertainty in the use of a location-aware application? How is that related to the management of granularity and the reference to multiple spaces?
• What are the influence of automatic positioning on the the practice of manual location disclosure?
• How can people-generated content help define multiple spaces and different levels of granularity?
• …

Relation to my thesis: Avenues to discuss with my advisor, then take a decision, stick to it, trim and polish the research plan. I would love to integrate some of the velib and bicing data analysis to any research avenue, but it seems that for the moment it will stay as a parallel (fun) research endeavor.

Two projects mapping the predicted real-time positions of trains. Since train operators do not disclose the actual location of a train, these services must use indirect ways to collect these data. Where Are Trains (France) parses online schedule boards of different stations such as the one of Paris Montparnasse. The position of the train obtained in real-time upon Arrivals and with at least a 1-h delay for departures. Then the system uses pre-builded time profiles to estimate the current location of trains by-passing potential stopovers. Similarly Train Map (Switzerland) uses train timetable, and does not yet show the actual GPS-positions of the trains. “But, as Swiss trains are almost always on time, most of the time the position is accurate”.

Tracking a train in Zürich with Train Map.
Relation to my thesis: The predictive approach used in this systems generates uncertain data that determine a location in space (where is it) and time (when was it there). Does the temporal granularity of these location information affects the decision making? In what kind of scenario would people rely on these informations

Mapping real-time train traffic follows the trend of vessel tracking (e.g. planes and boats)

Last Wednesday, I gave a 20min brownbag talk at the giCentre at the City University in London. The presentation was divided in several parts. First, I define the shortcomings in location-aware computing and their consequences that generate a socio-technical gap. I continued by highlighted that the problems do not necessarily lay in the immaturity of technologies, but also in the failure to match people’s own perception of space (granularity, multiple-spaces). Then I detailed the evidences of this gap from my studies and observations of the appropriation of location-aware applications (CatchBob! and Satnav in Taxi). That lead me to describe an approach that leverages digital traces to tailor location information and define user’s the area of attention and their perception of area of influence of points of interest. In that context, I described the Tracing the visitor’s eye project and briefly introduced the context of future experiments (WikiCity and the Wireless City). The slides are online (5.5MB).

The presentation generated lively exchanges with Jonathan Raper, Jason Dykes, Aidan Slingsby, David Mountain, Jo Wood that benefited me to frame of my thesis. Besides arguing on the potential of volunteer generated information (VGI), the discussion centered on the influence of the presentation of location information on the behavior of people (the difference in the communication in CatchBob! (passivity), multiplicity of the sources of information and location-information trunking for taxi drivers) and these behaviors influence the data (feedback loop in WikiCity, geotagging in Flickr). I was in fact advised to focus on how the co-evoluation between location-aware systems and their user’s practices/behaviors (data influencing the behaviors influencing the data).

Relation to my thesis: This week’s trip in the UK is about testing my ideas and approaches with a verity of experts from different fields (I got the pleasure to meet UCL’s Jon Reades to discuss urban planning and urban computing). I must admit that it is a truly rewarding experience to pick the brains of geographers, geovisualization experts and social scientists and have them criticize my work. Presenting and arguing on the current state of my research work should help me create a “meme” and that everybody starts to believe my “story”. Many people have now advised me to get back to my different experiments and (re)define what there is to study for each of them. Categories or research thems and specific question should help me focus on 1 specific aspect and help me find the gaps in “my story” (e.g. thesis).

A couple of paper that discuss the emergence of the large data generated by sensor now sharing our lives:

First, Data management in the worldwide sensor web draws the big picture in mentioning that now too much attention has been placed on the networking issues of distributed sensing and too little on tools to manage, analyze and understand the data. The authors ask the question weather we can design sensor networks with data quality in mind? They ask a very crucial question, but as often in location-aware computing, it is very unclear on who can claim what quality in location information is or in other words who can answer “how good is good enough?”. Of course it is important to manage temporal and spatial data and handle their inherent uncertainty (e.g. via probabilistic theory) or mask it (e.g. via interpolation) or play with it (seamful design). It seems clear now that my thesis is about acknowledging that situation (uncertainty in the location information, fluctuant quality in the data), but instead of aiming to produce “perfect data”, I plan to provide an understanding and solutions from a human and urban perspective. It comes, at the first place, with the observation of people experiencing location-aware systems in CatchBob!, and making use of location information, in my taxi driver (co-evolution, context and granularity). This observations help me accumulating evidences on the contextual factors influencing the granularity (≈human expectation of quality) of the location information used.
Balazinska, M., Deshpande, A., Franklin, M. J., Gibbons, P. B., Gray, J., Hansen, M., Liebhold, M., Nath, S., Szalay, A., and Tao, V. (2007). Data management in the worldwide sensor web. IEEE Pervasive Computing, 6(2):30–40.

Second, Citizens as Voluntary Sensors: Spatial Data Infrastructure in the World of Web 2.0 discusses that the most powerful sensor web is made of the 6 billion humans occupying Earth’s surface. This large collection of mobile and intelligent sensors will affect the processes by which geographic information acquisition and compilation (VGI: volunteered geographic information). The data generated suffer similar issues as a top down (authoritarian, centrist paradigm) when it comes to the fluctuating quality in the data and trust. However, the notion that citizens with means of taking measurements is at the source of the solution to the problems mentioned above. The analysis of how these “citizens” handle and annotate their measurements and observations allow to further understand the influencing factors in the use of location granularity. This is why I study Flickr users in their spatial annotation practice and in their use of geographic semantics.
Goodchild, M. F. (2007). Citizens as voluntary sensors: Spatial data infrastructure in the world of web 2.0. International Journal of Spatial Data Infrastructures Research, 2:24–32.

Third, the digital traces (shared measurements/observations) left by people in space allow to define a human description of space (e.g. citizen definition of a neighborhood). This type of sensor web data can only make sense with geovisualization, as the ones presented in Interactive Visual Exploration of a Large Spatio-temporal Dataset: Reflections on a Geovisualization Mashup. The authors explore the new opportunities for visualizing sensor web data to explain user behaviors. Tools such as Google Earth provide a quick support for visual synthesis and preliminary investigation of digital traces.
Wood, J., Dykes, J., Slingsby, A., and Clarke, K. (2007). Interactive visual exploration of a large spatio-temporal dataset: Reflections on a geovisualization mashup. IEEE Transactions on Visualization and Computer Graphics, 13(6):1176–1183.

Relation to my thesis: Each of these three paper give an overview of the main themes of my thesis that aims to take a human and urban perspective to define the quality of location information:
1. Issues in the quality in the location data (uncertainty).
2. New data to observe people handle/experience location granularity in order to collecting evidences.
3. New visualization to reveal how people perceive and describe the urban space

Late last year, a position paper I submitted to the Urban Mixed Reality workshop at CHI’08 was only accepted as a poster presentation. In this paper, we propose that digital footprints present an opportunity to the residents and tourists the ability to look down on the city and view the activities and their consequences. When this information is fed back to the community, it can inform the decision-making and change the behaviors through social navigation. However, the design of a social navigation service should take into consideration the lack of accuracy in space and time of digital footprints. Apparently, the unique reviewer who rejected the paper understood that we propose the use of digital footprints to support people in navigating (orientation/path/aims) in the city.

So I thought that instead of flying overseas to present a poster, it might be more relevant to leave this position paper online with its reviews, open for discussion and thoughts to the reader of this blog. Considering the encouraging comments of the second reviewer, I will most probably recycle it for future publications.

Leveraging urban digital footprints with social navigation and seamful design
Girardin, F., Nova, N., Dal Fiore, F., Ratti, C., Blat, J.

Abstract. The widespread deployment of mobile and wireless technologies increases the amount of recorded interactions between humans and the urban environment. The accumulation of these digital footprints provides new opportunities to reveal human behaviors in space. Beyond their utility to improve the quantity and quality of mobility data already available to urban planners and local authorities, this information can be returned to residents and visitors to enhance their perception of the space and inform their discussions and decision making. In this paper, we argue that digital footprints, when properly revealed, can act as social navigation cues to support the exploration of the city.

[Full paper - 104KB]

Reviewer 1 (reject):
The paper discussed ideas on how to make patterns of mobility and flow based on digital footprints available to tourists and residents. The discussion is based on two concepts – ‘social navigation’ and ‘seamful design’. I miss a reflection on previous studies on supporting people’s orientations/paths/aims when moving in a city through giving them visual information. This is a quite complex endeavor and it is not sufficient to provide a ‘vision’ without thorough grounding. For example, I don’t underdstand the usefulness of ‘seams’ (uncertainty of data, lack of timeliness, etc.) for people. You may want to read Bill Gaver’s paper on ‘Ambiguity as a resource for design’.

Reviewer 2 (accept):
The paper argues for the use of digital footprints as social navigation cues for the exploration of the city. Digital footprints are space and time referenced data that are produced by the increased amount of recorded interactions between humans and the urban environment. The paper presents an approach meant to leverage this kind of mobility data to support awareness of the overall dynamics of an urban space and affect the discussion and decision-making of residents and visitors in that space. Challenges inherent to the rendering of spatio-temporal data in mobile and urban environments are addressed by adopting a “seamful design” approach revealing the imperfection of the sensed data.

The contribution of the paper to an overall framework for the social use of mobility data is timely and likely to raise discussion. Suggestions for improvement: the authors may want to better explain the idea of “cultural views of mobility” and provide more examples on the kind of data that could be used and how their visualization would inform people’s behaviour in the urban space.

I have been on the field collecting more observations and interviews on the adoption of GPS by taxi drivers in Barcelona. On one hand I am interested in the co-evolution process and the modifications of the practices, on the other hand I focus on the context of use of the technology (other artifacts/tools, influence of landmarks, type of location information used (how, where, when, granularity). Here are a few notes that will need to be interpreted further. I collected data from a unbalance population of either find very experienced (15+ years: seniors) taxi drivers or novice in the practice (2- years: juniors). It has been harder to anybody in between. At least, it revealed dissimulates in the appropriation of the technology between the “seniors” and “juniors”.

Uniformity in the wild
First of all, there is a concenus that GPS an unbeatable tool when it come to reaching a specific destination in a village and leaving it (link to previous post). It is a strong reason to acquire a sat-nav system. But it is not really to be more productive and hence earn more, rather than feeling more serene (i.e. tranquility was the word). Some mention that it does not only tranquilize them but also their clients (”they know I cannot cheat them with it”, “it reassures them that I go to the proper destination”). When it comes to seniors, they now know that they have a “companion” (some refereed to their car with an affectionate “she” and their sat-nav with a “he”) can take them anywhere and avoid these moment when they had to search endlessly browsing the street directories of the suburbs of Barcelona, asking their colleagues via radio or mobile phone. One mentioned me that “the fear of getting lost with a client I felt in my stomach now disappeared”. Another was relieved that he did not have to use his radio to ask colleagues about particular location informations and wayfinding (even though many taxi drivers are extreme social animals). Currently the radio is uniquely used to get dispatching orders and professional questions (e.g. asking on special rates for some destinations). To get out of the unfamiliar environment, the driver would be either very cautious in the path taken to enter the area (”I am a very good observer”) or use their GPS with destination Barcelona. In their exit strategy, their reach their bookmark of saved destinations are hit “home”, “Barcelona, any street”, or “Barcelona center”.

Disparities in the urban space
However, when it comes to the context of the city, practices differ and the “Guia” comes into play. The “Guia” is a dense book of all the streets and POI in Barcelona. Experienced drivers have been using it all their professional lives and have partially learned the city with it. Another big part of the knowledge drivers acquire come from the clients themselves. They know all the tricks and secrets official books and commercial systems can’t tell. As the experience drivers contain most of the knowledge of the city in memory they rarely return to the Guia. In fact, they prefer using their sat-nav and hit the Guia only as fail-over (e.g. when the system does not contain a street number or lacks of information). Even after 20 years of experience, one drivers still perceived accessing the Guia as a demanding task. Compare to that, hitting the screen of his Tom-Tom was a pure pleasure. This process of accessing correct information is executed differently by the juniors I interviewed. First, they had mixed feelings about their experience with their GPS (”It is like my mobile phone, sometimes it does not work well”, “it is a potential problem”) in terms of quality of services and precision of the information. They very early reached the limits of the systems and the inaccuracy (erroneous information) of the system directly impacted the quality of their service (mostly because they had to deeply rely on it in the beginning). In consequence, they embraced the use of the “Guia” which is “more accurate and complete” (particularly when it comes to POI). In the city, they learned not to completely rely on their sat-nav (because it is hard for them to judge if the information is erroneous or not) and first open the “Guia” prior to enter an address. One told me that he learned to interpret the system and its errors. When a conflict emerges between his sat-nav and his “knowledge” he starts to “improvise” (in his own terms). Depending on the circumstances he would either switch it off or ignore it for a while.

The confirmation of the funnel
Confirming what I mentioned previously, the wayfinding practice takes place in 2 main stages. There is a separation difference between “ir” (to go) and “llegar” (to arrive) and the location information are different in each phase. First the driver has the knowledge of the area where to go or checks in the Guia to get a rough awareness of destination (one driver told me he would directly enter the address in his navigation system). This is done at any occasion at a traffic light or in dense traffic (if the driver knows he will have to go on the highway directly, he/she will check prior to leaving). It is only at the entrance of the area (e.g. a neighborood - it would be interesting to define it better) that the active mode of the sat-nav system starts by entering the full address (once again taking advantages of traffic lights and busy traffic).
Mobile phone
The mobile phone often available is seen as a social link to family and friends, and is very rarely used as a tool to support their work. One driver told me that he used it a few times to check if a table was available at a restaurant for clients (particularly foreigners). This is a service he gives as added value, but he “normally should not do it”.

Passive mode
As mentioned in a paper by Nick Forbes,there is a clear tendency to keep the system in passive mode at any time. It is done so mainly to keep track of the radars (seems to be more a priority for experienced drivers) and (for juniors) to keep an awareness of where they are (keep track of the proximate streets) and it was also very valuable to learn the city. On this point, an experienced driver explained me that younger drivers who use a navigation system do not gain knowledge of the city, because they follow the recommendations and “stop to think”. My contact with the younger generation proved somehow the opposite. They are eager to learn the city and the imperfection of their systems forced them to get knowledge from the “Guia” and from the clients.

Relation to my thesis: Working on The co-evolution of taxi drivers and their in-car navigation systems, I stereotyped a bit the practices based on my notes to reveal the main patterns. I keep 3 main tracks that are worth investigate more in details:
- different appropriation of the location-aware system depending on the overall knowledge and experience of the space. The shortcoming of GPS forced the younger generation to rely on the Guia, while the older generation has little use of the Guia now. This goes in the direction of understanding and supporting “mobilities” instead of mobility.
- a tool not to make more money, nor to improve the efficacy, but to tranquilize (”I can go everywhere”) and relax (e.g. passive mode to have a continuous awareness)
- funnel-type of access to location information (I will model the process, this is part of the understanding of granularity) with multiple sources (kwnoledge, sat-nav and Guia)

After 2 years of doctoral school and the defense of my DEA thesis (pdf), I guess I now hold a Master of Philosophy in Computer Science and Digital Communication. The slides of the defense are available here.

Relation to my thesis: The defense ended up being a bureaucratic formality. One feeling I have after having presented the same slides a few times this year is that I now drag and try to combine too many concepts together. To move forward, I will need to make choices and trim my scope and get more depth in the details.

In April of next year, I will attend in Boston the AAG meeting and participate to a session on “Situating Sat Nav: Questioning the TomTom Effect“. Organized by Chris Perkins and Martin Dodge, it aims to address the social effects, cultural meanings and political economy of in-car satellite navigation. I will be in the middle of a spectacular line-up:

Session One

Amy Propen: The Use of Sat Nav Systems: An Empowering Cultural Practice or Portentous of a Lost Geographical Imagination?
Don Cooke: The TomTom Effect: Industry Point of View
Allan Brimicombe and Chao Li: Sat Nav: Rising theft of a geo-engineered must-have.
Tristan Thielmann: Navigation becomes travel scouting: The augmented space of car navigation systems
Caren Kaplan: Precision Targets: Consumer Subjects, Militarization, and the Politics of Location.

Session Two

Fabien Girardin: The co-evolution of taxi drivers and their in-car navigation systems
Georg Gartner: Restrictions in mental representations of the world as a result of relying upon navigation systems
Jonathan Raper: The mistakes that satnavs make (and what they don’t know)
Alexander Klippel: Can we afford to provide cognitively inadequate wayfinding assistance?
Discussant: David M Mark

The abstract of my paper entitled “The co-evolution of taxi drivers and their in-car navigation systems” goes as follow:

In the recent years, the massive use of in-car navigation systems has symbolized the emergence of location-based services for wayfinding. This market success creates the opportunity to learn from a real-world use of present location-aware systems in order to inform the design of future applications. In that context, we are using an ethnomethodological approach to study the different ways taxi drivers rely on their navigation system. First, this work focuses on describing how location technologies impacted the wayfinding practices and similarly how the practice influences the appropriation of navigation systems. This co-evolution starts from the acquisition and setup of a navigation system to mastering the system shortcomings and limitations. Second, we study the criteria that steer a driver in selecting among the different modes of a navigation system and the other artifacts and tools (e.g. maps, street directories, landmarks) he or she uses for location awareness and wayfinding. Moreover, we are analyzing the role of context in this dynamic. That is where and when a driver accesses location information from the system, the external supports and the surrounding environment. We are currently collecting data from 20 semi-structured interviews each augmented by in-car observations of 1-hour ride. The study concentrates on the taxi drivers of the city of Barcelona, Spain. This community forms a massive population of early adopters of in-car navigation systems with a strong past practice of relying on mobile technologies and maps to support their work.

Relation to my thesis: I can’t imagine a better set of people to receive feedback on my taxi driver study.

Today I gave a talk to my research group on the current state of my investigation. The presentation covered the topic of my DEA work, the early results of Tracing the visitor’s eye project and their implication for a field experiment taking advantage of digital footprints to support social navigation and seamful design. One personal goal of this presentation was to prepare my DEA defense in November and next week’s presentation of Understanding of tourist dynamics from explicitly disclosed location information at the 4th International Symposium on LBS & TeleCartography.

Towards Reducing the Social-Technical Gap in Location-Aware Computing (PDF)

Relation to my thesis: Trying to put the pieces of the puzzle together. The overall argumentation still lacks of coherence. I will need to make some choices and get rid of some themes to explore others profoundly. Most of all, my research questions are now outdated. I need to refine them. More on that later.

Meeting to discuss my DEA thesis. We agreed that the focal point of my research shifted from uncertainty to granularity of location information. The first is a problem to solve (while sometimes being an opportunity), the latter is a source for interaction. In my model of the social-technical gap in location aware computing, I intend to define their relations. I hypothesize that uncertainty appears when a location system does not match the granularity of information expected by a user. So I keep the work I have done so far on the reactions (uncertainty) to fluctuating location information, and focus more on the factors influencing people to tune the information. This is what I need to further investigate in my ethno study of the taxi drivers (e.g. the funnel metaphor to access information, their use of neighborhoods, landmarks, addresses). Similarly with my Flickr study I could include the analysis of the textual description (i.e. tags) to understand how the users describe the granularity (for example: city -> landmark). Results of complete studies from different contexts could already be a nice outcome to define key aspect of human interaction with location information granularity in a mobile context. It could open the door to the definition of sub-issues (psycho, social, cultural, gender, …) that would be mostly outside of the scope of my thesis. No decisions have been made on further studies (let’s see what the outcomes of the 2 current studies), but we certainly don’t lack of ideas. The concept of granularity of location information is nothing new. However, it is worth revisiting it since “we use things that did not exist previously”.

As for the DEA thesis, he shared my mixed feelings. I believe I have not achieved a good breadth-depth ratio, trying to cover too many aspects of my research domain. Then I lacked of energy to argument the choice (why a mention to privacy? -> uncertainty as opportunity + studies in spatial cloaking) and linkage of the key concepts. However, I think the breadth of this initial scope will help me in the long run. We discussed that chapter 3 (literature review) was not well self-contained. That is that I did not argue enough for the choice of the topics, the perspectives I chose to cover them, and their relevance to my work. The last section (Discussion) clearly revealed that lack of connection between the concepts described and the future work. I should rewrite that part by focusing and arguing the key elements of my framework and their relations.

Relation to my thesis: Still not quite in the narrow part of the funnel… but working on it.