Date Published: June 20, 2018
Publisher: Public Library of Science
Author(s): Daniele Quercia, Luca Maria Aiello, Rossano Schifanella, Renaud Lambiotte.
Over the last few decades, public life has taken center stage in urban studies, but that is about to change. At times, indoor activities have been shown to matter more than what is publicly visible (they have been found to be more predictive of future crimes, for example). Until recently, however, data has not been available to study indoor activities at city scale. To that end, we propose a new methodology that relies on tagging information of geo-referenced pictures and unfolds in three main steps. First, we collected and classified a comprehensive set of activity-related words, creating the first dictionary of urban activities. Second, for both London and New York City, we collected geo-referenced Flickr tags and matched them with the words in the dictionary. This step produced both a systematic classification (our activity-related words were best classified in eleven categories) and two city-wide indoor activity maps which, when compared to open data of public amenities and sensory maps of smell and sound matched theoretical expectations. Third, we studied, for the first time, activities happening indoor in relation to neighborhood socio-economic conditions. We found the very same result for both London and New York City. In deprived areas, people focused on any of the activity types (leading to specialization), and it did not matter on which one they did so. By contrast, in well-to-do areas, people engaged not in one type of activity but in a variety of them (leading to diversification).
What makes urban communities successful? A vital public life, urban studies would suggest [1–3]. To promote that vitality, urban planners act upon the physical structure of neighborhoods. In her 1961 book, The Death and Life of Great American Cities, urban sociologist Jane Jacobs argued that city life is a product of the diversity of the physical environment . To promote diversity, four conditions are necessary: 1) neighborhoods must serve more than one function, attracting people at different times of the day and night; 2) city blocks must be small and have dense intersections, encouraging interaction between pedestrians; 3) buildings must be diverse in age and form, supporting a mixture of low- and high-rent tenants; and 4) neighborhoods must have a sufficient density of people and buildings.
In 1987, Recker et al. studied the relationship between activities and people’s mobility . They asked 665 individuals in 249 households to keep travel diaries in which they had to record all trips made. Being a manual process, the classification of activities was coarse-grained, in that, they were classified as being related to work, education, shopping, social, and recreation. A decade later, Janelle et al. still used activity diaries but were able to embark in a thorough study on how the use of neighborhoods changed over time . They did so by having their participants manually classify activities based on what-who-where: what type of activities, with whom they were done, and where. In 2004, Kahneman et al. published a Science paper that refined the way activities are elicited: they proposed a diary method to gather how study participants spent their time and how they experienced the various activities (e.g., in which mood they were) . In it, activities were richly classified into sixteen types.
This work proposes a new way of analyzing indoor spaces from data implicitly generated by social media users. The idea was to search for activity-related words on geo-referenced picture tags. To this end, we needed words and data, both of which are described next.
We then studied how the presence of certain activities was related to the neighborhood’s socio-economic conditions.
This work adds to the growing literature exploring the ways in which online data can be used to study the socio-economic development of communities [6, 30]. We demonstrated a significant link between neighborhood deprivation and indoor activities. Next, we discuss some of the limitations of our work and some of the opportunities it opens up.
Activities behind closed doors have been out of view of many measurement techniques. This work is the first in examining the role of social media in mapping activities happening in indoor spaces at city scale. Despite self-reporting biases in such data, we still found that, in their private or semi-public spaces, people expressed words of protest and reported sexual activities. More generally, we found evidence that, as opposed to people in deprived communities, people in well-to-do ones tend to diversify their activities. It was not a matter of picture volume or service penetration rates—people in well-to-do communities did allocate their time in a more diverse way. We hope to empower designers, researchers, city managers by offering them a number of methodological tools and practical insights to re-think the role of indoor activities in their work.