Assignment Blog Post on Lifestyle, Mobility and Location.

This Blog Post is in response to the Assignment #3 which can be found here or here. 

The Livehoods Project: Utilizing Social Media to Understand the Dynamics of a City

1.      What patterns are used by the authors to define a livehood cluster?

Authors define Livehood clusters based on “…the spatial proximity between venues as given by their geographic coordinates, as well as the social proximity” derived “… from the distribution of people that check-in to them.”
The authors use Foursquare check-ins by users in the targeted area to analyse and define Livehood clusters. The authors used only the Foursquare check-ins that were publicly disclosed by users on Twitter to collect the data, use Cluster mapping to define Clusters of user behavior which determined the borders of Livehoods within city municipal borders.

For each check-in, authors documented the user ID, the time, the latitude and longitude, the name of the venue, and the category of the place.

The data was further validated against semi-structured interviews with residents of the area under question, and industry experts from Real Estate Development & Urban Planners, in both Public and Private capacity.

There are three patterns of Livehoods based on this approach that Authors conceptualized:

-          Split – “Split patterns often show the different demographics or different functions that operate in certain city neighbourhood(s).”
For Example, in the same neighbourhood of  South Side Flats, we could see LH7 and LH8 being divided based on geographical and architectural differences, as well as population density, and local community habits and preferences (like greater nightlife., or desolation on roads).

-          Spill – “Spilled patterns typically reveal areas that are in transition or borders that are in flux.”
 For example, Shadyside and East Liberty, wherein, economic development in the area to the East of Shadyside, and introduction of Whole Foods Supermarket, had rapidly lessened the divide between the two neighbourhood, and was fast getting the neighbourhoods to have a fluid connect.
  

-          Corresponding – “Corresponding patterns indicate the strong influence municipal borders and geography have over local social interactions.”
As Authors describe, “in several cases, the Livehoods boundaries corresponded perfectly with the municipal borders indicating the strong role that neighborhoods do play in shaping people’s activity.” For example, the border between LH3 & LH4 was found to be 40^th St. Bridge, which was actually the exact border according to the municipal neighbourhood borders as well.

Based on their research, Authors defined that Livehoods can be influenced by the following factors:

-          Municipal Neighbourhood Borders – Although Municipal Neighbourhood Borders are usually rigid and take longer to change, because they are primarily based on functional resource allocation, Livehoods evolve with time and change in people’s behaviours, and hence are more dynamic.

-          Demographics – The Authors also found that “found strong evidence that the demographics of the residents and visitors of an area often played a strong role in explaining the divisions between Livehoods.”
“In addition, the lack of both users and venues data for certain areas provides another way of tracing its demographics.” As illustrated by the example of Hill District, one of the poorest districts, which did not even surface in the study, the lack of data reflected in the economic strata of inhabitants of that area, implying low rate of smartphone usage, or use of Foursquare (which was primarily younger educated population of 25-35 years). This also gives insight into a possible limitation of the study, with respect to the possible digital divide within various communities of the city.

-          Development & Resource – The Authors state that, “economic development can affect the character of an area.” For example, LH1 and LH2, where there was a  spill effect from between Shadyside and East Liberty, due to expansion of economic development on the Eastern parts of Shadyside, which was rapidly joining up both the neighbourhoods by creating a neutral connecting area between the two neighbourhoods.

Geography & Architecture – Authors state that “The flow of people through the streets of a certain area is shaped by the geography and the architecture of the place.” Example to validate this claim was some responses by the interviewees which indicated that there could be a distinction between LH7 and LH8, because there was a greater density of building or decrease in the width of the roads, in LH7, as opposed to LH8, which created a sort of split.

2.      What could be some of the benefits, for different stakeholders, of using social media data to understand the structure of a city?

Social Media Data gives the Urban Planners, or Real Estate Developers, as well as potential LBS businesses and Marketers various insights into user behaviours across the city, based on Check-ins.

Livehoods are one of the many ways of capitalization of such online social behavior of users in understanding city dynamics and ‘character’ better.

With greater penetration of Smartphones, and more prevalent check-ins across multiple media online, the data is more voluminous and insightful now than when this research was published, so the potential of this particular extension is immense.

This could help identify commercial opportunities for LBS businesses, retail businesses, and independent neighbourhood based services.

Some of the potential uses of understanding the structure of the city (but to limited to these) include:

1)       Providing Health Care & Transportation Services better and allocating resources appropriately to the potential needs of the population.

2)       Police Patrolling and Crime Mitigation

3)       Food Delivery and Homes Based Services Industry could vastly benefit from identifying user preferences across different geographies and demographics within the city.

4)       Urban Planners can allocate resources based on identifying population migration patterns within the city

5)       Real Estate Planners can identify potential new areas for Real Estate Development, and also the type of Real Estate Development (Industrial, Economic Housing, Gentrified Localities, and Retail Malls etc.)

6)       Population densities can also be observed and analysed using Social Media data, to better understand the movements of people around the city at any given point of time, and assign public transportation services accordingly (private and local municipality).

7)       Identifying areas for new Schools, Hospitals, and Market areas as well as Public Utility/Recreational Areas by Urban Planners.

Tweets from Justin Bieber’s Heart: The Dynamics of the Location Field in User Profiles

1.      How did the authors identify a user’s location indirectly?

Authors utilize their analysis of implicit information about location provided by Twitter users, within the explicit messages shared as tweets on the users’ profiles, and match the information against a predictive machine learning model that they developed, which analyses the proximity of key words, used within the tweets to a country or a state.

Hence, instead of relying on location fields, which authors found were significantly not providing accurate information regarding location of the users, (and substantial non-geographical information was being shared), authors focused on implicit information inadvertently and unintentionally shared by the users in their tweets.

2.      What can be some of the ethical and privacy implications of discovering users’ location indirectly in a social network?

Although the author underplays the privacy implication here, the data he furnishes points out that only 66% users furnished anything close to valid geographical location information. This implies that almost 1/3^rd of the users prefer not to disclose their locations or obfuscate their location data for various reasons, including preferring to use that field to rather evoke their interest or adoration towards a popular celebrity. At the same time, amongst the 66% of the users, there are users who only mentioned their continent, or country, or state, indicating a reluctance to share any specific details of their location. Most users preferred to stick to describing their city, as the most common location information, that was disclosed.

Hence, the authors conclude that there could be various ethical and privacy questions that could arise due to apparent reluctance of users to disclose the information by themselves. Authors suggest that before using such machine learning approaches towards any social networks, it would be advisable to first warn the users. And this predictive analysis approach could also lead to “users easily use this type of information to fool predictive systems.”

Another key point raised by the authors is that this sort of model could enable Location based services (LBS) further, but on the other hand, could also lead to location based “inference attacks”.

The Shortest Path to Happiness: Recommending Beautiful, Quiet, and Happy Routes in the City

1.      Describe a similar context where big data from social networks could be analysed and used to facilitate, or improve the livelihood of a place.

We could use a very similar approach as used by the Authors of the page to define user-generated recommendations on “when to visit” information on various places across the world. This could be done basically by cross-referencing the pictures or status messages and/or tweets uploaded by users of Social Networks, with the information shared about particular events the place is famous for – events, festivals, and/or weather conditions.

For example, in cold countries like Scandinavian countries or Canada, summer is really popular, and users generally post “bright and happy” messages and photographs in summer.

A model could be developed to analyse user generated content, which could crowd-source recommendations on when to visit particular spots across the globe real-time, based on analyzing such big data, and incorporating machine learning, to facilitate tourist influx based on user responses.

Places could also be rated based on the level of positive expression about the weather/events/festivals that users express, against the total number of active users in the area. The higher the rate and frequency of expression, the higher score a place gets, thereby facilitating a higher ranking in popularity.

This would be a good approach because, in addition to crowd-sourced ranking, it gives real time ranking, which could vary from year to year.

For example, if this year, Tallinn gets the highest ranking in terms of user appreciation during July, next year may not be the same case, if summer is weaker in Tallinn or “less appreciated” by the users, and instead, another place in another part of the world could be higher in rankings.

This mechanism of user generated rankings, could also influence tourism real time, if users accept and “listen” to the recommendations of these rankings.