Context
Bogota is the capital of Colombia. Both the perception of unsafety of its inhabitants as well as the amount of urban crime have been rising over the last years, with over 70% of the population feeling unsafe, and more than 120 thousand theft victims in 2023. The local government response to this has mainly focused on increasing the police presence and law enforcement. Is this the only solution? Can we as designers do something to reduce crime in the public space?
There has been a continuous debate among two principles for safety in the public space: on one side, Oscar Newman (1972) with the theory of “Defensive Space” which states that safety is provided, among other principles, by raising clear delineations between public and private, raising gates and fences to protect what is private, ensuring formal surveillance and demonstrating territoriality <a perspective that matches in some sense with the current and historical efforts of Bogota municipality. On the other hand, authors like Jane Jacobs (1960) and Jan Gehl (1996) have defended the idea of “Natural Surveillance”, arguing that safety in the public space is achieved by encouraging more people, activities and therefore “eyes in the street”, as well as activity in the ground floor, high permeability, entrances and windows facing the streets and mix uses.
In Bogota particularly, as well as many other in the Latin American context, this representation of “life in the street” materializes as different representation of informal economies. Although informal street vendors are a big part of the urban landscape, they are considered occupants of the public space, as well as an unfair competition to business owners that do have to pay for the use of their premises. Therefore, the strategies of regularization are commonly oriented towards their relocation and they are pushed away from the public space.
Why bring informal vendors and urban safety into one conversation?
- Because there are case studies demonstrating that regularization does not have to mean relocation, such as the case of India’s National Policy on Urban Street (2009).
- Because crime is still rising, regardless of the continuous efforts of solving it through police enforcement.
- Because although are two current problems in Bogota as well as many other Latin American cities, there is not enough research on how to both of them are linked
- Because informal vendors do shift the function of public space as a path for movement into ephemeral destinations for pedestrians to wait for an appointment, wait for a person, have a phone call or have a snack or a coffee.
- Because, although informal vendors do occupy public space, sometimes they are filling up urban voids that would otherwise be unused.
- Because, if proven that they could serve as assets to reduce crime, this could be the excuse that the municipality is missing for encouraging the development of policy to acknowledge them as part of the public space.
Research Framework
This research is an opportunity to search for correlations between the location of informal vendors and the occurrences of crime events in the public space, by spatializing the project and finding correlations between them and the morphology and dynamics of the public space. This would allow to answer questions like:
- Which physical characteristics foster the presence of informal vendors in the street?
- Which physical characteristics propitiate the occurrence of crime?
- How does the presence of informal vendors influence the occurrence of crime?
Methodology
The thesis is to be developed in three steps:
- Data gathering
- Stablishing correlations
- Figuring how to have an impact
This blogpost will focus on the current development of the first chapter, Data Gathering.
How to gather data?
The first attempt to retrieve information about crime and informal vendors was through official requests to the dedicated agencies within the local government, but the official access to the existing datasets was not allowed. Therefore, different strategies were developed for retrieving the information from alternative sources. The outcome is a unified pipeline for city scale implementation, that is fully scalable, reliable on open resources and fully developed in Python with no external software needed.
How to gather data about crime?
Although the access to the raw data of crime events is not possible, the police department of Bogota publishes monthly reports for each of the 19 Localidades in which the city is spatially divided. Each of these reports contain the aggregates for seven types of crime: i) murders, ii) personal injuries, iii) personal theft, iv) vehicle theft, v) motorcycle theft, vi) phone theft and vii) bike theft. For each of these types two tables are provided: i) the disaggregation by gender and ii) the disaggregation by spatial subdivisions called Zonal Planning Units (UPZ).
A Python pipeline was developed using the OpenAPI to create an assistant that receives each PDF and outputs a structured dictionary for generating two datasets: i) a timestamped dataset with the monthly sum of victims by crime type, Localidad and gender and ii) a timestamped and geolocated dataset with the monthly sum of victims by crime type, Localidad, UPZ, for the period 2016 to 2023.
By comparing the results with a deprecated dataset, this methodology was proven to provide 85% of the correct data, and the spatial distribution of both datasets show a very close visual correlation.
This pipeline functions entirely with open data and payment by demand of the OpenAI API of USD $5 for 900 documents.
How to gather data about informal vendors?
There are many representations of informal vendors in the city. Some are pedestrians themselves that carry products in their chest, some have fixed stand, others a moving cart, others exhibit their products in a car trunk, some lay the products in the ground, some travel around in the public transport, and even some formal business informally extend their activities into the street. For this research the ones that fall within the scope and interest for their location and further correlation are those who play a part in the pedestrian right of way, either by involving them as customers or just by sharing the same space.
A Python pipeline was developed using the Google Streetview API and the Mistral API for parametrically retrieving images along the pedestrian network and using the Vision capabilities of Mistral Vision model Pixtral to answer 4 specific questions or instructions.
- Is there an informal or street vendor in the image?
- What is distance of the informal vendor in the picture on a scale of 0 to 3 (0: not present, 1: far, 2: midrange, 3: near)
- What is the description of the informal vendor (if answer to previous question is positive, otherwise answer None) using 0 to 5 keywords?
- Provide a description of the context of the picture using 0 to 5 keywords.
This pipeline is 100% free and relies on globally available data with the ability to process approximately 150 images per hour with a current output of 11% out of 111 thousand images to be processed for a high resolution and city scale query.
How to measure the morphology of space?
For measuring the morphology of the public space, three concepts are taken from the principles of bill Hillier’s Space Syntax for performing a quantitative evaluation of the accessibility and connectivity of the pedestrian street network.
- Betweenness or Choice: is an indicator for measuring how traversed a path is.
- Closeness or Integration: is an indicator for measuring how close, proximal or central a path is related to the rest of the street network, as an origin or a destination.
- Visibility: is an indicator of how far can an individual reach given a radius and a set of obstacles, commonly buildings.
The Space Syntax methodology is commonly implemented using software such as UCL Depthmap, QGIS + PST Plugin or Rhino + Grasshopper. In this case, a 100% Python pipeline was developed to retrieve such indicators. The graph analyses were performed using OSMNX street network and IGRAPH library that optimizes computation time by running the back-end processing in C. The visibility analyses were performed using the library T4GPD that integrates the calculation of isovists given certain obstacles. Although commonly approached using buildings as obstacles, the pedestrian reach in the public space of Bogota is commonly also limited by other set of obstacles including pedestrian and vehicular vehicles as well as long sections of highways. For this reason, these were also weighted in the analysis.
How to retrieve data about the formal city dynamics?
Additionally, an extra layer of indicators is considered taking into account that normally they choose strategic locations where there is a preexisting cluster of activities, with examples such as commercial or educational clusters. For this, the Overture Maps places open dataset was retrieved.
With this set of points of interest, two types of indicators:
- Activity density: is a measure of the quantity of points of interest within a given area.
- Urban complexity: is a measure of the mix or diversity of uses. Takes its origin from the ecological background using the formula Shannon entropy formula.
Another set of indicators is integrated for taking into account the proximity of informal vendors to certain infrastructural elements, such as public transport stops (more specifically to Bogota Bus-Rapid-System, TransMilenio) or pedestrian and vehicular bridges.
Finally, for integrating any other contextual element, the urban characterization performed by the Pixtral Model allows to generate Word Clouds per different sets of polygons for performing a quantitative classification of the keywords provided.
What comes next?
This multicriteria data gathering pipeline allows to generate a city-scale battery of indicators. This will be the main input for further steps in which:
- Different regression models will be compared and assessed, mainly in the effort to compile a model that explains crime as a function of the spatial morphology and the presence of informal vendors.
- Temporal dimension of data will be integrated to take into account variations in yearly, monthly, weekly and daily horizons.
- The different scales and resolutions of data should be unified for accurate comparison.
- Informal conversations with informal vendors and pedestrians will be performed, to obtain a grounded and bottom-up perspective of the issues regarding the research topic including both user types’ perceptions of safety, crime, visibility, accessibility, among others.
- Identify tools that can take the statistical results of correlations as input for generating impact for making data-based policy for informal vendors regularization and proposing design-based solutions for fostering safer spaces.
