Dust is an ongoing research project developed for Iridescent Learning, an American NGO. The project’s aim is to provide a freely available, web based information visualization tool that supports parents in exploring and comparing the educational offerings (from Pre-K to High School) from selected major cities in the United States, currently: New York, Los Angeles, and the San Francisco Bay Area. By leveraging a step-by-step decision making process, Dust helps to evaluate and compare school profiles based on multidimensional data-sets composed of general information (e.g., enrollment, class size, number of teachers), school performances (e.g., subjects score and proficiency, attendance), and urban mobility (e.g., location, distances, transportation). Supported by geographical maps and close-up visualizations users can create custom profiles based on their needs and priorities and then perform a search for the most appropriate schools for their children. Dust aims to combine the capability of information visualization in depicting synthetic views of complex, multidimensional, and georeferenced data; with a rich, yet intuitive, web-user experience. The project aims to move away from a “by experts, for experts” design paradigm to a schools comparison information visualization “for the people” -providing real impact on their daily life, and future prospects, through improved choices.
School choice
What does it mean to say that a school is a good school? We could estimate the student-teacher ratio or we might measure the educational outcomes achieved by the students, or perhaps we could consider the number of student expulsions in the last year. However, all these variables, taken one at a time, gives us just a partial view of reality: they all are focused on a specific aspect and are unable to show the big picture. This is because the phenomenon we are observing has a complex, multidimensional nature. This kind of phenomenon cannot be described and measured by a single dimension, instead it must be defined by a structure of variables related to each other. The goal of the Dust project is to relate these variables through a step-by-step process. Users can observe and compare schools according to a custom profile. In order to accomplish this task, the choice of the best school has been conceptualized as a Multi-Criteria Decision Making (MCDM) problem.
These collective problems do not yield to a single “optimal” solution, instead it is necessary to use decision maker’s (i.e. parents) preferences to differentiate between solutions. In essence, Dust aims to implement a MCDM process through a visual interface, making it accessible and intuitive toward the purpose of school selection. In general, a MCDM process is composed of four attributes: a) the alternatives, b) the criteria; c) the evaluation system; d) the results representing a satisfying trade-off between the criteria. The alternatives include the schools and the criteria that correspond to the variables for each school. Respecting the evaluation system a rating classification model has been chosen—this provides a range of satisfying results, not just the supposed best result.
By adopting this system it is hoped that the user is encouraged to further explore the set of good schools. This allows users to gain a deeper knowledge about a group of schools prior to reaching a final decision
The visualization
The MCDM process is based upon multidimensional datasets. Even though finding information related to schools and education in the United States is relatively easy, especially due to the releases of open data, it is still difficult to obtain homogeneous data sets across different states and associations. A complex mash-up of several data-sets has been created by Iridescent and Factual (the technical partner of the project). In fact, harvesting, cleaning, and preparing digital data for analysis has been a crucial task, in order to turn accessible open-data into effective data use
One of the aims of Dust is to exploit the capability of data visualization techniques to assist in exploring multidimensional data-sets. By combining visualization with the communicative objectives of formal design disciplines, Dust provides even non-experts with a flexible tool. The resulting visuals depict nuanced representations of the complex data composition they are observing. When visualization is utilized as a mere transcript or a transposition of the data (from a numerical form to a graphical encoding) it may provide no new meaningful insights.
Therefore, in dealing with multidimensional data, mapping the numerical values alone is not always sufficient to gain new knowledge. Dust, however, implements the “visualization as a transformation process” paradigm which highlights how visualization tools are always part of a communication process, where the user experience and the context are key, value-added, drivers of the design
process.
Toolset description
Dust presents a pan-enabled and zoomable geographical map with an overlapping schools layer, one profiling panel, two tray panels, and a search field. While the position of each school on the map comes from its actual coordinates, the size and the color depend on two dimensions drawn from the data-set. The user can decide the dimensions to depict through a dropdown menu on the top-right tray panel. In this way three different dimensions of the dataset can be displayed simultaneously. Besides browsing the map, the user can also search a specific school, address, or point of interest by using the autocomplete search field. In case of a matching result, the map is centered on the found element.
As previously mentioned, Dust provides two modes of use: a free mode, where the user can freely navigate the map and select, compare, and search for schools; or a profiling mode where the user is guided through a step-bystep process in order to find the schools that could fit her specific needs. While the free mode has been designed for an open-ended exploration, the profiling mode has been conceived as a visual guide to perform a multi-criteria analysis. By using the profiling mode, a customized map will be drawn: the output is a restricted pool of schools arranged in ordered categories based on the implicit user evaluation: the good ones, the sufficient ones, and the bad ones
Technical consideration
Dust is a Web application – in beta at the time of writing – developed in Javascript and Python. The HTML/Javascript front-end uses several open source libraries. Geo referenced data are displayed on a map using Polymaps library (www.polymaps.org) with maps data coming from OpenStreetMap (www.openstreetmap.org) and tiles from CloudMade (www.cloudmade.com). Visualization elements are created using D3 library (mbostock.github.com/d3), while UI controls are built with jQuery UI (jqueryui.com/). Data are coming from several open data repositories and have been collected on a Factual database (www.factual.com). In order to call Factual’s APIs and perform some preliminary data cleaning and analysis, a small back-end application has been developed using the Python Web framework Django (www.djangoproject.com). Data related to geographical distances and transportation are retrieved by Open Transit Data
