From CS294-10 Visualization Fa08

Jump to: navigation, search

Assignment: Assignment 1a: Good and Bad Visualizations and Assignment 1b: Visualization Deconstruction and Redesign

Student: Jeff Bowman


Bad visualization

Black Metropolis, p183
Black Metropolis, p183
Black Metropolis, p4, for context
Black Metropolis, p4, for context

Source: Drake, St. Clair and Horace Cayton. Black Metropolis: A Study of Negro Life in a Northern City. Harcourt, Brace & World, 1945. p183. Referenced in Ananya Roy's reader for Environmental Design 100: The City, Spring 2008.

This visualization was published to accompany a 1945 analytical book of the African-American condition in Chicago ("Black Metropolis"). This particular area was studied through multiple maps in the same book; the map's first appearance is on page four (also pictured in this assignment). The map describes socioeconomic conditions in different areas in Chicago, with eight categories: "Hotel and Apartment Hotel", "Native-Born White Apartment", "Single Home and Two Flat, with Rent Over $50", "Apartment and Two Flat", "Foreign Born Apartment", "Single Home and Two Flast, with Rent Under $50", "Tenement and Rooming House", and "Negro Neighborhoods".


The map's main feature is the set of eight different categories that it overlays over the map of Chicago. The map provides a scale and direction, allowing it to be overlaid with other geographic data from other sources; however, some local information (presumably that's a river in the top right" becomes obscured due to the rich data and patterning.

However, while the categories seem more nominal than ordinal in some contexts, the authors place the data in a particular ordering to suggest that the categories are in rough order from "most affluent" to "least affluent", or something similar to that. While each label describes a certain section of the population, the ordering allows the map to highlight the darker and lighter areas, providing for an inherent summary of the form that ordinal maps generally take: The darker areas all conform toward one grouping, and within that the darkest areas are the most symptomatic. These categories correlate with color value, or arguably texture, to describe the single quasi-ordinal variable.


  • Different scales. While the image on page 183 fits with the rest of the book, particularly the introductory image on page 4, the image on page 4 has different boundaries and no purely geographic labels. The implication is that the reader will "figure out" where the problem areas are, as they are consistently the darkest among the pictured maps, and Chicago has relatively well-defined (if unlabeled) geographic features.
  • Too many patterns. The use of eight different patterns (ranging from solid white, dotted, hatched, crosshatched, solid, and three more in between) makes it time-consuming to identify which area corresponds to which pattern. While some of the visualization could have been distorted in the photocopying and subsequent scanning, the fact that the image is monochrome (and published in 1945) implies that the original still suffered from the same problem.
  • Background matching a pattern. A plain white background surrounds the image, and also indicates a particular type of data. This is especially troublesome because the white-colored areas happen to be very close to the edge of the map; the problem would be much less pronounced if the white areas were near to the center of the map.
  • Optical interference. The use of two types of diagonal hatching and their overlaid crosshatching produces a moire effect that can hinder analysis.
  • Data combination fallacies. Likewise, the diagonal crosshatching is unrelated to the two types of diagonal hatching, despite the reasonable assumption that the crosshatched area would have the attributes of both of the diagonal hatches.
  • Compression into one data dimension. The data is ordered from light to dark, despite having completely separate qualifications (e.g. hotels versus houses versus foreign tenants versus rental apartments at different rent amounts). While we can assume a natural ordering of some of the qualifications, and deduce that the authors ordered the rest of the patterns intelligently, the graphic does not lend itself to flexible analysis.


Naturally, removing the interference, background contrast, and combination fallacies is a matter of picking different patterns or colors, and adding relevant geographic labels.

In an ideal world, the graph would also be available as colored sections of equal saturation and value, when the ordering is not natural. In cases where the ordering is natural, changing saturation and value could effectively demonstrate the same point.

Finally, as is demonstrated elsewhere in the book and mentioned below, smaller identical maps (side by side) could demonstrate different data dimensions that could be compared across one another.


  • The reconstruction uses color, which was unavailable in the visualization's original form, which I chose over the task of encoding three separate maps with each of the three datasets. Also, I chose to keep the original map's quasi-ordinal map categories and color value encoding, but added color hue to encode the related nominal sets which are also included.
  • I included a small version of the original map, which provides more geographic context; while I wanted to add extra geographic information that was impossible to decipher in the original map, given the new capacity to clearly add additional black lines, I couldn't—I wasn't able to decipher it.
  • The use of a single texture (solid) allowed me to include relevant labels over general areas of the graph, which I produced from the aforementioned overview map.
  • Finally, the explanatory text at the bottom uses color references to ease use with the map.

Good visualization

Source: Graham, Stephen, and Simon Marvin. Splintering Urbanism: Networked Infrastructures, Technological Mobilities, and the Urban Condition. Routledge, 2001. p241. \Referenced in Ananya Roy's reader for Environmental Design 100: The City, Spring 2008.

Splintering Urbanism, p241
Splintering Urbanism, p241

This image describes the socioeconomic conditions in different areas of Washington D.C. In each of the three maps, the designer includes a different subset of three socioeconomic classes detailed elsewhere in the article: In the top left, "Money and Brains", "Blue Blood Estates", and "Young Influentials"; in the top right, "Black Enterprise", "Emergent Minorities", and "Downtown Dixie Style"; in the bottom left, "Bohemian Mix", "Young Suburbia", and "Furs and Station Wagons". These are overlaid over a map of the Washington D.C. area, including geographic features outlined in black.


Because the data is displayed in three maps, each of which contains three different mutually exclusive categories, the maps in combination produce nine separate nominal categories. (I assume the categories are nominal, rather than in a relative ordering as above, because the legends are ordered randomly and are not portrayed as being along one dimension.) The visualization uses color value to encode between the three features (and map areas that fall into no particular category). This data is correlated to the geographic features, which—in addition to aligning data between the graphs and easing comparisons—also allow the user to make geographic observations such as noting that minority activity is particularly strong near the Washington D.C. city limits and to their immediate east. While a scale is not provided, the map form of the data and identical map scales produce distance observations (such as which communities are farther from the city center) that also communicate relevant data.

Notable attributes

  • Smaller, multiple maps. This allows each map to only show a limited number of relevant attributes, while still allowing the data to be compared. Also, the maps are displayed on the same page immediately horizontal or vertical to a corresponding map, easing comparison.
  • More prominent geographic features. This is partially a byproduct of showing less information on each map, allowing more auxiliary information on each map.
  • Limited, easily-distinguishable colors without interference. Each map only has a couple of patterns, preventing visual interference, implied orderings, and reuse of the background color.

[add comment]
Personal tools