Setting out to determine how far human perceptions can be used to characterise cities accurately, the MIT Media Lab has developed a tool which asks people to assess pictures of urban areas and then quantifies their perceptions.
Even to the casual observer, the stark inequality of different urban districts, in terms of the economic or social differences which separate them from the rest of the city to which they belong, is usually apparent. A citizen’s perception of his/her town varies, according to whether a given area is residential, commercial, affluent, run-down, and so on. So far, however, our ability to assess different urban areas and the type of environment they represent remains unproven. Now three researchers from the Macro Connections Group at the MIT Media Lab – Philip Salesses, Katja Schechtner and Cesar A. Hidalgo – have developed an IT tool capable of building a cartographic representation of the immediate reaction that ordinary people have when shown pictures of a city area. The programme has been set up to measure and test contrasting urban perceptions based on empirical data.
To carry out their pilot study, the researchers set up an online site, Place Pulse, where Internet visitors were invited to take part in the experiment. Each participant is shown two photos taken in the same city and asked to click on one in answer to one of three questions, namely: Which place looks safer? Which place looks more unique? Which place looks more upper class? Once the answers have been collected, an algorithm is used to establish a response ratio for ‘for’ and ‘against’. This helps to refine the process so that a photo which elicits a favourable response is not always matched up against a view which provokes negative feelings. Once a sufficient number of evaluations have been made, the data is represented on a map, thus building up a picture of people’s empirical perceptions of a given urban environment.
Areas of harmony and discord
For their pilot study, the researchers looked at a number of boroughs and districts in four cities: New York and Boston in the United States and Linz and Salzburg in Austria. With the aid of input from 7,872 Internet users from 91 countries, they were able to establish comparisons between photos from Google Street View and those taken manually by their own team. A number of conclusions were drawn. First, people’s opinions on the US cities were more clear-cut than those on Linz and Salzburg in Europe, i.e. the differences between the ‘best’ scores and the ‘worst’ scores for the images shown were wider, and thus the variances in perception were significantly higher than those for the Austrian cities. In addition, the pilot study also compared the New York urban perception results with actual figures from the city’s police department. The MIT Place Pulse tool showed a strong correlation between negative urban perception and the location of violent crime.