Sampling, Crowd-Sourcing & Reliability - (E3) Salesses...and Hidalgo 2013

[HyunkuKwon]

Post questions about the following exemplary reading here:

Salesses, Philip, Katja Schechtner, and César Hidalgo. 2013. “The Collaborative Image of The City: Mapping the Inequality of Urban Perception.” PLoS ONE 8(7):e68400. doi:10.1371/journal.pone.0068400

[RobertoBarrosoLuque]

The method developed by Salesses et al. to measure inequality of perception through photo comparison and crowd-sourcing is very interested. My questions are on some of the applications that such methods could have:

• Can urban perception studies help guide policy?
• Can citizens provide input and feedback to their local governments through a similar approach?

[Willy624]

This paper is very clear in its goal, not trying to find any causality, only looking for a good measure for the information inside urban perception. However, I have some questions about the sampling method they use.

If the point of this measure if to capture an "universal" urban perception, then the sample size still seems to be too "WEIRD". If however, we only want to make sure this measure can be helpful to urban administers, the focus should probably be on those who lives in the city and a little bit of travelers.

In general, I believe this measurement can be used to solve urban design problems, yet it heavily relies on "who" the researchers choose to sample, which should be linked to the problem they want to investigate.

Though a bit off-the-mark (not entirely related to this paper), I am curious are there any methods social scientists have adopted to solve the sample being too "WEIRD" problem? Or in social science we should recognize there's cultural differences, hence we should instead try to restrict our sampling population and solve the non-generalizable questions?

[jinfei1125]

This is a very interesting paper and it's an excellent example using MTurk to crowdsource the work. It also reminds me of an Atlas constructed by MIT Media Lab to draw the distribution of inequality. It uses people's geographical traces and their estimated income level (according to the income level of their living neighborhood) to study the each locations' inequality. Now it has studied the distribution of several big cities including NYC, Boston, and Chicago. For example, this is the distribution of inequality in Hyde Park: If you look at other neighborhoods in Chicago, you will find that the inequality level of downtown is relatively low, but South Chicago is relatively high.

OK, back to the topic, this study mainly measure people's "perception", other than "facts" like the above study I mentioned, which uses the real data of people's movements during a day (though they have discussed some ethical concerns). Do you think people's perception is a good reflection of the reality?

[k-partha]

I'm excited to understand how we can better leverage the incredible temporal and spatial resolution in the data available. I found the Broken Windows theory - referenced by the authors in their explanation of the connection between crime and the urban environment - an interesting causal framework.

I'm interested in whether we can understand which came first - crime or poor built-up environment (the chicken or the egg). Are there any interesting research designs - that take temporality into account - that we can apply in this case to validate the Broken Windows theory?

[MOTOKU666]

It's a really interesting paper about how images may establish criteria for evaluating inequality. I initially thought there would be no human response involving and the researcher set a benchmark city. After reading the paper, it seems more accessible to have the people grading, comparing at the same time recording the time. I'm wondering whether it's possible to combine the benchmark(Maybe set by urban scientist and sociologist together) and the grading system.

[joshuabsilver]

Unsurprisingly, people looking at images can perceive inequality/safety in urban spaces. In terms of sampling, why didn't the authors collect geographical and demographic data on the survey participants? I agree with @Willy624 on this point. A more interesting research question, I think, would be how individuals from one part of a city perceive the landscape of other parts of the city. For example, can we know if individuals from the Bronx perceive themselves to be in a safe environment if shown images of high income, highly policed areas in Manhattan? As safe as people from other wealthy parts of Manhattan?

Also, why the focus on violent crime? Are homicides that common? This is an awfully narrow definition/measure of perceptions of cities. Can we generalize from perception of "safety" based on images to making claims that we have mapped urban perception?

[Rui-echo-Pan]

Similar to what @k-partha brought about, I'm also interested in the logic behind the correlation between urban perception and crime/homicides. As the paper implies that there are some implications behind the people's perception about the city, it'll be interesting to explore more.

[xxicheng]

I am also wondering about the significance of urban perception. Could you please offer more researches on this topic?

[lilygrier]

I am curious about whether a clustering algorithm applied to the images would divide them into similar groups based on respondents' perceptions of safety/inequality/etc.. Are there specific features that people consciously or subconsciously associate with safety? Is there an image-analog to the TF-IDF that could give a sense of an image's uniqueness?