In the following, we express all demographic proportions with respect to the pool of individual, identifiable artists for whom we could reach confident demographic inference.
These stories belong to culture.Third, and most importantly, we had each artist-museum record researched by at least five independent workers. A small number of workers completed unusually many HITs, with most completed in 30 seconds or less. Just nine months into , the combined number of works by African American artists acquired by museums total, so far is on track to become a ten-year record. Now, access to comparable works is tightly controlled—and far too expensive for most institutions to afford. These stories belong to culture. We manually corrected these inconsistencies, and similarly, we corrected two inconsistencies for geographic origin and five for birth decade. To identify a record as IIA, we required that it was evaluated by at least three workers, and that a majority of the workers classified it as IIA. Ref: arxiv. Geographic origin For the 42, IIA responses available for geographic analysis, we mapped national origin response to one of the seven GEO3 regions used by the United Nations [ 37 ].
As a consequence, the true proportion of women artists in our study may be slightly larger than what we report here. The combined stage 1 and stage 2 sample size of artist records drawn from each museum is shown in Column 7 of Table 1.
The diversity is not much greater for museums that specialize in modern art. We analyzed data submitted by the 17 most prolific workers, who each completed between and 1, HITs, and found that eight of them appear to have acted in bad faith, submitting large quantities of incorrect information.
Ethnicity Of 41, IIA responses available for ethnicity analysis, only provided a free-text ethnicity.Nonetheless, this shift is extraordinarily recent. We inferred all four demographic traits through a survey instrument deployed via crowdsourcing. We then scraped the catalog of holdings [ 12 — 29 ] using customized code we wrote with the R project for statistical computing [ 30 ], retaining a web link for each item. We excluded these responses from gender analysis as we could not confidently conclude any artists to have nonbinary gender. All subsequent questions allowed the worker to answer that the artist is non-IIA. We refer to cases meeting these requirements as confident gender inferences CGI. Data scraping For each museum in our study, we read the museum website Terms of Service and verified that gathering data for research would comply with them, as well as with United States Fair Use doctrine. Just nine months into , the combined number of works by African American artists acquired by museums total, so far is on track to become a ten-year record. Birth decade Of the IIA HITs available for analysis of birth year, we retained the 40, responses containing three or four digit numbers. The museum plans to co-organize a major Thomas exhibition in Acquisitions and exhibitions require expertise and resources that museums are constantly under pressure to invest elsewhere. Survey instrument Survey overview. In referring to specific museums, for concision, we again use the abbreviations of museum names as given in Table 1. As we will describe, we used this redundancy to form consensus demographic inferences. Though our survey instrument allowed for the possibility of multiple primary ethnicities, we found that only two out of our 10, IIA records were confidently inferred as such.
This database contains 68 of the artists that we also sampled for our study. This is nothing new. We refer to cases meeting these requirements as confident gender inferences CGI.
As we will describe, we used this redundancy to form consensus demographic inferences. The Virginia Museum of Fine Arts, for example, made its desire to become a leader in the field of African American art a prominent part of its strategic plan in Now, the late artist is the subject of a major exhibition at the Met Breuer in New York.