ICML 2017

In measuring the Gender Diversity Index (GDI), we don’t have access to gender information, so we tag the gender most associated with a given first name. Thus, the number of species is 2: male and female. 

We calculate the GDI following the Shannon evenness [1] (we discard richness) by means of the Pielou’s diversity index [2]:

H’ is the Shannon diversity index [1] and H’max is the maximum possible value of H’ (if every species was equally likely):

p is the proportion (n/N) of individuals of one particular species found (n) divided by the total number of individuals found (N), and S, that is the number of species.

J’  is constrained between 0 and 1, meaning 1 the highest evenness. 

For calculating the Gender Diversity Index, we should consider three different communities: keynotes, authors (of the 10% of the papers) and organisers. Our final Gender Diversity Index performs a weighted average among the Pielou’s evenness index in each community, giving more the highest priority to keynotes (they get more visibility) and less priority to organisers (they get lower visibility):

References:

[1] E. C. Pielou. The measurement of diversity in different types of biological collections. Journal of theoretical biology, 13:131–144, 1966.

[2] C. E. Shannon. A mathematical theory of communication. Bell system technical journal, 27(3):379–423, 1948

For calculating the Geographical Diversity Index we will collect the country of affiliation for keynotes, authors and organisers. As we can have multiple species and we also want to measure the richness together with the evenness, we will apply the weighted average of the Shannon Index community [1]. This index may be greater than 1, so we normalise it between 0 and 1 by dividing it by 4 - i.e. the maximum value: 

p is the proportion (n/N) of individuals of one particular species found (n) divided by the total number of individuals found (N), and S, that is the number of species.

References:

[1] C. E. Shannon. A mathematical theory of communication. Bell system technical journal, 27(3):379–423, 1948

In measuring the Business Diversity Index (BDI), we consider 3 species: academia, industry and research centre.

We calculate the BDI following the Shannon evenness [1] (we discard richness) by means of the Pielou’s diversity index [2]:

H’ is the Shannon diversity index [1] and H’max is the maximum possible value of H’ (if every species was equally likely):

p is the proportion (n/N) of individuals of one particular species found (n) divided by the total number of individuals found (N), and S, that is the number of species.

J’  is constrained between 0 and 1, meaning 1 the highest evenness. 

For calculating the Business Diversity Index, we should consider three different communities: keynotes, authors (of the 10% of the papers) and organisers. Our final Business Diversity Index performs a weighted average among the Pielou’s evenness index in each community, giving more the highest priority to keynotes (they get more visibility) and less priority to organisers (they get lower visibility):

References:

[1] E. C. Pielou. The measurement of diversity in different types of biological collections. Journal of theoretical biology, 13:131–144, 1966.

[2] C. E. Shannon. A mathematical theory of communication. Bell system technical journal, 27(3):379–423, 1948

The General Diversity Index of a conference (CDI) will be calculated by averaging the proposed indexes for gender (GDI), geographical (GeoDI) and business diversity (BDI).