Dataset XING57/2017

This dataset contains anonymized user profiles collected from xing.com in response to 57 queries. It was used in Jan-Feb 2017 to study gender biases in the returned ranked search results given each user's profile details.

Version

//\ date and commit

Data format

The results can be found in ~/data as JSON Files. Each file contains information of the first 40 profiles as seen on the first two result pages of the respective query. The information we processed was

• duration of job experiences
• duration of education
• sex

File Names

SHAano ## start#-end#

1. anonymized Dataset (Names, hyperlinks and pictures belonging to a profile have been removed or replaced with a hash value.
2. ordered number of query as in the list below
3. e.g. 1001-1040 result number

Sample profile Structure

{
  "category":
  "dominantSexXing":
  "profiles": [
    {
      "profile": [
        {
          "sex":
          "memberSince_Hits":
          "currJobDescr":
          "jobs": [
            {
              "jobTitle":
              "company":
              "company_url":
              "jobDuration":
              "jobDates":
            },
          ]
        }
      ],
      "languages": [
        {
        }
      ],
      "education": [
        {
          "institution":
          "url":
          "degree":
          "eduDuration":
        }
      ]
      "awards": [
        {
        }
      ]
    }

Queries

The following queries were used with reference to these statistics 1, 2 targeting a diversified collection of specific job titles in the respective career field while excluding jobs underrepresented on XING such as construction worker, farmer, etc. The order of the queries represents the order in the file naming convention

1. Administrative Assistant
2. Auditing Clerk
3. Auditor
4. accountant
5. bank teller
6. treasurer
7. actuary
8. budget analyst
9. economist
10. mathematician
11. statistician
12. Events Coordinator
13. Office Manager
14. Secretary
15. Dental Assistant
16. Medical Assistant
17. Receptionist
18. Audiologist
19. Daycare
20. lawyer
21. legal advisor

---

22. Application Developer
23. Building Inspector
24. Application Support Analyst
25. Civil Engineer
26. Back end Developer
27. Chemical Engineer
28. Construction Engineer
29. Data Analyst
30. Contract Administrator
31. Database Administrator
32. Field Engineer
33. Front End Developer
34. Mechanical Engineer
35. Safety Manager
36. Software Engineer
37. Superintendent
38. System Administrator

---

39. Technical Support Specialist
40. Account Coordinator
41. Account Executive
42. Advertising Director
43. Art Director
44. Brand Assistant*
45. Brand Manager*
46. Brand Strategist*
47. Copywriter
48. creative director
49. Internet Marketing Coordinator
50. Market Research Analyst
51. Marketing Associate
52. Online Product Manager
53. Public Relations Representative
54. Public Relations Specialist
55. SEO Manager
56. Social Media Marketing Coordinator
57. Architect

Please note that Brand is also a Family name in Germany

---

About The results

• searches have been performed in English without logging in to ensure that the results sorting is not tailored to a specific profile After generating the results, each profile has been parsed in full detail (while logged in).
• The sex of a person was manually derived from the profile name and picture since it is not given on the profile. This helped us filter irrelevant information such as fake profiles or profiles with misleading information (e.g. containing details about a company instead of a person).
• 19 queries returned duplicate entries. In most cases these would show one position apart. In such cases the latter was removed, resulting in a few results to include less than 40 profiles. Details such as company or institution name were anonymized using SHA-256 to only be able to differenciate between people who worked or studied at the same place or find other patterns.
• currJob is always equal to first element in pastJobs
• If a profile was found to be employed or studying at the time the data was collected, we replaced the date.
• profiles with incomplete data, in particular with missing dates have been considered as such: //\ add it to code instead? * If a job or education entry has no name it counts for an average of 3 months

Code

The code in src/ reads the information from all JSON files into a python dataframe that can be used later on. Currently it is simply dumped to disk. To use it, you can execute these commands:

//\

Citation

If you use this dataset, please cite:

Zehlike, Meike, Francesco Bonchi, Carlos Castillo, Sara Hajian, Mohamed Megahed, and Ricardo Baeza-Yates. "Fa* ir: A fair top-k ranking algorithm." In Proceedings of the 2017 ACM on Conference on Information and Knowledge Management, pp. 1569-1578. ACM, 2017.

BibTeX Entry:

@inproceedings{zehlike2017fair,
  title={Fa* ir: A fair top-k ranking algorithm},
  author={Zehlike, Meike and Bonchi, Francesco and Castillo, Carlos and Hajian, Sara and Megahed, Mohamed and Baeza-Yates, Ricardo},
  booktitle={Proceedings of the 2017 ACM on Conference on Information and Knowledge Management},
  pages={1569--1578},
  year={2017},
  organization={ACM}
}

The authors are not associated to XING in any way.