Investigating the Impact of Gender on Rank in Resume Search Engines

Chen, Le; Ma, Ruijun; Hannák, Anikó; Wilson, Christo

doi:10.1145/3173574.3174225

Cited by 129 publications

(85 citation statements)

References 62 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Such AI-informed decisions can thus lead to unfair treatment of certain groups. For example, in Natural Language Processing (NLP), résumé search engines can produce rankings that disadvantage some candidates, when these ranking algorithms take demographic features into account (directly or indirectly) (Chen et al, 2018), while abusive online language detection systems have been observed to produce false positives on terms associated with minorities and women (Dixon et al, 2018;Park et al, 2018). Another example where bias (specifically gender bias) can be harmful is in personal pronoun coreference resolution, where systems carry the risk of relying on societal stereotypes present in the training data (Webster et al, 2018).…”

Section: Introductionmentioning

confidence: 99%

Measuring Gender Bias in Word Embeddings across Domains and Discovering New Gender Bias Word Categories

Chaloner¹,

Maldonado²

2019

Proceedings of the First Workshop on Gender Bias in Natural Language Processing

View full text Add to dashboard Cite

Prior work has shown that word embeddings capture human stereotypes, including gender bias. However, there is a lack of studies testing the presence of specific gender bias categories in word embeddings across diverse domains. This paper aims to fill this gap by applying the WEAT bias detection method to four sets of word embeddings trained on corpora from four different domains: news, social networking, biomedical and a gender-balanced corpus extracted from Wikipedia (GAP). We find that some domains are definitely more prone to gender bias than others, and that the categories of gender bias present also vary for each set of word embeddings. We detect some gender bias in GAP. We also propose a simple but novel method for discovering new bias categories by clustering word embeddings. We validate this method through WEAT's hypothesis testing mechanism and find it useful for expanding the relatively small set of wellknown gender bias word categories commonly used in the literature.

show abstract

Section: Introductionmentioning

confidence: 99%

Measuring Gender Bias in Word Embeddings across Domains and Discovering New Gender Bias Word Categories

Chaloner¹,

Maldonado²

2019

Proceedings of the First Workshop on Gender Bias in Natural Language Processing

View full text Add to dashboard Cite

show abstract

“…For example, Kay et al only looked at simple metrics like average representation that fail to take order effects into account [34]. Other audits have used group representation in top K ranks [24], logarithmic discounting [9,23,51] and linear normalization by rank [38,46] to model the decay of attention. In this work, we argue that these ad hoc methods do not accurately model users' attention, and may lead to incorrect conclusions about (un)fairness of IR systems.…”

Section: Auditing Search Enginesmentioning

confidence: 99%

Quantifying the Impact of User Attentionon Fair Group Representation in Ranked Lists

Sapieżyński

Zeng

Robertson

et al. 2019

Companion Proceedings of the 2019 World Wide Web Conference

Self Cite

View full text Add to dashboard Cite

In this work, we introduce a novel metric for auditing group fairness in ranked lists. Our approach offers two benefits compared to the state of the art. First, we offer a blueprint for modeling of user attention. Rather than assuming a logarithmic loss in importance as a function of the rank, we can account for varying user behaviors through parametrization. For example, we expect a user to see more items during a viewing of a social media feed than when they inspect the results list of a single web search query. Second, we allow non-binary protected attributes to enable investigating inherently continuous attributes (e.g., political alignment on the liberal to conservative spectrum) as well as to facilitate measurements across aggregated sets of search results, rather than separately for each result list. By combining these two elements into our metric, we are able to better address the human factors inherent in this problem. We measure the whole sociotechnical system, consisting of a ranking algorithm and individuals using it, instead of exclusively focusing on the ranking algorithm. Finally, we use our metric to perform three simulated fairness audits. We show that determining fairness of a ranked output necessitates knowledge (or a model) of the end-users of the particular service. Depending on their attention distribution function, a fixed ranking of results can appear biased both in favor and against a protected group. CCS CONCEPTS• Information systems → Page and site ranking; Content ranking; • Human-centered computing → User interface design.KEYWORDS information retrieval; group fairness; ranked lists 1 We use "top" and "high" to refer to the numerically lowest ranks in lists, e.g., rank one, in keeping with the norms of the IR literature [10,27].

show abstract

“…While a small stream of CSCW and HCI studies have offered important insights about job recruiting [11,33], there has been a lack of attention to the interviewing practices of evaluators. Job interviews play a critical role in the hiring process of many labor markets.…”

Section: Cscw and Hci Literature On Hiringmentioning

confidence: 99%

Are You One of Us?

Chua

Mazmanian

2020

Proc. ACM Hum.-Comput. Interact.

View full text Add to dashboard Cite

To better understand how we can broaden participation in computing, this exploratory study examines the interview process for elite internships at established technology companies. Through conducting 36 interviews with evaluators at technology companies, we find that in addition to technical competence, evaluators often assess internship applicants based on explicit and implicit signals of industrial fit, organizational fit, and individual fit. These evaluative criteria are reminiscent of prior literature linking biases in hiring to social class background. By reflecting on how our findings relate to previous studies, we suggest that evaluators' assessments of fit are potentially linked to social class background, and this might be an invisible factor contributing to hiring biases at technology companies. Given that hiring only culturally similar employees can have negative individual, organizational, and societal consequences, we propose strategies for evaluators to broaden their evaluation perspectives and to enact inclusive interviewing practices. We conclude with a call for further research on the role of social class background in the hiring process. CCS Concepts: • Social and professional topics → Cultural characteristics, Employment issues.

show abstract

Investigating the Impact of Gender on Rank in Resume Search Engines

Cited by 129 publications

References 62 publications

Measuring Gender Bias in Word Embeddings across Domains and Discovering New Gender Bias Word Categories

Measuring Gender Bias in Word Embeddings across Domains and Discovering New Gender Bias Word Categories

Quantifying the Impact of User Attentionon Fair Group Representation in Ranked Lists

Are You One of Us?

Contact Info

Product

Resources

About