Letters of recommendation (LORs) are an important and widely used evaluation criterion for hiring, university admissions, and many other domains. Prior work has identified that gender stereotypes can bias how recommenders describe female applicants compared to male applicants in contexts such as faculty positions and undergraduate research internships. For example, female applicants are more likely to be described using communal adjectives (e.g., affectionate, warm) while male applicants are more likely to be described using agentic adjectives (e.g., confident, intellectual). In this paper, we investigate (i) the extent to which these differences in language affect readers’ impression of applicant competitiveness and (ii) the efficacy of a mitigation strategy: visual highlighting. Our findings suggest that simple changes in visual salience through highlighting language more commonly used to describe women can negatively affect readers’ evaluation of candidates, while highlighting the language more commonly used to describe both men and women can reduce the effects of the bias.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Visual Salience to Mitigate Gender Bias in Recommendation Letters

  • Yanan Da,
  • Mengyu Chen,
  • Ben Altschuler,
  • Yutong Bu,
  • Emily Wall

摘要

Letters of recommendation (LORs) are an important and widely used evaluation criterion for hiring, university admissions, and many other domains. Prior work has identified that gender stereotypes can bias how recommenders describe female applicants compared to male applicants in contexts such as faculty positions and undergraduate research internships. For example, female applicants are more likely to be described using communal adjectives (e.g., affectionate, warm) while male applicants are more likely to be described using agentic adjectives (e.g., confident, intellectual). In this paper, we investigate (i) the extent to which these differences in language affect readers’ impression of applicant competitiveness and (ii) the efficacy of a mitigation strategy: visual highlighting. Our findings suggest that simple changes in visual salience through highlighting language more commonly used to describe women can negatively affect readers’ evaluation of candidates, while highlighting the language more commonly used to describe both men and women can reduce the effects of the bias.