Hiring Algorithms and The Gender Gap

Reis, C., Adjerid, I.

Preparing for Journal Submission (Target: Management Science)

Governments have introduced regulations that require organizations to disclose when using hiring algorithms. We test the impact of such disclosure on the gender gap in applicant pools for male-dominated jobs. Across two longitudinal field experiments, with over 3500 participants, we found that men show higher persistence and opportunism when encountering hiring algorithms, thereby widening the gender gap. A phenomenon we term the “overflowing” pipeline problem.

​​

(A)I can! How AI-Based Performance Appraisals Increase Self-Advocating Behaviors for Women Through Psychological Safety

Thompson, P., Reis, C., Raveendhran, R., Gordon, M.

Preparing for Journal Submission (Target: Organization Science)

Research reports that women are less likely than men to ask for a promotion or raise, especially if they anticipate biases in how their performance will be rated. Grounded on the psychological safety framework, we show that that the integration of artificial intelligence (AI) into the performance appraisal process can make employees (especially women) feel more psychologically safe and, in turn, increase their self-advocation behaviors. These findings were corroborated across 3 studies.

​​

Me, Misinformation, and (A)I: The Role of Magic, Myths, and Metaphors

Reis, C.*, Brown, N.* (* indicates equal contribution)

Preparing for Journal Submission (Target: MIS Quarterly)

Misinformation is pervasive online and is considered one of the main threats to society. Drawing on the magic, myths, and metaphors (3Ms) theory, we reveal that people are more willing to believe chatbot hallucinations than fake news propagated on social media. This is due to people's beliefs of 3Ms surrounding generative AI tools. These results were corroborated across 3 studies.​​

​

Bias Encoded: Human Raters and Algorithmic Content Moderation

Reis, C., Brown, N., Seref, O., Wang, A.

Data Collection

Companies often outsource a large share of the content moderation work to workers in foreign countries. Through a mixed-methods approach (qual + quant), we examine how such practices impact content moderation outcomes, particularly when such decision-making is used as training data for large language models (LLMs) responsible for moderating online information.​​