Mentorship plays a crucial role in career advancement and organizational continuity, particularly in academia where junior scientists benefit from guidance from senior colleagues. Traditional mentorship models often center on formal supervisory relationships (e.g., thesis advisors). However, this study expands the definition to include informal mentorship from multiple senior collaborators who may not have formal supervisory roles. This is especially relevant in the context of team-based research collaborations prevalent in modern academia. Prior studies on academic mentorship have limitations such as focusing solely on thesis advisors, relying on self-reported data, and limiting their scope to single disciplines. This research leverages the rich dataset of academic publications to study mentorship more comprehensively. By analyzing millions of collaborations from multiple disciplines, the researchers overcome these limitations and provide a more robust investigation into the relationship between mentorship quality and the long-term success of junior scientists. The study also contributes to the ongoing discussion on gender equity in science by examining the mentorship experiences of both female and male scientists, considering the gender of both the mentor and the protégé.

The authors review existing literature on the benefits of mentorship, highlighting its impact on career progression and organizational cohesion. They discuss how mentorship can act as an equalizing force, particularly for underrepresented minorities, providing role models and access to networks. The paper then contrasts its approach with prior studies on mentorship in academia, noting its advantages in terms of the breadth of mentorship considered (beyond thesis advisors), the use of objective scientific impact data, the inclusion of multiple disciplines, and the scale of analysis. The paper also differentiates itself from related work, such as Li et al.'s focus on coauthorship with top scientists regardless of mentoring role, and Ma et al.'s study of formal mentorship and its potential inverse correlation with protégé success post-mentorship. This study uniquely focuses on informal mentorship, post-mentorship impact independent of the mentor's involvement, and the gender dynamics of mentorship.

This study utilized the Microsoft Academic Graph (MAG) dataset, containing detailed information on 215 million scientists and 222 million papers. The researchers addressed the name disambiguation problem and used external data sources to establish gender and affiliation rank. Junior and senior scientists were defined based on their academic age (years since first publication), with the first seven years classified as junior years. Mentor-protégé pairs were identified based on coauthorship criteria, including shared discipline, US-based affiliation, and a limited number of coauthors. This resulted in a sample of 3 million unique mentor-protégé pairs across ten disciplines and over a century of research. A survey of 167 randomly selected protégés confirmed the mentoring aspect of the identified relationships, showing high percentages of protégés reporting having received various forms of advice and support from their senior collaborators. Mentorship quality was assessed using two measures: 'big-shot experience' (average citations per annum of mentors prior to mentorship) and 'hub experience' (average degree of mentors in the collaboration network prior to mentorship). The mentorship outcome was measured as the average citations of protégés' publications during their senior years (after seven years of their career) excluding any papers coauthored with their mentors. Coarsened Exact Matching (CEM) was employed to compare groups of protégés with varying levels of mentorship quality, controlling for several factors including the number of mentors, publication year, discipline, gender, and affiliation rank. To investigate gender dynamics, the researchers compared outcomes based on different numbers of female mentors and contrasted the mentors' gains (citations from papers coauthored with protégés) based on the gender of both the mentor and the protégé.

The analysis revealed significant associations between mentorship quality and protégé post-mentorship impact. Higher 'big-shot experience' was strongly associated with increased protégé impact (up to 35%), while 'hub experience' showed a weaker association (up to 13%). The 'big-shot experience' proved to be a more significant predictor than 'hub experience', suggesting the scientific impact of mentors is more critical than their network size. This association remained robust across various factors such as discipline, affiliation rank, number of mentors, mentor age, and protégé gender. Importantly, the study found that increasing the proportion of female mentors was associated with a decrease in the post-mentorship impact of female protégés (up to 35%). Furthermore, female mentors experienced a significant reduction in their gains (18% fewer citations) when mentoring female protégés compared to male protégés. Male mentors' gains did not show a significant difference based on the protégé's gender. These findings suggest that, counterintuitively, female protégés might benefit more from male mentors.

The findings highlight the complexity of mentorship dynamics in academia and challenge the assumptions underlying current diversity policies. While same-gender mentorships are often promoted to retain women in science, the results indicate that opposite-gender mentorships might lead to greater success for female protégés. This calls for a reassessment of diversity policies and a consideration of both first and second-order consequences. Several factors could contribute to these observations, including potential time constraints on female mentors due to committee work or differences in research topics chosen by female mentors and their protégés. The study also underscores the mutual benefits of mentoring, as mentors' gains are influenced by the gender of the protégé. However, the paper acknowledges the limitations of observational data in fully explaining these observed patterns.

This study makes significant contributions to our understanding of academic mentorship by examining informal mentorship across various disciplines and using large-scale data analysis to assess scientific impact. The findings challenge existing assumptions about the benefits of same-gender mentorship for women in science. The results underscore the importance of a nuanced approach to diversity policies that account for the complex interplay between mentor and protégé gender. Future research should explore potential mediating factors and mechanisms that drive these observed relationships, such as differences in mentorship style, research topic selection, and access to resources between male and female mentors.

The study's findings are based on observational data and cannot establish causal relationships. While the survey helped verify mentoring relationships, self-reporting biases could still exist. The focus on US-based affiliations limits the generalizability of the findings to other contexts. The specific mechanisms behind the gender-related findings require further investigation.