Affective Language Patterns in Text-Based Communication Predict Relational Outcomes in Online-Mentoring for Girls in STEM

• In Event: AERA Roundtable Session Wednesday 11:45 am
  In Roundtable Session: Mentoring, Training, Measurement & Technology

Abstract

Online mentoring is a promising measure to increase girls’ interest in STEM and support their learning. Understanding emotional tone in this setting is crucial, especially when identifying early signs of disengagement preceding premature closure of mentoring relationships. This study investigated variation in emotional tone in text-based communication within a one-year online mentoring program for girls (N=827, M=13.77 years, SD=1.96 years) in STEM, comparing 5,017 messages from dropouts with 6,556 messages from non-dropouts. Linguistic predictors for premature closure were derived by combining natural language processing based on large language models with classical text analysis utilizing validated word dictionaries. Findings reveal distinctive patterns in mentees’ emotional and STEM-related communication with mentors versus peers and demonstrate its predictive value for premature match closure.

Authors

• Claudia Uebler, University of Regensburg

• Albert Ziegler, Friedrich-Alexander-University of Erlangen-Nürnberg

• Heidrun Stoeger, University of Regensburg