When one teaches, two learn: the bidirectional learning process between supervisors and PhD students

In contexts involving supervisors and students, knowledge transfer is commonly assumed from the former to the latter. This paper challenges that assumption by investigating the bidirectional knowledge transfer between PhD students and their supervisors. We consider 40,852 PhD students who graduated in the STEM fields in France between 2010 and 2018. Focusing on Artificial Intelligence (AI) knowledge transfer, we find that a student supervised by a supervisor with AI knowledge is 10 percentage points more likely to write a thesis in AI than a student supervised by a supervisor without AI knowledge, indicating an AI knowledge transfer from supervisors to students. We also find that a supervisor with no prior AI knowledge but exposed to a student with AI knowledge is 15 percentage points more likely to publish an article with AI content than a supervisor exposed to a student without AI knowledge, indicating an AI knowledge transfer from students to supervisors. These results confirm the bidirectionality of the learning process.