About the job
Join Our Team
At Doctolib, we are on a mission to revolutionize healthcare access and enhance the quality of life for healthcare professionals. As a part of our commitment to responsible artificial intelligence, we are seeking a Master’s level intern to assess biases in a pediatric dialogue system by analyzing interactions between AI and patients.
Collaborating with Doctolib's data scientists and researchers from LISN (Université Paris-Saclay), you will help ensure fairness and reliability in our AI health solutions. One of the key objectives of this internship is to create an annotated and anonymized dataset that will be published as open-source. This five-month internship will allow you to work on a project with significant societal impact, at the intersection of natural language processing, health, and AI ethics.
Your responsibilities will include:
- Conducting a literature review on bias assessment methods in health-related language models
- Analyzing a corpus of anonymized dialogues between the AI system and pediatric patients/parents to identify potential biases
- Defining and implementing an annotation scheme to characterize identified biases
- Building an evaluation framework based on demographic characteristic disaggregation
- Applying this framework to the dialogue system and comparing results with other autoregressive models

