Large Language Models as Online Mediators


As the digital landscape continues to grow, the need for effective online communication and mediation becomes increasingly crucial. Language models, especially large-scale models, have shown great potential in understanding and generating human-like text. To harness the power of language models and explore their application in online mediation, we propose a 6-month internship for a highly motivated Master’s student. This internship aims to develop large language models as online mediators, facilitating smoother and more efficient online interactions.


The primary objective of this internship is to explore the development and deployment of large language models for online mediation purposes. By leveraging the capabilities of these models, the intern will contribute to the advancement of digital platforms and tools that promote meaningful and respectful communication, while minimizing misunderstandings, polarisation, and conflicts.

Internship Tasks and Responsibilities

  1. Literature Review: Conduct a review of existing research and studies related to large language models and online mediation.
  2. Development: Develop an application that intervenes in discussions to mediate it if it goes off track or to enhance it to prevent it from going awry.
  3. Evaluation: Propose human and automatic metrics to assess the quality and utility of the model(s)


Ideas that will be explored include, but are not limited to:

  1. Dataset creation
    • Synthesis of online discussions going off-track.
  2. Development
    • Designing targeted interventions with LLMs.
  3. Discussion enhancement
    • Providing feedback after a comment is written but before it is posted as a way to give a second chance for its author to modify it.
    • Assist users by suggesting rephrasing of opinions poorly expressed.
    • Provide arguments (or counter-arguments) that can help maintain civil, fair, and fruitful discussions.
  4. Evaluation
    • Assessing the effectiveness of the methods proposed


  • Familiar with Python
  • Creativity, initiative and proactive spirit
  • Knowledge of Linux and related tools


  • Experience in Machine Learning and Deep Learning
  • Experience in Natural Language Processing
  • Experience in PyTorch and HuggingFace librairies

Send me your transcripts and CV: [email protected]