Researcher

Philipp Müller

Philipp Müller is a researcher in multimodal AI who analyses human social behaviour, from eye contact and backchannels to engagement. After studying psychology and computer science at Saarland University, he completed a PhD at the Max Planck Institute for Informatics in 2020 with a thesis on sensing and anticipating social behaviour. In 2021, he began organising the yearly MultiMediate challenge at ACM Multimedia, whose editions have covered eye contact detection, backchannel detection, bodily behaviour recognition and cross-cultural engagement estimation. By 2025, he was a Senior Researcher at DFKI in Saarbrücken, and since November 2025, he leads the Embodied Social Interaction group at the Max Planck Institute for Intelligent Systems in Stuttgart, which combines psychology with multimodal AI.

Timeline

  1. 2016–2020

    PhD student in Collaborative AI

    He earns a Bachelor's degree in Psychology and Bachelor's and Master's degrees in Computer Science at Saarland University, and spends six months as a Visiting Student Researcher at Stanford University during his master studies. From 2016 to 2020, he is a PhD student in the Collaborative Artificial Intelligence group.

  2. 2018

    Detecting low rapport in groups

    In 2018, he and coauthors Michael Xuelin Huang and Andreas Bulling are first to investigate automatic prediction of low rapport during natural small-group interactions. Using facial features, their method detects low rapport with an average precision of 0.7 against a chance level of 0.25.

  3. 2020

    A thesis on social behaviour

    He completes his PhD at the Max Planck Institute for Informatics with the thesis "Sensing, Interpreting, and Anticipating Human Social Behaviour in the Real World" for Saarland University in 2020. Also in 2020, with Ekta Sood and Andreas Bulling, he presents the first method to anticipate averted gaze in natural dyadic interactions, a multimodal LSTM network that analyses non-verbal facial cues and speaking behaviour.

  4. 2021

    Launching MultiMediate

    In 2021, he starts organising MultiMediate, a yearly ACM Multimedia Grand Challenge on automatic social behaviour analysis, from DFKI's Cognitive Assistants department in Saarbrücken. The first edition poses eye contact detection and next speaker prediction in group discussions, using the MPIIGroupInteraction dataset of three- to four-person conversations.

  5. 2022–2023

    Backchannels and body language

    In 2022, he is first author of MultiMediate'22, which annotates 7234 backchannel instances in the MPIIGroupInteraction dataset for backchannel detection and agreement estimation, and he coauthors BBSI, annotations of 15 body language classes across 26 hours of group interaction. In 2023, MultiMediate'23 adds engagement estimation on the NOXI database and bodily behaviour recognition annotated with the BBSI scheme.

  6. 2024–2025

    Engagement across cultures

    In 2024, he is first author of MultiMediate'24, the first challenge on multi-domain engagement estimation, which trains on the NOXI database and tests on new languages and on MPIIGroupInteraction group discussions. In 2025, MultiMediate'25 turns to cross-cultural engagement estimation and adds NOXI-J, a corpus of Japanese and Chinese interactions. By 2025, he is a Senior Researcher at DFKI in Saarbrücken.

  7. Since November 2025

    Leading Embodied Social Interaction

    In November 2025, he starts the Cyber Valley Max Planck Independent Research Group Embodied Social Interaction at the Max Planck Institute for Intelligent Systems in Stuttgart. The group works at the intersection of multimodal AI, psychology and robotics, and he applies its methods in psychiatric contexts, supporting clinical assessment and the evaluation of patient–therapist relationships. MultiMediate'26, which he organises in 2026, adds child-child and child-robot play interactions from the PInSoRo dataset.

Career

  1. Since 2025Max Planck Institute for Intelligent SystemsGroup Leader
  2. Stanford UniversityVisiting Student Researcher
  3. DFKISenior Researcher

Education

  1. 2020Saarland UniversityBachelor's and Master's degrees
  2. 2020Max Planck Institute for InformaticsPhD student

Sources and credits

Photo: We have no published photo of Philipp Müller yet. Send us a free-licensed photo.

  1. Philipp Müller | Collaborative Artificial Intelligence collaborative-ai.org. Checked 22 October 2026.
  2. Philipp Müller - Max Planck Institute for Informatics mpi-inf.mpg.de. Checked 22 October 2026.
  3. Past Challenges | MultiMediate: Multi-modal Behaviour Analysis for Artificial Mediation multimediate-challenge.org. Checked 22 October 2026.
  4. Philipp Müller – SCCAI 2025 sccai2025.gtu.ge. Checked 22 October 2026.
  5. Bodily Behaviors in Social Interaction: Novel Annotations and State-of-the-Art Evaluation arxiv.org. Checked 22 October 2026.
  6. MultiMediate'23: Engagement Estimation and Bodily Behaviour Recognition in Social Interactions arxiv.org. Checked 22 October 2026.
  7. Cyber Valley continues to grow | Cyber Valley cyber-valley.de. Checked 22 October 2026.
  8. Distinguished Lecture Series: Philipp Müller | Event | Institute for Stochastics and Applications | University of Stuttgart isa.uni-stuttgart.de. Checked 22 October 2026.
  9. People | MultiMediate: Multi-modal Behaviour Analysis for Artificial Mediation multimediate-challenge.org. Checked 22 October 2026.
  10. Overview | MultiMediate: Multi-modal Behaviour Analysis for Artificial Mediation multimediate-challenge.org. Checked 22 October 2026.