JOB DESCRIPTION

Hybrid Intelligence (HI) is the combination of human and artificial intelligence (AI), augmenting human intellect instead of replacing it, and developing AI which works with and for humans.
Hybrid Intelligent AI systems should actively calibrate the trust they elicit from humans they interact with by actively identifying cases of over/under-trust, and then acting to address these. To do this, they require an understanding of their human partners’ trust perceptions and how these dynamically evolve. Currently, trust perceptions are typically measured using explicit assessments, such as questionnaires at static moment in time. However, such measures lack efficiency in providing feedback for trust calibration in human-AI interactions, and effectiveness because they struggle to capture fine-grained information about how human trust perceptions dynamically evolve.
In this project, you will be working to address this issue by developing approaches using multimodal sensor data for the implicit assessment of human trust perceptions. In particular, you will work on answering the following research question: How can we enable artificial intelligence systems to efficiently and effectively assess the dynamic development of humans’ trust evaluations during collaboration for improved calibration behavior?
Given the recent advances in multimodal language modeling (e.g., MLLMs), this project will explore approaches that leverage unstructured language data describing elements of human trust perceptions and associated reasoning processes during interactions captured with a “Think Aloud” (TA) protocol. The project will involve collecting multimodal data about trust behavior; identifying data collection approaches for effective and efficient modeling of trust perceptions and calibration; and addressing a cycle of trust behavior, from understanding human trust, to adjusting agent behavior, which in turn influences human trust.
This project will be supervised by dr. Bernd Dudzik and dr. Myrthe Tielman. That means the candidate will be a part of both the research group of Pattern Recognition and Interactive Intelligence, both within the Intelligent Systems section of Computer Science, EEMCS. Additional co-supervisors will be Prof. Dr. Mark Neerincx (II, TU Delft & TNO) and Prof. Dr. Dan Balliet (VU Amsterdam).
This project is part of a larger research effort within the Hybrid Intelligence Center, where you will have the opportunity to collaborate with researchers across multiple universities and disciplines.

JOB REQUIREMENTS

We are looking for a candidate who meets the following essential criteria:

• A Master’s degree or equivalent (or about to graduate with one) in a relevant field (Artificial Intelligence, Computer Science, Data Science, Cognitive Science, etc.)
• A good command of spoken & written English
• A curiosity-driven mindset and a willingness to learn
• Some experience with ML or NLP
• Some experience in collecting multimodal datasets or running experiments with human subjects

We encourage you to apply even if you do not meet all the criteria above as long as you are willing to acquire the relevant skills.

Additionally, the following criteria are appreciated:

• Experience with interdisciplinary research projects.
• Strong analytical and conceptual modelling competencies
• Good programming skills (preferably Python), including ML methods and libraries.

Please refer the Job description for details