Supervisor: Andrew Yates, University of Amsterdam
Co-supervisor: Piek Vossen, Vrije Universiteit Amsterdam

Short description:
We will support the investigation step of the scientific method cycle through a scientific assistant that helps identify relevant literature (at the document level) and helps pinpoint relevant information within this literature. 

Supervisor: Annette ten Teije, Vrije Universiteit Amsterdam
Co-supervisor: Herke van Hoof, University of Amsterdam

Short description:  
We will support the investigation step of the scientific method cycle through a scientific assistant that helps identify relevant literature (at the document level) and helps pinpoint relevant information within this literature. 

Supervisor: Antske Fokkens, Vrije Universiteit van Amsterdam
Co-supervisor: Shihan Wang, Utrecht University

Short description:  
In linguistic tasks, not all errors are equal. Which errors are more important than others changes according to how a model will be used. We will investigate how to use reinforcement learning to guide models to a specific behavior in computational linguistic tasks, prioritizing some forms of correctness over others.  Shihan Wang (UU)

Supervisor: Catha Oertel, Technical University Delft
Co-supervisor: Piek Vossen, Vrije Universiteit Amsterdam

Short description:
Successful communication within a multi-party meeting depends on the creation of a common ground between interaction partners: what is relevant for oneself, what is relevant for the interaction partner(s). We focus on memory creation based on affective responses to conversational instances and their relation to real-world objects supporting the interaction. 

Supervisor: Joost Broekens & Aske Plaat, University of Leiden
Co-supervisor: Kim Baraka, Vrije Universiteit Amsterdam

Short description: 
How can we use affective signals grounded in the learning process of an RL agent as feedback to a human teacher to close the teacher learner loop? Emotion simulation and means of expression on robots and agents studied in the context of human-agent interaction. 

Supervisor: Daniel Balliet, Vrije Universiteit Amsterdam
Co-supervisor: vacancy

Short description:
CoDa is the first databank to represent an entire field of studies in the social sciences (human cooperation) in a machine-readable way which can be searched to select studies for on-demand meta-analysis. We will generalize CoDa to factors from economics (e.g., altruism, trust), sociology (e.g., social capital), and clinical psychology (i.e., treatments of depression), we will develop queries that can produce on-demand meta-analyses, and develop an interactive technique to create queries and update meta-analyses (e.g., text of the manuscript).  

Supervisor: Davide Grossi, University of Groningen
Co-supervisor: Pradeep Murukannaiah, Technical University Delft

Short description: 
We will develop a methodology for supporting participatory democratic decision making online. The methodology should support (1) users in participating in online democratic deliberation and reflecting on the alignment between their choices and values, (2) policy makers in aggregating users’ choices and opinions in a principled and transparent manner. 

Supervisor: Erik Bekkers, University of Amsterdam
Co-supervisor: Daniel Pelt, University of Leiden

Short description: 
We will develop techniques for collaborative human-computer image annotation of training sets for deep learning tasks. Our techniques will suggest relevant annotations to the human annotator, will indicate inconsistencies in the human annotations, and will use concepts from geometric deep learning to handle shapes of image annotations.

Supervisor: Frans A. Oliehoek, Technical University Delft
Co-supervisor: Jan-Willem vd Meent, University of Amsterdam

Short description: 
We will investigate how Bayesian methods can be used in multi-agent reinforcement learning (MARL) as a way to infer models of human partners, in order to collaborate with them more effectively.

Supervisor: Hayley Hung, Technical University Delft
Co-supervisor: Dan Balliet, Vrije Universiteit Amsterdam

Short description: 
We know that there are critical moments in social interactions that determine whether to select someone as a collaborative partner. The goal of this project is to bridge the artificial perception and action loop by considering the idea of building a HI Social Guru. This is carried out through (i) goal setting, (ii) enlightenment and (iii) empowerment via the HI Social Guru. The important thing about a Guru is that they assist humans in self-reflection in social settings without needing to interfere directly (synchronously) with the interaction. That is, they provide wisdom and enlightenment asynchronously from the interaction.