The Social Cognitive Systems group uses a middleware for inter-process and inter-agent communication that is based on principles of incremental updating: IPAACA, the incremental processing architecture for artificial conversational agents and robots. IPAACA exists for C++, Python and Java and is currently communicating using the MQTT protocol.
For six years in a row, Rust has been voted the most loved programming language by the Stackoverflow developer community since 2015. Its verbose and helpful compiler, safety guarantees and fearless concurrency are some of the features that make this modern systems language so appealing. We are looking for a student who likes the challenge of reimplementing IPAACA, while learning Rust in the process.
Contact: Sebastian Kahl

The task is to extend or redo the existing web-based visualization prototype. Additionally, the existing PRIMO backend can be improved or extended.
Contact: Jan Pöppel

NWI Master – Summer 2017–Winter 2017/18
A student created an interactive visualization for our PRobabilistic Inference MOdules (PRIMO) using d3.js allowing one to create, view and modify (Dynamic) Bayesian Networks in a browser and perform (conditional) inferences in an intuitive interface.
Supervised by Jan Pöppel

Psychology Bachelor – Winter 2017/18
An interlocutor's eye gaze can be a very strong cue in interactions. Virtual agents can use their eye gaze alongside a speech utterance to improve the user's understanding. In this project, we investigated a) the precision and accuracy of humans perceiving a virtual agent's gaze in virtual reality, and b) the unique contribution of gaze by comparing cases of redundant, additional and conflicting eye gaze and speech.
Supervised by Sebastian Loth

NWI Master – Winter 2017/18
The project developed a simple Bayesian model to make predictions and recommendations for placing events in users' calendars (date and time), based on the their event history.
Supervised by Hendrik Buschmeier

ISY Master – Winter 2014/15
The project aimed at developing an approach for recognising a speaker's mental state (agreement, disagreement, uncertainty) based on verbal/vocal characteristics of their spoken feedback signals (‘ja’, ‘nein’, ‘okay’, ‘m’, ‘mm’, ‘hm’, ‘mhm’, etc.). The project developed an online crowd-sourcing platform for data collection and a signal processing and classification pipeline that used PocketSphinx, openSMILE, and Dlib.
Supervised by Hendrik Buschmeier