Kathrin Janowski M.Sc.
Links
Research Interests
Virtual Agents, Social Robots, Human-Computer Interaction, Human-Robot Interaction, Multimodal User Interfaces
My research focuses on the development of multimodal dialog systems for intuitive communication between humans and computer-controlled social agents.
Among other things, this includes the generation of believable, life-like behavior for virtual characters and robots which allows the user to understand their cognitive state. Furthermore, I am interested in modern input technology like speech, facial expressions and gestures because those agents, in turn, should be able to understand naturally-occurring human behavior.
Projects
Decision-Theoretic Personality-Based Turn-Taking
For my doctoral thesis I am examining different factors which influence the turn-taking in dialogue, and how these mechanisms can be applied to artificial characters.
The focus is on behavior patterns which express the personality and the interpersonal stance towards the interlocutor.
Externally Funded Projects
Supervised Student Projects
Description
For this project I supervised two students who programmed an emotional game-playing companion. For this purpose two different personalities were modeled - a friendly partner and a hostile rival - who react differently to game events and express this verbally as well as using facial expressions.
This application was used in the context of the ForGenderCare project in order to examine how elderly people interact with social robots.
Publications
"Adaptive linguistic style for an assistive robotic health companion based on explicit human feedback"
Hannes Ritschel, Andreas Seiderer, Kathrin Janowski, Stefan Wagner, Elisabeth André
12th ACM International Conference on PErvasive Technologies Related to Assistive Environments (PETRA), 2019
Internal Chair Projects
Local Scientists
Description
This demonstration was created for the "Lange Nacht der Wissenschaft" ("The Long Night Of Science") which took place on May 5th 2018 in Augsburg.
Here the social robot Reeti serves as a nutrition advisor who informs the user about the calorie content of the offered drinks. By means of custom-made drink holders and an electronic scale the system can detect how much the user drinks of which juice. Based on the amount of calories in the cup a reward is calculated for the information strategy used beforehand, in order to adapt the probability that it is chosen in the future via reinforcement learning. This way the robot is supposed to learn whether the user is convinced more easily by pointing out high or low calories.
In the winter semester of 2018/19 this system was also used to illustrate potential applications of social robots in retirement homes. In the context oft the course "Multimedia 1: Usability Engineering" a group of students presented this recommendation robot in three different retirement homes in the region of Augsburg. Afterwards, they discussed potential uses and the requirements for this technology with the inhabitants and the care staff.
Publications
"Towards a robotic dietitian with adaptive linguistic style"
Hannes Ritschel, Kathrin Janowski, Andreas Seiderer, Elisabeth André
European Conference on Ambient Intelligence (AmI), 2019
"Drink-O-Mender: An Adaptive Robotic Drink Adviser"
Hannes Ritschel, Andreas Seiderer, Kathrin Janowski, Ilhan Aslan, Elisabeth André
3rd International Workshop on Multisensory Approaches to Human-Food Interaction, 20th International Conference on Multimodal Interaction (ICMI), 2018
Media Coverage
"Robi forscht und hilft. Studenten mit Roboter im Awo-Heim"
Aichacher Zeitung, January 12, 2019
"Künstliche Intelligenz - Sind Maschinen die besseren Menschen?"
"Stationen", BR, June 13, 2018
Local Scientists
M.Sc. Kathrin Janowski, Dr. Hannes Ritschel, Dr. Florian Lingenfelser, Dr. Johannes Wagner, M.Sc. Markus Häring, Prof. Dr. Birgit Lugrin
Description
The simplest form of empathy is mirroring the emotions which are being observed from the interlocutor. Humans usually do this subconsciously.
Our current prototype employs the Social Signal Interpretation Framework (SSI) to analyse the tone of voice and the facial expressions of the user. The recognized emotions are then mapped to numerical values along the two axes "pleasure" and "arousal".
The robot we use, "Zeno", is a RoboKind R50 built by the Texan company Hanson Robotics. He has a human-like body and additional servo motors under a synthetic skin which allow for the adaptation of his facial expressions. The appliation which controls his behavior continually adapts his facial expression and head pose to the emotional state which was recognized by the SSI pipeline.
Additionally, Zeno has several lists of sentences which he can use to comment on the situation when user pauses in their speech. The exact moment when the robot talks depends on various factors, for example the intensity of the current emotional state or the time which has passed since the last comment.
Media Coverage
"Roboter Zeno - Ist es bescheuert, Gefühle für eine Maschine zu haben?"
"Puls“, BR, March 24, 2016
"Extrem Robotern - Ein echt netter Typ"
Deutschlandfunk Nova, March 7, 2016
"Algorithmus der Gefühle - Menschliche Roboter"
SWR2, May 2, 2016
"Robo Sapiens - Können Roboter den Menschen ersetzen?"
"X:enius", arte, June 18, 2015
"Empathische Alice"
"Quarks & Co", WDR, November 12, 2013
Local Scientists
Dr. Gregor Mehlmann, M.Sc. Kathrin Janowski, M.Sc. Markus Häring
Description
For this application the human's task is to place different objects on the given fields. The robot's instructions for this are kept intentionally ambiguous in order to make the human ask for clarification.
The eye tracking glasses worn by the user, as well as special markers on the objects, enable the system to detect which object the user is currently looking at. This in turn enables the robot to resolve ambiguities in the user's spoken question, so it can answer correctly.
Furthermore, this information allows the robot to show gaze behavior which expresses joint attention. For example, it follows the user's gaze to the respective objects or makes eye contact when the former looks at the robot.
This eye contact also serves as a signal for conversational floor management. When the user asks a question, the robot delays its answer and waits until the user looks at it directly.
All these behavior patterns serve to establish the common ground in order to avoid misunderstandings or at least resolve them quickly.
Publications
"Modeling Grounding for Interactive Social Companions"
Gregor Mehlmann, Kathrin Janowski, Elisabeth André
KI - Künstliche Intelligenz, Special Issue on Companion Technologies, 2015
"Exploring a Model of Gaze for Grounding in Multimodal HRI"
Gregor Mehlmann, Kathrin Janowski, Markus Häring, Tobias Baur, Patrick Gebhard and Elisabeth André
16th International Conference on Multimodal Interaction (ICMI), 2014
"Modeling Gaze Mechanisms for Grounding in HRI"
Gregor Mehlmann, Kathrin Janowski, Tobias Baur, Markus Häring, Elisabeth André und Patrick Gebhard
21th European Conference on Artificial intelligence (ECAI), 2014
Teaching
Bachelor
Practical Course "Interaction Design and Engineering for Health Care Applications"
- WS 2020/21
- WS 2021/22
Multimedia Project
- WS 2019/20: "Multimodal Human-Robot-Dialogue with Aisoy"
- WS 2015/16: "The Always-On Clone - Social Activity Manager"
- WS 2014/15: "Wizard-Of-Oz Control for Robots"
Practical Module
- SS 2016: "Emotional Expressions for Reeti"
- SS 2015: "RoboGames - Social Games with Reeti"
Lecture "Multimedia Grundlagen II"
- SS 2015
Seminar "Fundamental Issues in Multimodal Dialogue and Interaction"
- SS 2015: "Verbesserung der Langzeitinteraktion von interaktiven Robotern und Menschen" („Improvement of long-term interaction between interactive robots and humans“)
- SS 2015: "Social Interaction with Game Playing Agents“
Supervised Student Theses
- SS 2015: "Entwurf und Implementierung eines Emotionsmoduls für einen sozialen Roboter" („Design and implementation of an emotion module for a social robot“)
- SS 2015: "Persönlichkeitsbasierte Verhaltensanpassung eines sozialen Roboters" („Personality-based behavior adaptation of a social robot“)
Master
Practical Course "Interaction Design and Engineering"
- WS 2020/21
Lecture "Multimedia 1: Usability Engineering"
- WS 2019/20: "Human-Robot-Interaction - Empathic Smalltalk Application for Elderly People"
- WS 2018/19: "Human-Robot-Interaction - Requirements Analysis in Retirement Homes"
- WS 2015/16: "Context-aware Recommendersystems for Elderly People - Social Robots"
Practical Course "Usability Engineering"
- SS 2016: "Dialogue Application for a Social Robots - Cooking Assistant"
- SS 2015: "Context-aware Recommendersystems for Elderly People - Social Robots"
Seminar "Advanced Topics in Multimodal Dialogue and Interaction"
- SS 2015
Presentations
Talk
"Künstliche Höflichkeit und Frechheit. Wie erhält ein Pflegeroboter das passende Auftreten?"
("Artificial Politeness and Insolence. How does a Care Robot get the appropriate Demeanor?")
4. Interdisziplinäres Dialogforum „Mensch und Endlichkeit“,
Goldegg (Austria), September 21, 2021
Talk
"Was machen soziale Maschinen heute schon?"
("What are social machines already doing today?")
Careum Dialog 2020, Zürich, February 13, 2020
Workshop
"Wie man einen Roboter zum Leben erweckt"
("How to bring a robot to life")
- Girls' Day (2014-2019, 2021-2022)
- Tag der Informatik (2015, 2016, 2018, 2019)
Exhibit
"Lange Nacht der Wissenschaft"
("The Long Night of Science")
Augsburg City Hall, 2018
Exhibit
"Fädeln, Löten, Programmieren - Mechanische Wesen im Figurentheater"
("Threading, Soldering, Programming - Mechanical Creatures in Puppet Theater")
at the museum of the "Augsburger Puppenkiste", 2014-2015
- robot presentation at the vernissage
- robot exhibit (October 2014 - April 2015)