The implications of conversing with intelligent machines in everyday life for people's beliefs about algorithms, their communication behavior and their relationship building

The content on this page was translated automatically.

Communication with machines is steadily increasing in both quality and quantity. This does not remain without effects on the human communication culture, the mental models of users and the forging of relationships. Against this background, the project "The implications of conversing with intelligent machines in everyday life for people's beliefs about algorithms, their communication behavior and their relationship building", which is funded by the Volkswagen Foundation, investigates how voice assistants can be further developed, their effects empirically tested and ethically and legally evaluated. Together, design proposals will be developed to strengthen the advantages of the technology and minimize its risks. The project is part of the funding initiative "Artificial Intelligence - Its Impact on Tomorrow's Society" of the Volkswagen Foundation.

In addition to the project group on constitutionally compatible technology design at the University of Kassel, project partners include social psychologists from the University of Duisburg-Essen (consortium leader), ethicists from the Protestant University of Applied Sciences in Nuremberg, and computer scientists from the University of Bielefeld. The research questions address the topics of transparency, communication, and relationship building. The project takes a life-span perspective that includes vulnerable groups such as children and seniors. The project started on 4/1/2019 and will run for four years.

To answer the research questions raised, three scenarios will be addressed. These address different applications, user groups and the three levels of transparency, communication and relationship. In all scenarios, machine learning is used to optimize the interaction between humans and machines. The first scenario addresses children interacting with talking devices and thus the most vulnerable group. In terms of technical developments, it asks how an AI system can be designed to be transparent and self-explanatory for children. The focus of the empirical studies to be conducted in the scenario (three studies with 160 children aged 7-10) and the ethical analyses is on relationship building. In laboratory and field studies, children interact with an enhanced commercially available voice assistant to select music and audio plays. The legal analyses are devoted to the question of how the particularly vulnerable group of minors can give informed consent as well as use the system in a self-determined manner.

The second scenario analyzes how adults interact with a health app that makes health-related suggestions through a conversational interface. The technical work addresses how machine learning can be used for recommendations by making the algorithms explainable and more transparent (to researchers and users). Empirical research (three studies with 150 adults) focuses mainly on the mental models that users form of the system - depending on their transparency. Ethical analyses reflect on which mental model about the system is appropriate, while from a jurisprudential point of view questions of the way information is given, self-determined action and consent are in the foreground.

In the third scenario, seniors are considered in the context of Ambient Assisted Living. Here, the focus is on communication with a virtual agent that helps with scheduling daily appointments. Technical work focuses on combining machine learning (reinforcement learning) with classical dialog modeling methods. Empirically, two studies (with 55 seniors over the age of 60) will be conducted to show to what extent the seniors adapt linguistically to the system. In addition, the types of relationships formed will also be addressed. Ethical and legal analyses focus on the problems of artificial systems in care settings.

In addition, a long-term survey will be conducted involving 100 families over three years who use speech assistants in the home setting. Participating families will be interviewed twice a year about usage patterns, evaluation, mental models, emotional responses, and relationship formation.
The project will further be complemented by a Citizen Science approach. Due to the high social importance of the topic, the public is to be involved not only as participants in the empirical studies, but as active lay scientists who can ask relevant questions and provide answers in workshops and further studies.

Project partner

University of Duisburg-Essen, Social Psychology - Media and Communication:
Prof. Dr. Nicole Krämer (Coordination)

University of Bielefeld, Sociable Agents Group and Machine Learning Group:
Prof. Dr.-Ing. Stefan Kopp and Prof. Dr. Barbara Hammer

Nuremberg Protestant University of Applied Sciences:
Prof. Dr. Arne Manzeschke

Project info


April 2019 - March 2023

Project Officer:
Prof. Dr. Alexander Roßnagel

Contact person:
Dr. Christian Geminn


Project history