How human must a robot be before we can rely on it?

The aim of this interdisciplinary project is to examine how human-like traits in AI and robots impact our interaction with technology and with each other. The project is focusing on two aspects of resemblance to humans. 

The first aspect is agency, i.e. whether we perceive that the AI or robot has intentions, feelings and a “will of its own”. The second aspect is anthropomorphism – the question of whether the robot looks and sounds like a human.

Whether AI and robots should possess these characteristics is an open question. The researchers believe the answer likely depends on a number of factors. They consider that psychological and ethical aspects are rarely taken into account, and that there is very little understanding of cause and effect. The researchers therefore intend to study how agency and anthropomorphism impact psychological and social factors that are critical for lasting social interaction and human decision making, such as trust, reciprocity and honesty. Based on empirical observations, they will also examine the ethical factors involved.

One challenge of studying social factors like trust is that they are extremely difficult to measure reliably, partly due to their subjective nature. The researchers are therefore using experimental methods, known as economic games, which are used in behavioral economics. They intend to modify them, however, so they involve an AI or robot. This will enable them to study how subjects in the experiment make economic decisions on the basis of trust, reciprocity, altruism and honesty.

The researchers can manipulate the agency and anthropomorphism of AI, enabling them to study how economic decisions are affected. Another key task is to learn how these factors evolve over time. To study this, they will also be developing an artificial financial advisor with which the subjects will regularly interact.

The project is a new interdisciplinary collaboration between researchers in computer science at KTH Royal Institute of Technology, in behavioral economics at Stockholm School of Economics and in psychology at Stockholm University.

Project:
Relying on a Robot: How agency and anthropomorphism in human-centered AI affect social decision making

Principal investigator:
Gabriel Skantze

Co-investigators:
Anna Dreber Almenberg
Jonas Olofsson

Institution:
KTH Royal Institute of Technology

Grant:
SEK 6 million