"How is it to be a human?" – Enabling a social robot to have open conversations with humans on fundamental topics

Bachelor Thesis

What makes us human? Such a simple question still has no clear answer. The goal of this thesis is to investigate whether social robots can help shed a new light on the topic, by providing the unprecedented opportunity to discuss “what makes us human” with an intelligent artificial agent. At the same time, the thesis provides the opportunity to investigate the effect that holding such a conversation has on a person’s perception of the robot and its intelligence.

Concretely, in this thesis you will: (1) enable Pepper/NAO to hold open conversations by bridging with Open AI GPT-3; (2) custom train the language model if needed to improve its accuracy; (3) design a simple experiment with human participants to validate the accuracy of the language model and the effects that such an interaction has on the participants’ perception of the robot and its intelligence; (4) collect and analyse the outcomes of the experiment.

This thesis requires you to (1) be human and (2) proficient with programming in Python. Experience with OpenAI GPT language models, the Pepper/NAO robots and their programming tools (Choregraphe, NAOqi) is advantageous, but not necessary.

This thesis will give you the opportunity to (1) gain first-hand experience on state-of-the-art language models, (2) work with the most popular social robots on the market, (3) learn how to design user studies and (4) have a lot of fun hearing what humans have to say about themselves.

Target group: Computer Science

Starting date: immediately

Supervisor(s): Barbara Bruno (barbara.bruno∂kit.edu), Romain Maure (romain.maure∂kit.edu)