School of Science and Technology

Robot duck debugging : Studying the use a conversational agent in Thinking-Aloud Pair Problem Solving

Motivation and scope

"Thinking-Aloud Pair Problem Solving" (TAPPS) is a technique used to allow an individual to be more efficent at problem solving (references). This techniques requires a problem solver who will verbally articulate their thought process and problem solving, and non active listener whose role is to keep the problem solver talking, ask clarifying questions, and encourage the problem solver to explain their thinking process out loud. This technique is also often referred as "rubber-ducking" in the computer science world. In this context, the rubber duck is used as a proxy for an active listener presence, but leaves aside the active role.

The goal of this thesis is to explore whether an cognitive system can act as an "improved" rubber duck, by re-introducing some of the active role of the listener, especially inciting the problem solver to keep talking. To do so, the student will have to implement an existing framework to predict backchannel usage from a conversational agent, and conduct a user-study to evaluate the impact of these backchannel on the
TAPPS process.

Specific Tasks

1. Review and analyze litterature on backchannels for conversational agents
2. Implement existing algorithm to predict backchannel usage in a non-embodied agent
3. Validate the developped approach with human users

The code and models developed during this thesis will be reused in future research at AASS. Appropriate acknowledgement to the student will be given in case of any publication using the developed code.

Necessary Skills

- Excellent programming skills
- Willing to learn and conduct user-experiments


During this project you will learn how to work with conversational agents. you will also learn how to design and perform proper user periments. Depending on the project resluts and student's motivation, a publication in a peer-reviewed venue is possible.


Jennifer Renoux