Project Info

Adaptive Model of Robot Performative Autonomy

Tom Williams

Project Goals and Description:

This project aims to implement an adaptive model of robot performative autonomy, a strategy in which robotic agents intentionally perform lower levels of autonomy than they are truly capable of in order to raise human teammate's situational awareness (SA) levels. The goal of the model is to accurately predict the human's SA in a collaborative task and adjust the robot's behavior accordingly. When the model predicts that the human's SA is high, the robot will adopt a more autonomous stance and carry on its tasks without asking questions to the human. When the model predicts that the human's SA is low, the robot will intentionally ask questions to the human in order to raise their SA levels. The implementation of this adaptive model is interesting because it may increase the efficiency with which human-robot teams can perform collaborative tasks. In addition, compared to results from our previous work, it may produce better user ratings about the robot in terms of key metrics such as trust, teaming quality, and perceived intelligence.

More Information:

Grand Challenge: Reverse-engineer the brain.

Primary Contacts:

Tom Williams, | Rafael Sousa Silva,

Student Preparation


The interested student should have some experience with experimental design, and should have taken at least one 400 or 500-level robotics course.




Understand how Markov models work and how they can be used alongside basic reinforcement learning algorithms. This will be used to explore the implementation of a Q-table as a means to keep track of how the robot should adaptively respond to different human SA states.


The student will have weekly meetings with a graduate student mentor, who will closely monitor their progress, set weekly goals that will keep the project moving forward, discuss relevant papers with them, and provide guidance and support whenever necessary. This includes support and mentoring related to model logistics, programming, and planning/running experiments.


Share This