Advisor — Master Thesis "Comparing the Mechanisms of Plan Generation in Humans and Machines via Failure Modes"
Supervised and Examined by Oliver Brock & Eva Wiese
Student
Antonia Köngeter
Abstract
We examine AICON, a previously introduced method where we use interconnected components and gradient descent to determine new actions, on its problem solving ability. This is done by testing it on planning problems and comparing it to humans. We define the failure and success modes of AICON and humans in order to explore AICON’s biological plausibility. In order to determine AICON’s failure modes in planning problems we select the Tower of London Test, a popular test in clinical psychology. Hence, we compare it to human data with the characteristic parameters of the different problems. We found that AICON and humans have a correlation in their success rates and the extra moves needed to solve a problem. The characteristic Tower of London parameters have similar effects on the success in AICON and humans. Therefore, AICON and humans have similar patterns in planning problem solving. Furthermore, this might indicate that humans do not always solve planning problems by planning ahead, but in a partially AICON-ic way.
Related Publication
Publication pending.
More Information
More information can be found here.