Robotics and Biology Laboratory

Rational Selection of Exploration Strategies in an Escape Room Task

© DALL·E 2

A robot solving an escape room

Project Description

In the psychological and biological fields, it is commonly assumed that the mind disposes of a repertoire of different strategies, which it utilizes in solving cognitive and behavioral tasks. This repertoire is also known as the mind's toolbox. However, the mechanisms underlying the selection of the proper strategy from the toolbox are still not well known. The framework of strategy selection as rational meta reasoning introduced by F. Lieder & T. L. Griffith, 2017 assumes that the strategy selection from the toolbox is based on a subjective assessment of two key variables, which are a trade off against each other:

  1. The strategy Accuracy for a given task.

  2. The Costs necessary to implement the strategy.

However, the accuracy and costs of a strategy in a given situation are usually not directly accessible to the decision maker and need to be inferred. 

In this project we aim to contribute to the Ecologically Rational Strategy Selection through providing a better understanding by focusing on how the crucial inferences of the framework's parameters for strategy selection are cognitively and computationally implemented and how they can be improved.

Experiments

We try to answer these questions through multi-stage experiments. In the first phase, we aim to identify strategies that people employ or develop when solving new problems. In the second experimental phase, we attempt to provide participants with a strategy toolbox based on the identified strategies and study the mechanisms of their strategy selection from this toolbox. This allows us to evaluate existing hypotheses and generate new ones. In the final phase, we plan to design new experiments to challenge our hypotheses and gain further insights into how people solve problems and how the strategy toolbox contributes to it.

The first two phases of the experiment are conducted using a simulated Lockbox task. In the LockBox task, participants are presented with a 3x3 grid on a computer screen containing six linear sliders. The sliders are randomly distributed on the cells of the grid in each trial, except for the target slider, which is always located in the upper right cell and marked with an orange outline. The sliders are randomly oriented vertically or horizontally, without functional distinction. Initially, all sliders are positioned either downward or to the left, depending on their orientation. Each slider has a binary state, meaning it can only be at one of the two ends of its corresponding slit.

The goal of the task is to solve the LockBox by moving the target slider to the other end of the slit. At the beginning of each task, the target slider is typically blocked by one or more other sliders, which, in turn, may be blocked by others. A blocked slider cannot be moved until all blocking sliders are placed in the required state. The dependency structure of a LockBox can be expressed as a directed acyclic graph. How the sliders block each other is hidden and must be explored by the participant by attempting to move the various sliders.

A Haptic Device is used to manipulate the LockBox

Please note: Once you watch the video, data will be transmitted to YouTube/Google. For more information, see Google Privacy.

Funding

This project is funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany's Excellence Strategy - EXC 2002/1 "Science of Intelligence" - project number 390523135.