Predicting disruptiveness of interrupts using cognotove models
- Bachelorarbeit -
Description:
Background
AI assistants are currently everywhere to be found. An attempt to make them more interactive and maybe even pro-active needs further research of the disruptiveness of interrupts in the human multitasking ability.
Machine-Learning is a modern and efficient way to train AI for a wide range of use-cases. One of the problems here fore is data availability, especially when it operates with human data, like an assistant. Gathering enough data from tests with humans to train an AI sufficiently is extremely time consuming and requires an enormous amount of participants.
Objective
This thesis researches the usability of cognitive models for predicting the disruptiveness of interrupts by building a cognitive model to solve a task comparable to a human, including mistakes. The goal is not to train the model to do the task as efficient as possible.
Methods
It’s part of a project with the University of Delft and evolves around the development of a cognitive model based on ACT-R that provides human-like data in solving a task. An AI makes suggestions which could help solving the task more efficiently. The future part of the project is to use the model's data for machine learning purposes to improve the quality of the robot’s suggestion's and make them less disruptive for the human. The ideal result would be an AI which always suggests useful steps at time- slots where the human is least disturbed by them and therefore, they can work with one another without getti ng in the way of thoughts or actions.
Anforderungen/Kenntnisse:
Machine Learning, Cognitive Modelling
Bearbeitung:
Timo Wiesner
Betreuung:
Prof. Dr. rer.nat. Nele Rußwinkel
Institut für Informationssysteme
Ratzeburger Allee 160 ( Gebäude 64 - 2. OG)
23562 Lübeck
Telefon: 0451 / 3101 5700