Using Personality Models as Prior Knowledge to Accelerate Learning About Stress-Coping Preferences

Abstract

Nominee for the Best System Demonstration Award presented by Ann Nowe and Shih-Fen Cheng

The management of (dis)stress is an important factor for a long and healthy life. But everyone's stress is different, so is everyone's stress coping strategy. Within this demonstration we introduce PeSA, the Personality-enabled Stress Assistant, an agent-application that accounts for this individualism bringing together several agent techniques: Reinforcement learning is applied to learn about preferences of the users, prior knowledge and knowledge transfer is used to accelerate the learning process, agent mirroring is utilized to enable communication and offline functionalities. Based on that, PeSA's main functionality is to support the user during stressful phases by advising coping strategies that fit with the personality of the user. The user can rate these advices, thus, providing a reward/punishment signal.

Poster presented @ AAMAS 2016

Click to download an A0 version of the poster.

@InProceedings{Ahrndt2016Using,
  Title                    = {Using Personality Models as Prior Knowledge to Accelerate Learning About Stress-Coping Preferences (Demonstration)},
  Author                   = {Sebastian Ahrndt and Marco L"{u}tzenberger and Stephen M. Prochnow},
  Booktitle                = {Proceedings of the 15textsuperscript{th} International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2016), Singapore},
  Year                     = {2016},
  Editor                   = {John Thangarajah and Karl Tuyls and Catholijn Jonker and Stacy Marsella},
  Month                    = {May},
  Organization             = {IFAAMAS},
  Pages                    = {1485--1487},

  Abstract                 = {The management of (dis)stress is an important factor for a long and healthy life. Yet, stress affects people differently and everyone manages stress in different ways. In this paper we introduce PeSA, the Personality-enabled Stress Assistant, an agent-based application that accounts for this individualism. PeSA merges several agent techniques: Reinforcement learning is used to learn about preferences of the users, prior knowledge and knowledge transfer is applied to accelerate the learning process, agent mirroring helps to enable communication and offline functionalities. Based on these mechanisms, PeSA guides through stressful phases by proposing coping strategies that are tailored to the personality of each individual user. Users can assess these advices and thus provide a reward or punishment signal that helps PeSA to improve its suggestions.}
}
Autoren:
Sebastian Ahrndt, Marco Lützenberger, Stephen Marc Prochnow, Sahin Albayrak
Kategorie:
Tagungsbeitrag
Jahr:
2016
Ort:
John Thangarajah, Karl Tuyls, Catholijn Jonker, and Stacy Marsella (eds.) Proceedings of the 15th International Conferences on Autonomous Agents and Multiagent Systems (AAMAS 2016), Singapore. pp. 1485-1487. IFAAMAS. ISBN: 978-1-4503-4239-1