Master
English

Computational Modelling of Pavlovian and Instrumental Learning under Stress - A Reinforcement Learning Approach

Number of pages52
Master program titleMaîtrise universitaire en neurosciences
Defense date2023-08-28
Abstract

Background: Pavlovian and Instrumental learning are fundamental in shaping reward-seeking. They compete and interact to guide behaviour. Stress has been observed to impair Instrumental goal-directed behaviour and facilitate habits, relevant for clinical disorders characterised by habit-based compulsive reward-seeking. The impact of stress on the interaction between Pavlovian and Instrumental systems remains poorly understood.

Method: To investigate their interaction and whether acute stress moderates it, twenty-eight male participants undergo a Pavlovian-Instrumental conflict task on three consecutive days. Anticipatory gaze direction and gaze dwell time on regions of interest are measured with an eye tracker. Three computational models are simulated and fit to the data, to better understand the dynamics between how the systems learn, as well as investigate possible free parameters differences between the stress and control conditions.

Results: Behaviourally, a significant Pavlovian interference on the Instrumental system can be observed (Pavlovian bias), reducing the effectiveness of goal-directed behaviour. Computationally, an integrative model that accounts for both systems as interacting components accounts for the behavioural data best. Extracted free parameters show interesting incipient trends, but are not sufficiently conclusive.

Conclusion: This study provides evidence for the Pavlovian system’s role in interfering with Instrumental goal-directed action at both behavioural and computational levels, but future research is needed to assert whether acute stress does indeed have a moderating role in this interference.

Keywords
  • Pavlovian Conditioning
  • Instrumental Conditioning
  • Reward Learning
  • Computational Modelling
  • Rescorla-Wagner
  • Acute Stress
Citation (ISO format)
TUDOR, Maria-Cristiana. Computational Modelling of Pavlovian and Instrumental Learning under Stress - A Reinforcement Learning Approach. Master, 2023.
Main files (1)
Master thesis
accessLevelRestricted
Identifiers
  • PID : unige:176412
116views
14downloads

Technical informations

Creation13/04/2024 09:12:54
First validation16/04/2024 07:29:25
Update10/06/2024 13:54:42
Status update10/06/2024 13:54:42
Last indexation01/11/2024 09:14:35
All rights reserved by Archive ouverte UNIGE and the University of GenevaunigeBlack