A computational model for learning from repeated traumatic experiences under uncertainty.

Area of Science:

• Computational neuroscience
• Behavioral psychology
• Trauma studies

Background:

• Posttraumatic stress disorder (PTSD) involves inflexible threat perception and behavioral adaptations.
• Existing models often overlook non-associative threat responses.
• Traumatic events can lead to lasting changes in how individuals perceive and react to threats.

Purpose of the Study:

• To propose a novel computational theory of PTSD incorporating non-associative threat responses.
• To model PTSD as an adaptation to the frequency of traumatic events using reinforcement learning.
• To explain how recent threat prediction errors influence future threat perception.

Main Methods:

• Development of a reinforcement learning momentum model.
• Testing the model with data from a mouse model of PTSD involving footshocks.
• Analysis of epidemiological data on PTSD incidence and trauma exposure.

Main Results:

• The proposed model successfully fits data from a mouse model of PTSD.
• Unpredictable footshocks in one context accelerated threat learning in a novel context.
• The model aligns with epidemiological findings on trauma frequency and early life trauma.

Conclusions:

• PTSD may stem from adaptation to the average of recent threat prediction errors, not solely specific associations.
• This theory offers new insights into the mechanisms underlying PTSD.
• The model generates novel predictions for the development and treatment of PTSD.