Today's trauma paper is "Neural Computations of Threat", Levy & Schiller 2020, Trends in Cognitive Sciences. https://t.co/uwiaSJv4l9
Much of this is very similar to trauma therapy stuff, but from a cogsci / neuroscience angle so sounds more reputable and scientific. ;)
The overall frame is familiar: anxiety disorders and PTSD are manifestations of a similar underlying mechanism which involves fear learning. Either can be learned as the brain comes to predict that a particular stimulus is predictive of danger. https://t.co/l5oyEICEr6
A learned threat prediction can be triggered and made available for reconsolidation through several processes, including extinction (repeatedly experiencing the trigger without experiencing something threatening), counterconditioning (associating it with something positive), https://t.co/RgSBh5axcs
or doing sensorimotor interference (e.g. playing a game of Tetris, disturbing the process of reconsolidating the memory and thus weakening it).
(I don't see coherence therapy-style reconsolidation mentioned in this figure but CT gets an explicit callout later.) https://t.co/ZPXA5RXylT
Like this distinction: in a "safe" state there's no threat (e.g. predator) in the immediate future; in a "pre-encounter threat" state, there's no visible threat but one may emerge at any moment; in "post-encounter threat", the predator hasn't spotted you but you have spotted _it_ https://t.co/kIfumNKVwT
Anxiety disorders probably involve the brain misclassifying a "safe" environment as a "pre-encounter" or maybe even a "post-encounter" one; PTSD involves the brain thinking that the threat is even closer, and triggering associated responses. https://t.co/9hB7RbDBAH
Interestingly this associates fear and anxiety with "higher cortical areas" and only "freezing, escaping and panic" to earlier processing. A more distal predator allows for "model-based planning"; thinking about how to deal with it and running mental simulations. (See rumination) https://t.co/zwF6g2qFNz
Suppose some action gives you a shock 80% of the time; if you know to expect a lack of shock about 20% of the time, that's expected uncertainty. But if suddenly the whole probability distribution changes, giving you a shock only 30% of the time, that's unexpected uncertainty. https://t.co/hggCG8efB8
In natural environments, having a lot of unexpected uncertainty is common, so people are naturally good at quickly adjusting their reward predictions.
Anxious people tend to learn more from recent punishments but to be slow to adjust their learning overall. This makes them naturally averse to novel situations. https://t.co/STDmteCr7i
(The paper calls anxiety and PTSD as involving "aberrant neural computations" but if you're invoking a Bayesian framework anyway, why not talk about them being different priors on the environment being threatening or unpredictable? Would seem natural to me.) https://t.co/mPV3R3GI94
Talk about memory reconsolidation in threat learning. Sounds like nobody has refuted any of the major claims in e.g. Unlocking the Emotional Brain ( https://t.co/hf4voSbOHc ) yet. https://t.co/pwNIX4Hp7A
Here's the promised Coherence Therapy shoutout. https://t.co/RCS0VZ56m7
Lots of Words about "so what determines your action when you think there's a threat that needs to be dealt with", which boil down to "here are some brain areas we think are associated with this stuff, but we don't really know what happens".
(So a normal neuroscience "answer".) https://t.co/oW6dYjUed9
In the end, they think that all this stuff is about two core problems: "making predictions about long-term accumulated outcomes and tracking uncertainty
in the environment".
And some people thought all this emotion and trauma stuff was somehow distinct from Bayesianism. ;) https://t.co/zQZjgyDNTX