Bayesian Physics And Anxiety

When something bad happens and we’re interested in learning from our mistakes to stay alive in the future, we react in a few different ways

1. Automatically learn a stimulus-response connection, assigning to the immediately preceding stimulus the negative feeling we are currently experiencing
2. Dig further back into the past and replay things that might have been different: Isolate the stimulus with the least effort to change, hence small meaningless changes like tying a shoelace can be coaxed into showing a large meaningful change on preventing/causing the bad thing that happened.
3. Dig further back into the past: Isolate the stimulus that had the largest contribution to bringing about the negative outcome, the MVP
4. Dig further back into the past: Isolate the stimulus with the highest ratio of effort to change / impact on outcome. Small nudges are more effective learned behavior than large sweeping actions. Least Effort Player / Most Valuable Player

But is there a theory that organizes and unites these 4 strategies as relevant in particular spheres of influence?

Our understanding of cause and effect is linguistic and not physical. In Physics, things merely happen one after the other, and cause and effect are just positional circumstances arranged from left to right on a timeline. The correct calculation if we were really interested in problem-solving and root cause analysis is P(C/E) = n(C and E) / n(~C and E). I didn’t get the job, and I start to blame my shoelace, but first think about all the times I didn’t get the job and how many of those had shoelace issues. Instead we do a masterful linguistic shift and calculate P(~E/~C) = n(~E and ~C) / n(E and ~C). This math reflects our frame of reference when we play out the issue in our minds. Geometrically, we travel backwards to the point where we’re tying our laces, then execute a change in action, and then march forwards into a whole new situation ~E which is the opposite of E (not getting the job), with some gratuitous embellishments of winking at a supermodel who will become my wife, saving a kitten from a burning building, and winning the lottery, none of which have anything to do with my problem but all of which contribute heavily to the sweetness of life if only ~C had been executed.

We don’t have a great way of understanding the physics of mind. We copy-paste the real-world analog of Newtonian mechanics. Hence we have motivation, drive. Our will has power. We do mental work with psychic energy. We use the Force. But our physical intuitions fail us when considering mental work. We form metaphors of physical work. Like ‘getting the ball rolling’, the idea that the first decision and act of will is the hard part, overcoming the friction due to the weight of the ball, and then Newton’s first law kicks in and less mental work is required to ‘keep the momentum going’. We understand upward spirals, where every act of will is like the gap in a cyclotron that provides an electric field to accelerate a subatomic particle every time it makes a pass. The electron spirals outwards with ever-increasing speeds, until it is flung out tangentially at the end to travel at constant velocity until it collides with something else. We understand the downward spirals of misfortune. Drugs, depression, death. Each day is harder than before. Each day our momentum is dampened and that dampening is inversely proportional to our momentum such that each day our reduced momentum makes for increased dampening.

There is no compelling physical constraint determining that will power must follow laws of classical mechanics, there is only the linguistic convenience of the latter being broadly grasped by the human mind. The PFC is an energy-source like the cyclotron gap, periodically providing an accelerative force that we call motivation/drive/action leaving the rest of the time up to Newton’s first law. We give the toy-car a push now and then, it travels for a bit, slows down due to friction, and eventually comes to rest. Now it needs another push. Our PFC isn’t on rational mode the entire time, it needs to be summoned while the ANS hums away in the background. That’s why we have the metaphor of auto-pilot mode, or the CEO showing up for important board meetings and if a crisis is reported but otherwise delegating activities to competent managers.

If the CEO is absent for 10 straight actions and that ends up in a bad situation, our anxious minds calculate which of the 10 actions had the largest impact on the situation. Talking on the phone while crossing the road, I got hit by the bus. If I calculate P(C1 for cellphone/A for Accident) and find it to be high, I should basically be able to bail out of this predicament right away without need for reflection. But the problem is how do I calculate n(~C1 and A). Now this is basically n(C2 and A) where C2 is whatever happened just before the cellphone, like missing the subway. When you go to calculate n(C2 and A) you similarly run into the denominator problem that requires you to calculate n(C3 and A). Technically this is in an infinite problem but only for anal retentives. Practically, all I need to do is check if P(Cx/A) > P(Cx+1/A), if it is I can then ignore all x+1 to infinity. If not, then I go backwards until I hit the inflection point. But what if P(C1) itself is made higher because of C2? If P(C1/A) is higher than P(C2/A) but P(C1/C2) is higher than P(C1/~C2) then does C2 not gain relevance to A despite a smaller conditional probability?

All of this takes mental energy, and we calculate that it just isn’t worth the time or effort, unless something really traumatic happens and we give ourselves the luxury of sitting down to process it properly. But what if everything is traumatic? Or if we give ourselves the luxury of sitting down with absolutely everything to ‘process it properly’, abacus and log tables in hand? The rationalists among us will find a way to spin this well. After all, if I can identify actions that reliably lead to both large gains as well as losses, then that’s an investment in a future lifetime’s worth of good decisions. This is hugely problematic, not because we’re weak-willed creatures constantly falling short on execution of strategic goals, but because there’s a math error in this reasoning, we’re assuming that our prior probabilities of gain/loss values assigned to actions are radically wrong, and radically revised by our considered Bayesian updating of post-probability, hence changing our life forever. This gives us a more refined view of the relative strengths of the 4 reactions I started with -

1. Reactive behaviorism: Small negative shocks
2. RCA Shoelace: Large Negative shocks beyond my control, this is just fantasy without any predictive utility, an exercise to protect against learned helplessness perhaps. Do something, anything.
3. RCA MVP: Large Negative shocks that are both in my control and totally uncharacteristic
4. RCA LEP/MVP: Large Negative shocks that are in my control but not totally uncharacteristic

The classical mechanics of a CEO-mind are intuitively easy for us to apply to an understanding of anxiety. The CEO is an important woman with large constraints on bandwidth. We make 35,000 decisions in a day and 34999 of them are stupid, too stupid to require her signature, and yet we go to her with these stupid forms to get signed. Either

1. We’re too stupid to sign it ourselves: If poor training means middle management is too incompetent or unconfident to sign off on trivialities, then the solution is a radical retraining program. This is what manifests in our lives as habit-formation. Just shut up and listen to the CEO mandate exercise everyday for 2 years until you don’t need to ‘decide’ to exercise.
2. Consequentialism: If the CEO insists that the decision between brushing my left molar 3 times clockwise and 3 times anti vs 2 and 2 is hugely significant and requiring her personal attention, then I get stuck in an inefficient mode of micromanaged operation. I use the word Butterfly Effect too much and with too little understanding. I am an anxious person.
3. Paranalysis: If the CEO insists that the wrong decision, of brushing my left molar 3 times clockwise and only 2 anti, requires everyone to put away all their work for the day and go on a week-long intense off-site staff brainstorming meeting to attack the RCA of this momentous event, then I get stuck in a different micromanaged inefficiency. I use the word Butterfly Effect too much and totally differently, along with things like ‘everything is connected’. I have epiphanies about humiliating teeth-brushing incidents at scout camp. My life is full of meaning. But I am an anxious person.

This is a non-trivial problem. ‘Live in the moment’ is as facile and impractical as it is misleading and woefully incomplete. What do I do in the moment I am living? There is a subset of moment-living that certainly reduces anxiety. It is ‘act in the moment’, not live. At any moment, we may have 10,000 decisions to take, 10,000 past actions to analyze and 10,000 future actions to plan, but only say 20 current actions open to acting upon. Of these, 10 are foolish, involving cheesecake, Facebook, and MS-Calendar. 5 are smart, involving a skipping rope, a book and a dildo. 5 are brilliant, involving a thesis document, a divorce lawyer, and an elaborate plan to murder the wealthy uncle who loves me dearly. At any point we just need to maximize the payoff of the action and then erase it from memory entirely. How much of our anxiety is just the inability to identify and evaluate the actions available to us right now. We instead marinate in the fruits of actions sowed long ago and the potential fruits that may be reaped many years from now of actions to be sown today, with no CEO bandwidth for the actions to be sown today.