Posted on Leave a comment

Bayesian Depression, Affective Forecasting, and Psychedelics – First Principles

I’ve previously written about how to be a “high-functioning depressive,” one who faces recurrent periods of depression/blueness but manages these periods effectively. However, at times depressed individuals might claim that life is not worth living in the first place, that struggling to just stay afloat is not worth all the suffering it entails. Many of these people end up killing themselves.

Although morbid, suicide is an interesting social phenomenon that produces some important questions:

  1. What are the proximal causes of suicide? That is, what are the short-term risk factors that trigger someone to kill themselves as opposed to choosing to live another day?
  2. What are the distal causes of suicide? Are there certain risk factors—genetic predisposition, childhood trauma, lack of a tight social circle—that make someone more likely to kill themselves?
  3. Are more people truly suffering than we think? Do these people not kill themselves for fear of how it will affect their friends and family, some sort of social Nash equilibrium/Schelling point? If this social accountability mechanism were not in place, would more people kill themselves?
  4. Setting aside how family and friends are affected, given the evidence these suicidal individuals have, is killing themselves a “logical” decision from a Bayesian perspective? That is, are they accurately forecasting what their future will be like—rife with suffering and therefore not worth living—or are they being myopic and making a decision that, if they had omniscience, they’d regret (i.e. things will get better, but they just have to wait)?
  5. Related to the previous question, on average, how does subjective well-being vary throughout one’s lifespan? Do many people kill themselves young, before quality of life tends to improve (myopia), or does quality of life slowly decline over time, at which point people decide to just end it all (prescient)?

Why ought we care about these questions? Well, to understand how to maximize the quality of human experience, we first need to understand what causes humans to suffer. What is it like in the bowels of hell? What gets us stuck there? What is the neurobiological signature of these states? And, most importantly, if we’re stuck, how do we get out?

A Bayesian Approach to Depression and Suicide

To begin addressing these questions, in this post I’ll focus numbers 4 and 5: when people decide to kill themselves, are they making a logical decision?

I’ll be approaching this question with cognitive biases in mind. Humans are products of evolution, and the software it produced has lots of bugs. One such bug is that humans are poor at affective forecasting—predicting how some action or decision will make them feel in the future (for a good examination of this topic, read Stumbling on Happiness by Daniel Gilbert or this chapter he co-authored on the subject). This is critical when it comes to depression and suicide, as many people kill themselves when they think their future is misery and there’s no possibility of things improving.

I’d like to point out it’s quite possible that, given certain pieces of evidence about your quality of life, you arrive at the conclusion that life is on the whole a negative, and therefore choose to out yourself. This action is not illogical per se, though it is extremely selfish and may emotionally destroy your loved ones.

What’s the Difference Between a Good and Bad Day?

To make sense of this, consider the best days of your life. Perhaps you saw a friend you hadn’t seen for a while and enjoyed some great conversation; or maybe your application to medical or graduate school was accepted; or maybe you experienced some of the most profound, tranquil moments of your life while under the influence of certain psychedelic compounds. When you have these sorts of days, life seems worth living.

Now consider the darkest moments of your life. Perhaps a loved one died; or maybe you broke up with a long-term partner; or perhaps you experienced a string of bad days—a week, a month, or multiple months—that led you to believe that your life will always be like this, that you’ll never get out of this darkness. Painted into a corner, it may seem like life will never be worth living again.

Everyone deals with dark days and setbacks, and though these moments seem insurmountable when you’re experiencing them, most resilient people eventually overcome them and bounce back.

But, suppose you didn’t get over these bad days. Suppose that your life became a string of horrible days with no good days in sight. Triggered at first by some small failure or setback, things start to spiral out of control. If your life became a string of horrible days with no good days in sight, how would you react?

Into the Darkness

Everything you previously enjoyed–food, exercise, sex–slowly ceases to be enjoyable to you.

This anhedonia causes you to isolate yourself. You spend more time at home and less time interacting with other human beings. You find it harder to get out of bed every morning; why should you? What’s the point of getting up in the first place? Just to go through the motions, performing the same quotidian routine? What for? There’s no future you’re working towards, no light at the end of the tunnel. You’re stuck in some perverse version of Groundhog Day. Your life is monochrome.

You think of how other people don’t have the luxury of staying in bed. Other people less fortunate than yourself have real jobs and real families to provide for. They have real responsibility. You, on the other hand, are just a lazy sack of shit that can’t even manage to get out of bed and get work done. You’re privileged, so caught up in your first-world problems, and you hate yourself for this.

You may begin eating less (a common symptom of depression), or depending on your personality, you may begin eating more (the symptoms of depression vary widely).

You begin listening to “dark” music like metal (Metallica and Iron Maiden are some of your favorites) or folksy, sad music (e.g. Nick Drake) because you sympathize with it more than upbeat, happy music. Music used to have a profound effect on your mood, but now it just acts as noise to drone out the self-loathing.

You think that there’s something wrong with you, some quirk of how your brain works that predestines you to feel this way. You’ve tried talk therapy, you consistently exercise, eat and sleep well, but something is always a bit off.

Maybe it’s your lack of friends and truly intimate, vulnerable relationships? Maybe it’s your lack of purpose? But, you think, these lacks are your fault. Why can’t you just be normal like other people? Why are you so different? Surely no one else feels this way?

This is Normal and You’re Not Alone

Lots of people feel this way. Even many stereotypically “successful” people, despite the trappings of their success, feel this way.

A Review of Bayes Theorem

Bayes Theorem is a tool used to change belief in the face of new evidence. For example, consider the following example from the Data Skeptic podcast: you’re at the farmer’s market and someone, let’s call him Doug, is selling boxes of fruit. Doug says he is selling unmarked boxes of fruit for a reduced price. He tells you some of these boxes have all apples, some have all oranges, and others are half apple and half orange, with a ⅓ probability for each type of box. Doug chooses a box at random and pulls out an apple. What is the probability that the box he pulled from is filled with 100% apples?

Your hypothesis is that the box pulled from contains 100% apples. The evidence you’ve acquired is that 1 pick from the box resulted in 1 apple.

To get the updated probability (called the posterior), you need to take your prior (the probability of your hypothesis before any evidence) and multiply it by the likelihood of the evidence you got (an apple) given your hypothesis that the box is 100% apples. Finally, divide by the probability of getting an apple in any situation.

So, the prior probability of your hypothesis is ⅓, as each type of box is equally likely to be picked from all the unmarked boxes. The likelihood of seeing an apple given that the box is filled with only apples (i.e. assuming your hypothesis is true) is 100% .

Now, the probability of getting an apple in general is the probability of getting an apple from a certain type of box, P(A|half and half), times the probability of getting that type of box, P(half and half). This is called a conditional probability.

So for the 100% apples box, your probability is 100% * ⅓, giving you ⅓. For the half and half box, it’s 50% * ⅓, giving you ⅙. And for the 100% oranges box, you know that you will never get an apple, so ⅓ * 0 = 0.

Add up ⅓ and ⅙ to get ½. This is the denominator. The numerator is ⅓. Divide to get 66%, your confidence level in the hypothesis that the the box chosen from contains 100% apples. We have just updated our priors from 0.333 to 0.666, a substantial change.

Now suppose he pulls out another fruit from this box and it’s an apple. What’s the probability of your hypothesis now? To make a long story short, it now becomes 80%. As more evidence is in favor of your hypothesis, your confidence–that is, your level of belief–in that hypothesis increases.

Predictive Processing and the Bayesian Brain

Per Friston and Seth, the brain is a statistical organ.

My basic argument is that when people commit suicide, they are deciding that life is not worth living. They come to this conclusion because of the evidence they are presented with: internal physiological states (e.g. “feeling blue”), social feedback (e.g. a lack of human contact), etc.

We can define a belief not as some airy-fairy psychological construct, but as a high-level neurobiological state that an organism instantiates. These beliefs influences our top-down predictions about the world.

For example, if I’m hungry (I am interoceptively aware of a prediction error related to the homeostasis of the organism that is me) and I believe there is food in the fridge, then I will try to reduce the prediction error by looking in the fridge, getting the food, and eating it. This is a form of active inference.

Suicide as active inference

When people are depressed, they can get stuck in a positive feedback loop in which they’re unable to summon new evidence. Consider someone who has been feeling blue and deals with social anxiety. They have certain hypotheses/beliefs about the world (no one likes me, I have no friends, etc.) that reduce the amount of active inference they can do; that is, if you’re too afraid to go out into the world and try to meet new people–i.e. Find evidence that contradicts your priors–then you’ll never revise those priors. Thus, these people get “stuck in a rut”. Their priors are too strong and they’re unable to perform the exploratory behavior necessary revise them.

We are information-processing machines. We predict the world around us through a generative model. Put in certain parameters and you’ll get certain outputs.

The action one takes–killing themselves–is a form of active inference. It is a prediction of sorts.

Psychedelics and Affective Forecasting

Psychedelics are a way of resetting your priors. They allow you to bypass the typical action needed to revise your priors, allowing you a shortcut to see the world with a new set of eyes.

When someone is stuck in a depressive state, they only experience an extremely circumscribed set of psychological states. Their brain is stuck on repeat, stuck in the same ruminative loops.

Psychedelics help your brain get out of these loops (see Michael Pollan’s forthcoming book How to Change Your Mind).

Related Research Questions: Hedonic Setpoints, Wireheading, and Psychedelics

All the questions about suicide prompt some of the following questions about depression and anhedonia more generally:

  • Why are some people’s hedonic setpoints—that is, their happiness or reward setpoints—lower than others’?
  • Why are some people’s hedonic state more variable than others’?
  • Why are some people more resilient in the face of stress and tragedy than others?
  • Clinically speaking, depression is an extremely heterogeneous disorder, with symptoms ranging from anhedonia to lack of appetite. How ought we revise our nosology—that is, our classification of diseases—to better treat diseases like depression?

I find the optimist in me asking some of the following questions too:

  • How great is the range of human experience? How high are the peaks of tranquility, meaningfulness and pleasure, and how low are the valleys of madness, pain and suffering?[1]
  • What role do alterations of consciousness, whether via meditation or psychedelics, play in experiencing these range of psychological states?
  • What are the neurobiological substrates of these positive states, and can we “hack” them (i.e. take a shortcut to achieve them) or is there truly no free lunch when it comes to biology? Some have termed this hacking of our psychology wireheading; it ranges from electrically stimulating the pleasure centers of your brain (probably not a viable long-term solution) to engineering your brain and body to maximize positive and states and minimize negative states, perhaps through the use of psychedelics, or simple practices like exercise, fasting and meditation.

[1] I should add that through all of this, the skeptical philosopher in me asks if there exists such a thing as a positive or negative mental state, or if it’s all just physical information-processing. But, as I’ve stated before, unless you want to venture off into the more esoteric parts of Buddhism or become a bat-shit-crazy eliminativist, quining qualia (as Dan Dennett termed it) is probably not the way to go. So, I’ll just assume that there do exist positive and negative mental states.

Leave a Reply

Your email address will not be published.