Detecting Mania

In hindsight I had the data to see the episode a week in advance.

The details are hazy. Mostly lost to time.

The graph is not.

Tracking my first manic episode

A perfect inflection point up. A hockey-stick line. A point on 2/7 where the graph continues steadily upward until 2/14; a deviation from its normal, flat persona.

And I know when the manic episode “activated” — activation being the tipping point from unhindered mania to mania under spiritual influence. My episode activated the night of 2/13.

It’s possible to see such things come and go, you know. I saw a DNA spiral rotating around the pitched ceiling’s apex. A DNA spiral in spirit space, letting me know that nature was changing the features of my life. I didn’t know it at the time — it’s hard to know the significance of things in the moment — but that was a significant turning point in my life.

The system — had it been monitored by someone not under its monitoring — would have caught the episode before nature would have what it needs to activate the episode.

I was in an outpatient hospitalization program. Essentially daily therapy to help with my treatment resistant depression. The program provided lectures on different therapy modalities as well as some group and individual therapies.

They had every chance to spot my mood swing upwards, except they were viewing it through my lens. From my perspective I was still depressed because I still felt much of the depression right up until that week. I was reporting my mood daily and nothing about that report was used to detect my episode.

The professionals aren’t to blame here.

I was able to spot the episode a week in advance, and had they been given the same data they’d have been able to do the same. They were operating from a perspective that allowed my own cognitive distortions to creep in and disrupt the unbiased collection of mood statistics. Modern therapy fails because it relies upon the patient for ground truth reporting and I was reporting to them that I was depressed. I was reporting to myself I was depressed and it stuck, even as my mood moved upwards.

It was the start of “the game” for me.

An intense, alternate reality game taking place behind the scenes of everyday real life. Gain points by working in favor of your clan in the game. Get others to work on projects that support your clan. Work in secret by working through other projects.

It was all cloak and dagger. A conspiracy behind the scenes.

It’s all too benign to write about. It’s just…~obvious. I was adding a layer of intensity that did not need to be there for life to be worth living. It’s obvious that we all gain in life by supporting our clan be it family, friends, or co-workers. Everybody has agendas — hidden or otherwise — and the clan does what it can to make people work on its projects.

When put like that the delusion loses all its sparkle and pop.

It’s boring. Mundane. Obvious. No need for alternate reality or cloak and dagger to tell the same story, so it feels forced, faked, and wrong to add these details to life. They don’t enable any additional modes of thought or new behaviors, so they’re extraneous.

Except they are exactly the kinds of thoughts my delusions make appealing.

I’ve heard them described as sanctum and punctum. One is the “wow” factor — sanctum — and the other is the technical factor — punctum. Get a perfect shot of something 100\% boring? Pure punctum. Get horribly framed out of focus, poorly composited picture of a hilarious subject? Pure sanctum.

For me the game was about adding sanctum to a life of punctum. The belief that those around me were playing this game too and just trying to engage me into playing the game. The problem was my delusions often turned towards being excited by carnage. I was afraid I’d feel pressure to create such carnage as a sanctum-generating act. I once told the angels to play “cache hammer” and “2-bit flip” on people’s computers to cause transient outages without lasting downtime.

Cache hammer” was a game of interrupting the Modified Exclusive Shared Invalid (MESI) protocol of computers. Find a cache line — the base unit of memory in a processor under one view — that was in use on another core and spam the invalidation action, taking an exclusive lock. This won’t break things, but they will slow them down a hell of a bit. And the slowdown would be hard to detect without tooling most people aren’t familiar with using in production.

2-bit flip” was about crashing computers by flipping more bits in memory than can be error corrected.

Between the two, the angels seemed to play cache hammer more. Plausible deniability.

It’s about acting contrary to what’s expected in one area to make progress in another. Life’s all about how we resolve competing interests, and my delusions were presenting as another party looking to compete for my attention.

Acts of service to prove loyalty.

A willingness to live with the consequences of breaking things in furtherance of the game.

The game” is a slippery beast. The problem with delusions is they build upon truths. The closer they are to truth, the harder they are to pin down. And “the game” was all about adding a little manic shine to each and every action, and about taking some big leaps of faith to help the universe.

I just didn’t get it. I saw others’ success and way with people and ascribed to it the actions they take that I don’t see. And so I would set myself to enumerating the options. That’s how psychosis gets its teeth into me. I’d enumerate all the options, but fail to stay tethered to reality.

I saw a world in which machines controlled our every emotion simply because they controlled social media. Ostensibly, this is true. It’s been shown that social media affects mood, and the algorithms that affect social media’s presentation and curation wings rarely involve humans in the data plane. The underlying form is one of machine influencing mood.

When you’re biologically predisposed to psychosis, discovering such an underlying pattern is a dopamine rush unlike no other. It rapidly gave way to a world in which human and machine were clashed in an epic battle.

The machine makes sure no one sees the post where you ask for help — it wouldn’t go viral anyway.

The machine makes sure that everyone is kept in line.

The machine makes sure that everything is useful.

Rapidly, “the machine” turns from the concrete data centers ruled by impersonal flow charts to society as a whole. To secret undercurrents that come about when all the competing interests of society rise together.

But I saw no evidence of any of this.

All I saw was others’ success with getting their ideas to take root in peoples’ minds.

The game got bad enough I had to take precautions at work.

I kept myself out of production and off the on-call rotation for I feared that during an incident the game might rear its head and encourage a choice to the detriment of my employer. I feared that while on-call I might be asked to serve the game and would do so before my judgement and awareness could kick in.

This was mainly a hypothesized threat, but I had delusions that I could dispel during that day that were just strong enough that a 3~a.m.~version of myself would not be able to dispel them. I was afraid an outage overnight might balloon at my hand. I’d be responsible for it. There’d be no insanity defense; besides, a defense doesn’t revert the damage done.

It was a tough position to be in.

Eventually, I concluded that the only thing I could do to protect my employer was to resign. It wasn’t the most rational thing to do, but I was burning out and had delusions daily. I didn’t want to be responsible for a misstep on my part harming my co-workers or the company.

It sticks with me: A perfect hockey-stick-line up starting on 2/7 where I’d gone off the rails from my life maintenance tasks. I’ve invented a signal for detecting when my background tasks of life fall off. And, for me at least, the background tasks fall off when I become couch-bound with depression or alight with manic energy, so it’s a signal of bipolar activity.

There’s something to be said for such a signal. Technology can take on the burden of computing and monitoring the signal, connecting humans together when the signal deems it necessary.

I wanted to build such an application. Something that captures the signal I had, monitors it with some number of nines of reliability, and calls family, friends, or caretakers when the signal falls outside the acceptable range. I investigated building this and concluded that it wasn’t profitable as a standalone endeavor. There are just too many variables across people to get it right without significant resource expenditure.

For that reason I want to give away the system in hopes that someone with resources picks it up and put it in front of people.

The system is fairly simple: A rhythm is a recurring task of life. The system asks the user to identify every rhythm in their life and capture evidence of it digitally. Examples of rhythms include doing laundry, cleaning out the vegetable drawer, and buying new underwear.

These are all things we do a finite but frequent number of times in our lives and failure to do them often correlates with mental health conditions.

That’s the premise anyway. I stumbled upon it by happy accident. I wasn’t really intending to track my rhythms, but I discovered that the task system I was using kept history of when I completed each rhythm. At peak I had $\sim 150$ different tasks, all on repeat of various forms to capture the rhythms they represent. A 1:1 between tasks and rhythms.

That’s what makes this such a hard idea to sell. It’s just a TODO list at the end of the day. A smart, recurring TODO list. Nothing more, so selling it at a high price is a hard sell. Plus, it takes dedication and engagement to enumerate life’s tasks. For me, it was easy because I had to do it anyway for my autistic side. I need to make explicit the maintenance tasks of life; else, they end up falling by the wayside and life gets progressively more crusty with time.

The fact that it detected mania is incidental and anecdotal. I’m describing it here for the benefit of all, but I’d be happy to follow it to a home where other similar ideas I have can live as well.

Formally, the metric to track is the cumulative number of days that tasks are overdue. If you have five tasks, each overdue by a day, it’s five points. If you have one task that’s five days overdue, it’s five points. One point for each day that a task is overdue, computed for each task, and then summed across tasks. It’s a simple metric that provides adequate signal for us to latch onto.

There’s always some background noise, so the goal for this new metric is not zero. Rather, the goal is stationary. In a healthy application of the system, the value will wander slightly day to day, but stay within a reasonably small range. The day one falls out of step with tasks is the day it begins to drift up, commensurate with how far out of sync they are.

I’ve lost pieces of the process I used to find this signal, otherwise I’d open source what I have. It’d be nice to have a home for it with people better able to tend to it than I. My reliance upon the tool is the exact reason I should not host it; a personal shortcoming.