Everything is Predictions
Delphiosis: The Model - Prediction - Action Framework & Its Usefulness
“There are these two young fish swimming along, and they happen to meet an older fish swimming the other way, who nods at them and says, “Morning, boys. How’s the water?” And the two young fish swim on for a bit, and then eventually one of them looks over at the other and goes, “What the hell is water?”
Sick parable, right? Its core idea is: fish are idiots… Or, more accurately, we can be so immersed in a thing that we don’t notice it. This idea can apply to lots of subjects: Math, Chemistry, Language, History, Microbiology, ketchup. For each of those topics, you could argue “society is enveloped by ______, but we don’t stop and think about it.” For example, viruses are everywhere and nobody talks about viruses. Ehh, maybe that’s not the best example. Still, look up how many viruses are in a drop of seawater. In spite of that fact, viruses don’t stop anybody from paddle boarding. The Malibu Police do that.
In addition to that list, the Idiot Fish parable also applies to predictions. We humans are swimming twenty thousand leagues under a sea of predictions but don’t notice. This might seem like a silly philosophical point to make, but it is practical because certainty that emerges from predictions is more useful than truth handed down from authorities.
I’ll loop back around to the usefulness of predictions later, but let me say first that its not just my idea that predictions are important. Here’s a quote from Lex Fridman’s interview of Nobel Prize winning Psychologist Daniel Kahneman:
Its very clear that animals have perceptual systems and that includes an ability to understand the world, at least to the extent that they can predict, they can’t explain anything, but they can anticipate what’s going to happen and that’s the key form of understanding the world.
And here’s Lex talking to Demis Hassabis, the CEO of DeepMind:
“What’s interesting about GATO (DeepMind’s General AI) is, it’s pushed prediction to the maximum in terms of like, you know mapping arbitrary sequences to other sequences and predicting what’s going to happen next. Prediction seems to be fundamental to intelligence.”
These two very smart dudes, who have spent their careers studying cognition, agree predictions are a big deal. If predictions are fundamental to understanding the world, then you can view all living things as moving through time based on a model - prediction - action framework. Before you do the things you do, you predict that doing them is the right thing to do. Or, a better way of saying it might be: living things use models. Those models are the basis of predictions. And action is based on predictions. Every action a living thing does is the result of the best prediction it has come up with out of several other imagined possible actions.
Now, “everything moves through time base on a model-predictions-action framework” is a lot of letters to type over and over again. And I don’t know how to use macros. So I decided it was easier to invent a word for this idea. I came up with “Delphiosis,” combining the oracle at “Delphi” with the Greek root “biosis” meaning life, or way of life. I thought it was catchier than my first idea, Tzatzikiosis. Ancient Greek scholars reading this might dunk on my shitty portmanteau, but goddamn it, I’m a big boy.
Below I’m drilling into the steps of Delphiosis to flesh out what I mean.
Model
I’m not talking about the kind that Jordan Peterson gets disgusted by on twitter. I’m talking about mental models, models of behavior. But more than that too. Living things don’t need a brain, or even organs to use models. Seeds use models. Think about it, how does a seed know when to germinate? It has DNA, and encoded in that DNA is a set of instructions that tell it, “when you sense the temperature is x degrees, and there’s y amount of moisture, shoot your little stem thingy out and lets fucking grow.” That’s a model.
Models are simplified patterns representing reality. They’re a substrate for predictions. The factories from which predictions are built.
Humans have thousands, maybe millions of interconnected models in our heads that we draw upon to make predictions.
Prediction
Think about the act of knocking on a door:
You knock because you predict that the force of your knuckles colliding with the door will create a sound on the other side of the door.
Also, you’re predicting that someone on the other side of the door will hear that sound
And then you’re predicting that upon hearing the sound a person will walk over to the door and open it for you.
This activity, that’s so simple you can do it automatically, has at least 3 predictions nested inside of it. Predictions are everywhere, but we don’t think about them.
The good news is, we’re excellent at making predictions. Not at betting on NFL games type predictions, but tiny asinine predictions like taking steps. Before you take a step, you’ve made a prediction that taking this step moves you in a direction you want to go. You almost never trip when you’re walking because you’re so good at predicting where to step. You didn’t used to be. It took about a year of your little infant brain trying and getting it wrong before you starting getting good at these kinds of predictions.
As you learned and grew you got better at doing things by creating sharper and sharper predictions.
Action
When you order an Orange Fanta at Applebee’s, you can tell that you consciously decided, or predicted, that it is your best beverage option at that time and place. But even things like instincts, reactions, or involuntary movements, come out of some type of model and prediction. For example, humans have an instinct to recoil at the sight of a snake.
That instinct is a vestige of our evolutionary heritage where our primate ancestors were often bitten and then eaten by snakes. Humans are hard wired to do this without thinking. But even without a conscious prediction, this instinct still fits the model - prediction - action framework. The “Snake = dangerous” model is coded into us. Seeing, or thinking you see a snake is the prediction, and jump back is the action.
Similar Models
Delphiosis isn’t wildly original. It’s a stripped down version of other trusted models.
An OODA Loop is a military concept that describes the same model - prediction - action framework with slightly different language.
In OODA world they use two steps, “observe & orient” instead of just “Model”. The other difference is, OODA is circular. But not everything living things do is circular like this. For example, seeing a snake and leaping backwards won't impact your future reactions to seeing snakes.
Delphiosis aligns with the Scientific Method too:
Observing and researching sound like model building, right? And what is a hypothesis but a prediction. The scientific method has more steps, is more rigorous, and helps us create certainty about the nature of existence. It gives us a way of externalizing and replicating Delphiotic processes living things have always done. OODA and the Scientific Method use different wording, and have different applications, but there’s a similar pattern underlying both of them, Delphiosis.
“Truths” Are Just Fancy Predictions (Hold my Beer While I Pretend to be Smarter than Every Philosopher Ever)
Now let’s get crazy.
There are a multitude of different lenses through which you can view the world. You can look at everything as being either good or evil, right/wrong, Democrat/Republican, Yankees/Red Sox. Pick your poison.
Among the many available lenses the true/false lens is treated as being super consequential right now. “Misinformation”, “fake news”, and “gaslighting” are the hot buzzwords in America’s raging fight about what the truth is. This lens is less bad than lots of the others available, but it has a major downside. True/false claims depend on the authority of the source making the claim. As in, The FDA claims it to be true that all meat should be cooked to well done and that no one should eat raw oysters. They don’t say, “if you get fresh oysters from a reputable place and you’re not pregnant, you’ll probably be okay” they just say “nobody”. Losers. The Catholic Church says its true that Jesus Christ is risen and you better not masturbate. Your mileage may vary on those.
I hold it to be true that No Doubt is the greatest band ever, and if you disagree, you’ll be damned to hell where you’ll be forced to listen to the song “Bad Company” by the band Bad Company, from their super creatively named album, “Bad Company,” on repeat for all eternity.
Speed limits, drinking ages, weed illegality, sodomy laws, women’s suffrage, slavery. All laws are predicated upon the ideas that, there is a legitimate authority, that authority can tell what “The Truth” is, and that authority makes laws that comport with “The Truth.” But, as anyone who’s followed politics anytime, anywhere will tell you, there are a lot of bad, stupid laws on the books.
Truth is a pretty slippery idea when you look at it this way. There’s at least a dozen competing definitions of what constitutes “truth” and then there are an almost infinite number of authorities who claim to be in sole possession of “The Truth,” including Trump’s new app. When you watch the world with your true/false glasses on, you have to decide how trustworthy authorities are. And that’s hard to do.
When you see how often authorities happen to be wrong, you see “truths” for what they are: predictions. There’s no way a “truth” could not be a prediction, or have predictive implications. “Truth” statements are just predictions some group pretends are 100% certain.
I’m going to write more on what makes an authority more versus less trustworthy, but I think there’s three big factors: transparency, updates, and humility. To expand, we should be looking for authorities that are clear about the model they’re using to come up with predictions, that update their models when relevant new data comes in, and that can predict with humility about how certain they are. I don’t know if any authorities like that exist or not, but there’s something close.
Prediction Markets
Certainty that emerges from a market is more useful than truth that’s dealt out by an authority because of markets’ decentralized ability to form consensus.
Free markets can price things, including predictions. As more participants and knowledge flow into a market, prices get sharper. If a market is off, it creates an opportunity to profit by making a better prediction. Markets reset themselves in the face of new information, and free markets are mostly transparent. They’re far from perfect, but healthy, free markets are an amazing mechanism for fostering cooperation among people. All of that is to say, markets are closer to my description of a trustworthy authority than any other institution I can think of. If I’m correct, prediction markets can do everything a true/false binary can, but more.
From the delphiotic point of view, “The Truth” is the bet with the shortest odds. To elucidate, consider the following statement: “The sun rises in the east, and sets in the west.” From a true/false viewpoint that is a truth statement, or a fact. It gets entered into Wikipedia as a fact, and then it's up to Wikipedia readers to either trust or not trust that fact. From the Delphiotic perspective, “The Sun rises in the East” is a prediction that you can bet on tomorrow morning. The type of odds you would get for that prediction tell you how “true” it is, or how certain we are that it will happen as predicted. What kind of odds would you be willing to give someone who thought the sun would rise in the west tomorrow? 1000 to 1? 1000000 to 1?
As Scott Alexander has written about, existing prediction markets are flawed, and I don’t think you can actually get odds anywhere about the behavior of the sun. Either way, our ongoing societal fight about truth seems to be getting more and more vicious. To some degree this is because people on opposite sides of the political spectrum have wildly different models of how the world works. Its possible better prediction markets might create better shared models by showing people with bad models that their models suck. If you are betting with your terrible model against sharp models, you will eventually either go broke, or update your model. And, this doesn’t even require the politician or pundit to place bets themselves. As Unusual Whales has shown, you can model Jim Cramer or Nancy Pelosi’s predictions without their participation. Expository models like these have fantastic potential for keeping authorities honest.
We need better prediction markets, but that too is only the beginning. We need a much bigger infrastructure for cataloging, arraying, sorting, and connecting predictions. Stay tuned for more ideas about how to do this.
Black & White
There have been a spate of books in recent years warning about the dangers of black and white thinking. Kahneman’s “Thinking Fast and Slow,” Julia Gaelf’s “The Scout Mindset,” Annie Duke’s “Thinking in Bets,” and Jonathan Haidt’s, “The Righteous Mind” and “Coddling of the American Mind.” I don’t recall seeing in any of these books the idea that true vs false is just another instance of black and white thinking. It is though! And everyone must abandon truth or we’re all going to die! (See what I did there?)
I’m not suggesting we throw away truth, or the true/false lens, but we need to understand the limitations of viewing the world this way. We like truth because we want things to be simple. But that’s not what the world is like. Its messy and truth is hard.
I’m not arrogant, or stupid enough to think that I can settle a debate that the most famous philosophers of all time have been wrestling with. Truth as an idea, and in any particular case, is still super important, but also super elusive. Delphiosis is just a tool for trying to push aside some of our human shortcomings, so that we get our asses kicked by existence a little bit less.
Will it work? I don’t know.
What do you predict?
I'm with you on most of this, but you lost me with the societal battles. We're being torn apart for fun & profit. Here's an example of how like minded most Americans are. Somewhere very close to 10% of the people want to ban all abortion. Somewhere around 8% want to allow all abortion, all the way to 39 weeks. Around 70% say 15 weeks is OK, the balance is shortly thereafter.
Currently, one party holds all the cards at the federal level, owning the Senate, House, & White House. They could easily pass a law providing a legal right to abortion at 15 weeks. They'd have 70% of the population behind them. Do they do this? No, they stoke the battle for fun & profit.