The Three C's of Human Framing: Causality, Counterfactuals, Constraints
- Difficulty
- Moderate
- Time to result
- ~ongoing to results
- Steps
- 5
- Confidence
- —
The core model from Cukier's book Framers, offered as the answer to what humans can still do that machines cannot. Human reasoning runs on three building blocks. Causality is the innate, instinctive read of cause and effect — so strong that we see it where it does not exist, but it makes the world predictable and repeatable. Counterfactuals are the ability to fill in what is not there: you have never had molten metal poured on you, yet you run from it, playing the game two ticks ahead on information you do not have. Constraints are the discipline of narrowing an imagination that can render anything, down to the options that let you act in time. Together these form a mental model, or frame, which you can consciously swap when the situation changes.
Origin
Extracted from Deep Dive with Ali Abdaal
How to run it
- 1
Name the causal story you are running
State explicitly what you believe is causing what. Humans are primed to see cause and effect, which makes the world workable but also makes false causes feel obvious.
Watch out We are so good at causality that we constantly mistake it and get it wrong; treat your causal read as a hypothesis, not a finding.
- 2
Generate counterfactuals
Ask what would happen under conditions you have never experienced. Fill in the blanks from adjacent knowledge and play the situation two ticks ahead rather than waiting for the data.
Pro tip This is precisely the cognition machines lack — it is where a human adds value over a model trained only on what has already happened.
- 3
Apply constraints deliberately
Imagination can render anything, which is useless on its own. Cut the space of possibilities down to the most appropriate options for the decision you face, so that you can act inside the time you have.
Pro tip Constrain by the requirements of the given moment, not by habit — the right constraint at an earlier time may be the wrong one now.
Watch out Too few constraints produces fantasy; too many reproduces yesterday's frame.
- 4
Reframe when the situation changes
Adjust the model to the here and now. The point of holding a frame consciously is being able to swap it when the requirements of the moment differ from the ones the frame was built for.
- 5
Apply AI inside the frame, not instead of it
Use machines for what they exceed us at, while keeping the framing — cause, counterfactual, constraint — with the human. Stay the controller of the AI rather than the object of it.
Pro tip Cukier's test: are you a subject in this world or an object being pushed around in it?
Watch out Handing off cognitive functions rather than tasks is where the flexibility advantage is lost.
In the wild
Cukier's illustration of counterfactual reasoning: you have never had a hot furnace pour molten metal on top of you. You have no training data for it. Yet when the lava comes gushing out of the steel mill you run for your life, because you have filled in the information you do not have from information you do. He calls it a magical orchestrational form of cognition — machines do not have it, other animals do not have it, and it is what lets a person act correctly the first time a situation ever occurs.
→ Correct action in a scenario with zero prior examples — the exact case where a statistical model has nothing to infer from.
Cukier applies the frame-swap to his own life. A hundred and fifty years ago there were intellectual justifications for slavery and laws that enshrined it, and today that reads as bat crazy. So he asks what people in a hundred years will look back at his generation and find inexplicable. One candidate he names is buying water in a plastic bottle — maybe that is heinous and he simply cannot see it from inside the current frame. The exercise is deliberately turning the causal-counterfactual-constraint machinery on your own accepted defaults.
→ A repeatable self-audit that surfaces the assumptions your current frame makes invisible.
Common mistakes
Trusting the causal story because it feels obvious
The same instinct that makes the world predictable manufactures causes that are not there. Feeling certain about cause and effect is evidence of how the brain works, not evidence about the world.
Confusing unconstrained imagination with thinking
Rendering endless possibilities in the mind is not the skill; narrowing them wisely so you can act inside the available time is. Cukier treats constraint as a component of cognition, not a limitation on it.
Handing off the framing itself to AI
Applying AI to problems is a duty in his view; delegating the choice of frame is the failure. That is the difference between being the controller of the AI and being controlled by it.
From the transcript
“there's three things in particular the first is cause ity the second is counterfactuals and the third is constraints”
“the second is counterfactuals we fill in the blanks we see what isn't there”
“these three features of cognition causality counterfactuals constraints are the three sort of building blocks of our mental models of our frames”
From the episode
Economist's Insider Reveals How Humanity Can Outsmart Risks Of Artificial Intelligence