Synced 2026-04-04 17:14 UTC

Atoms

7 atomic claims in the graph. Each one is a reusable unit of thought with provenance, boundaries, and explicit capture context.

Last synced 2026-04-04 17:14 UTC 7 captures upstream 3 artifacts built so far
mechanism economics

Fast cognition optimizes for coherent stories before careful accuracy checks

Fast judgment produces an immediately usable interpretation by assembling a plausible story from partial cues.

“Paraphrase: quick thought builds the most coherent account it can from the evidence at hand.”
Reasoning and limits

Because: Speed comes from compressing evidence into a coherent narrative early, while slower checking happens later if it happens at all.

Boundaries: Useful for rapid orientation, but unreliable when the environment is noisy, adversarial, or statistically unintuitive.

Thinking, Fast and Slow 1 capture 1 related atom 1 artifact
new Open
heuristic economics

Cognitive fluency is often misread as evidence of truth

Ideas that are easier to process can feel more credible even when the underlying evidence is weak.

“Paraphrase: ease of processing can masquerade as evidence.”
Reasoning and limits

Because: Fluency reduces friction in judgment, and the mind treats that low friction as a cue for familiarity, safety, or correctness.

Boundaries: Fluency can correlate with truth in familiar settings, but it breaks when repetition, style, or framing create false confidence.

Thinking, Fast and Slow 1 capture 0 related atoms 0 artifacts
new Open
concept causal inference

Counterfactual reasoning is what makes causal models useful for explanation and choice

A causal model becomes decision-relevant when it can compare the observed world with plausible alternative worlds.

“Paraphrase: causal understanding reaches its highest utility when it can answer what would have happened otherwise.”
Reasoning and limits

Because: Intervention answers what happens when we act, but counterfactuals answer whether a different action or condition would have changed the outcome.

Boundaries: Counterfactuals only help when the underlying causal model is credible; observational correlations alone cannot support them.

The Book of Why 1 capture 0 related atoms 1 artifact
new Open
concept causal inference deep learning

Causal representations should be invariant across environments

Representations that capture true causal structure remain stable under distribution shift, unlike purely statistical features that exploit spurious correlations.

“Causal models can be seen as the correct abstraction level for generalizing across domains.”
Reasoning and limits

Because: The ICM principle states causal generative mechanisms are autonomous modules — changing one does not affect others. Representations aligned with these mechanisms inherit their invariance.

Boundaries: Assumes the causal graph is stable across environments. Breaks under structural causal changes. Also assumes we can identify the correct causal variables.

Toward Causal Representation Learning 1 capture 2 related atoms 1 artifact
new Open
mechanism causal inference

The Independent Causal Mechanisms principle: causal generative processes are modular and autonomous

Each mechanism in a causal system operates independently — changing one mechanism does not alter the others.

“The mechanisms of the causal generative model are autonomous and do not inform or influence each other.”
Reasoning and limits

Because: Nature's generative process factorizes into independent modules corresponding to edges in the causal graph. This is a structural assumption about how the world generates data.

Boundaries: ICM is a modeling assumption, not provable from data. In tightly coupled systems, mechanisms may not be cleanly separable.

Toward Causal Representation Learning 1 capture 1 related atom 1 artifact
new Open
mechanism causal inference deep learning

Multi-environment data provides the supervision signal for causal representation learning

Observing data across multiple environments provides the contrastive signal to identify causal vs. spurious features — causal features stay stable, spurious ones shift.

“Distribution shifts correspond to local interventions on the causal model, providing a natural supervision signal.”
Reasoning and limits

Because: Single-environment data is ambiguous — both causal and spurious features predict equally well. Multiple environments break this symmetry because only invariant features persist.

Boundaries: Requires sufficient diversity in environments. If environments only vary along non-informative dimensions, the signal is too weak.

Toward Causal Representation Learning 1 capture 1 related atom 1 artifact
new Open
critique causal inference deep learning

Disentanglement alone is insufficient without causal structure

Learning statistically independent latent factors does not guarantee that factors correspond to true causal variables or support interventional reasoning.

“Without further assumptions, unsupervised disentanglement is fundamentally impossible.”
Reasoning and limits

Because: Disentanglement methods optimize for statistical independence, but independent components can be rotated arbitrarily without changing the likelihood. Only causal structure breaks this symmetry.

Boundaries: If true causal variables happen to be statistically independent, disentanglement may approximately recover them. But this is a special case.

Toward Causal Representation Learning 1 capture 2 related atoms 1 artifact
new Open