The question is not whether a model sounds intelligent. The question is: what part of reality is it good at seeing, what does it miss, and how can we test what it predicts?
The most dangerous model is often a useful model applied where it does not belong.Reality Skill is partly knowing which tool to reach for, and when to put it down.
A forecasting model, a human-values lens and a systems map should not be judged as if they were interchangeable. First identify the job. Then choose the tool.
Scientific method, Bayes, causal inference, forecasting.
Systems, constraints, incentives, values, games.
Alternatives, opportunity cost, information value, robustness.
Red team, inversion, AQAL, de Bono, premortem.
Start with the decision. Identify what job needs to be done. Only then choose the model.
A library of clever labels. Every model should create a better question, a prediction, or a choice.
This library is arranged by job, not prestige. The goal is not to collect smart-sounding frameworks. The goal is to choose a tool that improves what you see, ask, decide or verify.
These are the most foundational models in Reality Skill. They help replace intuitive certainty with testable beliefs, calibrated confidence and clearer causal thinking.
Turn beliefs into hypotheses, define what would count against them, test, and update.
Start with what was plausible before, weigh new evidence, and update proportionally instead of jumping from certainty to certainty.
Separate correlation from intervention. Ask what would likely have happened if the supposed cause had not occurred.
Make testable probability estimates, score them against outcomes, and learn whether your 70% really behaves like 70%.
Before explaining why your case is special, look at comparable cases and their base rates.
Decision models matter once alternatives are real, resources are limited and uncertainty remains. Their job is not to predict perfectly, but to improve the action chosen now.
A choice is not good in isolation. Compare it with the best realistic alternative, including doing nothing.
Research only the uncertainties capable of changing the action. More information is not automatically more value.
Attack the plan before reality does. Invert the goal and identify failure modes, hidden dependencies and preventable stupidity.
When the same problems return, the explanation is often structural. These models help you move from isolated events to constraints, feedback loops and leverage points.
Look for feedback loops, delays, stocks, flows, rules, information and leverage points behind recurring events.
Complexity often contains leverage. Find the bottleneck that governs throughput before optimizing everything else.
Metrics reshape behavior. Distinguish the thing you care about from the proxy you happen to measure.
Break a huge uncertain question into estimable parts. Often one or two variables dominate the result and tell you what to research.
Human behavior rarely has one cause. Incentives, values, identity, personality, culture and strategic interaction all matter. The goal is better hypotheses, not faster labeling.
People respond to money, status, power, safety, convenience and social rewards, often more reliably than to stated intentions.
Use values as motivational hypotheses about priorities such as achievement, security, benevolence, autonomy and tradition.
Personality traits can improve expectations about typical behavior, but situations still create substantial variation.
Useful for hypotheses about what feels legitimate, important, threatening or admirable. Treat the taxonomy as a lens, not ground truth.
Check individual interior, individual exterior, collective culture and collective systems before accepting a one-dimensional explanation.
When actors respond strategically to one another, model incentives, information, credible commitments and repeated interaction.
Some models help you move when certainty is impossible. Others help you generate alternatives or verify whether the inputs to your reasoning deserve trust in the first place.
Observe, orient, decide, act, then reorient as new evidence arrives. Learning speed can matter more than first-try perfection.
Clear, complicated, complex and chaotic contexts demand different decision modes. Sometimes analysis is right; sometimes you need probes and adaptation.
Deliberately generate alternatives and change perspectives before optimizing inside the first frame that appeared.
Leave the source, inspect provenance, triangulate independent coverage, and trace claims back toward primary evidence.
The skill appears when you can use the right model on a consequential decision, then return later and see whether reality agreed.