Learn it → Apply it → Save as Skill
Every purposeful action ripples past its target. Alongside the effect you intended, a decision can shift incentives, change behavior or move a problem somewhere new — often in ways nobody planned for and nobody notices until later.
Dr. Wallace’s Favorite Mental Models is designed to turn useful ways of thinking into reusable tools. Each model gives you a fast visual overview, deeper interactive learning, practical application exercises and an AI-ready Markdown skill you can save for future use.
Instead of merely reading about mental models, you can gradually build your own AI skills library: a collection of portable reasoning skills that can help ChatGPT, Claude, Grok and other AI assistants apply the mental models you value.
Whenever you change a rule, incentive, system or habit, the people and systems around it adapt — sometimes the way you hoped, sometimes not. The law is a reminder to look past the direct effect before you act, not a prediction that things will go wrong.
Important: ripples aren't always negative — some are neutral, some are even beneficial. The model asks you to look for them, not to assume the worst.
Most useful when an action will pass through other people's incentives, habits or adaptive behavior — not just straightforward physical cause and effect.
The sequence pushes you from the outcome you want toward the outcome the whole system might actually produce.
State the intended effect precisely: what should change, for whom, and by when. A vague goal makes every later step harder.
Map the first-order effects: what happens directly and mechanically once the action is in place.
Probe second- and third-order effects: who else is affected, what incentives shift, and how people might adapt or work around the new rule.
Decide with contingency built in: name what you'll monitor and how you'll respond if a side effect shows up after launch.
Open a use case to see what the model changes.
A team introduces a bonus tied to one visible metric, such as call volume or tickets closed. Employees respond rationally to the incentive itself, not to the underlying goal it was meant to represent, and the measured number rises while the outcome it was supposed to reflect — genuine customer satisfaction — does not.
Rent control caps what landlords may charge, intended to keep housing affordable. Economists have widely documented a common side effect: when caps sit below market-clearing levels, landlords can under-maintain units, convert them to condominiums, or withdraw them from the rental market, effects that can tighten the very shortage the policy aimed to ease. The size of this effect varies by market and policy design and is debated, but the general pattern is well established.
Someone bans a habit outright — no sugar, no phone after 9pm — instead of setting a workable limit. The all-or-nothing rule can produce a secretive workaround or a bigger relapse than a moderate rule would have, because it leaves no room for the ordinary lapses any real routine includes.
A streak counter or daily-login reward is added to increase engagement. Usage goes up, but some users now open the app purely to protect the streak, and a feature meant to signal genuine value becomes a source of low-grade obligation for part of the user base.
A teacher grades homework on completion rather than accuracy to reduce stress. Turn-in rates rise, but some students learn that effort, not correctness, is what's rewarded, and stop checking their own work — the opposite of the deeper habit the policy was meant to build.
A team adds a cache in front of a slow database to cut response times. Latency improves, but many requests can now expire at once when a cache entry lapses, sending a sudden spike of simultaneous queries at the database — a new failure mode created by the very fix that removed the old one.
This is an illustrative example designed to teach the method.
Expand for the sociological origin and how this model relates to two others already in the series.
The systematic treatment is usually traced to sociologist Robert K. Merton's 1936 paper, "The Unanticipated Consequences of Purposive Social Action," which proposed a typology of ways purposive action can produce effects the actor did not intend — including outcomes that undercut the actor's own goal. Related intuitions appear earlier and elsewhere, including in economic writing on the unplanned effects of individual choices, but Merton's paper is the field's standard reference point for naming and organizing the idea.
It's worth separating two cases that get blurred together. Some consequences are genuinely hard to foresee given what was knowable at the time. Others are unintended only in the sense that nobody wanted them, but a modest amount of diligence — a pilot test, asking who else is affected, checking incentive alignment — would have surfaced them in advance. Treating every miss as equally unforeseeable lets the avoidable ones go unexamined.
The "cobra effect" names one specific pattern within this broader model: an incentive designed to fix a problem instead rewards behavior that recreates or worsens it — the label comes from an often-retold account of a bounty on cobras in colonial-era India said to have encouraged cobra breeding rather than reducing the wild population. Goodhart's Law — "when a measure becomes a target, it ceases to be a good measure," already covered elsewhere in this series — is a related but narrower principle about what happens specifically to metrics under optimization pressure. This model is the broader frame; the cobra effect and Goodhart's Law each describe one particular way it shows up.
"Everything has side effects" can become a reason to avoid deciding at all. The model is meant to inform a choice, not to justify postponing it indefinitely.
After the fact, it's easy to insist a ripple "was obviously predictable." Judge foresight by what was actually knowable before the decision, not by what became obvious once the outcome was in.
Calling a result "unintended" can quietly excuse a consequence that reasonable diligence would have caught. Unintended is not automatically blameless.
Some effects really are hard to predict. Many others are catchable with a short pilot, a pre-mortem, or simply asking who else the change touches. Don't skip the checkable ones.
A side effect nobody planned for can matter more than the goal that produced it. "We didn't mean for that to happen" is not a reason to ignore it.
The analysis is not a one-time checklist. Many ripple effects only become visible after the decision is live, which is why ongoing monitoring is part of the method, not an optional extra.
the chain from action to effect is short, mechanical and well understood; mapping ripples adds overhead without adding insight.
the fastest way to learn the real effects is to test at small scale and observe, rather than reason abstractly about what might happen.
a mature field already has well-documented second- and third-order effects for this exact kind of decision; redoing the analysis from scratch wastes existing expertise.
This browser tool structures your thinking. It does not send your information anywhere or perform AI analysis.
Save as Skill converts the model into portable Markdown instructions that teach an AI when and how to apply the Law of Unintended Consequences, what questions to ask, what evidence to seek, what mistakes to avoid and how to structure its analysis.
Save the skills you find useful and organize them into your own reusable library. The Markdown is intentionally model-neutral, so it can be used wherever an AI system accepts uploaded files, project knowledge, custom instructions or reusable text context. Platform capabilities differ, so this is a portable skill file rather than a claim of one-click installation.
No. Some are neutral and some are beneficial — a change aimed at one goal occasionally produces a better outcome than the one intended. The model is about noticing effects beyond the target, not assuming they must be negative.
"Side effect" is a broad, neutral term for any secondary effect of an action, often something a system is deliberately monitored for, as in medicine. "Unintended consequence," as used here, usually points at effects that emerge because other people or systems adapt to a change — through incentives, workarounds or feedback loops — rather than a fixed byproduct of the mechanism itself.
No. Complex systems with adaptive participants mean some effects are only visible after the fact, no matter how carefully you plan. The realistic goal is to catch the foreseeable ones, reduce the size of the surprises, and build in monitoring so you notice and adjust quickly when something unplanned shows up.