# Behavioral UX > A library of behavioral science principles for interface design, where every principle also ships as a plain markdown guide so an agent can read it without parsing HTML. Each principle has a page for people and a guide for agents at /agents/.md. Every guide carries the same sections in the same order: what it is, the evidence and the conditions under which it holds, how to apply it, what to avoid, live examples, references with DOIs, and links to related principles. Claims come from published research and are cited; the guides state the limits of each effect, not only its upside. Every guide carries a grade for the research behind it in an `evidence` frontmatter key — strong means replicated by independent groups or shown in large field experiments, moderate means it replicates with known moderators or modest effects, mixed means an inconsistent record or an effect that holds only under the conditions the guide spells out, and weak means a popular account resting on small, unreplicated or criticised work, where the effect may be near zero or reverse. A `categories` key on every guide names the product surfaces the principle applies to — onboarding, pricing, retention — so an agent can tell from the frontmatter whether a principle is likely to bear on the screen it is building. The headings below group the principles by subject, not by surface. ## Motivation - [Labor illusion](/agents/labor-illusion.md): People value a result more when they can see the work that went into it, even if it means waiting longer. - [Endowed progress effect](/agents/endowed-progress.md): People are more likely to finish a task when they feel they have already started it. - [Implementation intentions](/agents/implementation-intentions.md): People are more likely to follow through when they decide exactly when and where they will act. - [Fresh start effect](/agents/fresh-start-effect.md): People are more motivated to pursue goals at the start of a new week, month or year. - [Effort justification](/agents/effort-justification.md): After paying a real cost for something, people revise upward what they think it was worth — the cheaper conclusion is that the cost was wasted. - [Goal gradient effect](/agents/goal-gradient-effect.md): People push harder as a goal gets close, and slow down again the moment it resets. - [IKEA effect](/agents/ikea-effect.md): People put a higher value on something they built themselves — but only if they got it finished. - [Loss aversion](/agents/loss-aversion.md): Losing something you already hold can outweigh never getting it — but mainly at large stakes, and far less reliably than the textbook claim. - [Overjustification effect](/agents/overjustification-effect.md): Pay someone for something they were already doing for its own sake, and they do less of it once the payment stops. - [Sunk cost fallacy](/agents/sunk-cost-fallacy.md): People keep paying into something because of what they already paid, even when the money, time and effort already spent cannot be recovered and should not bear on the next decision. - [Variable reward](/agents/variable-reward.md): People repeat an action more when its payoff is uncertain and resolves at once; attached to money, the same mechanism drives gambling harm. - [Zeigarnik effect](/agents/zeigarnik-effect.md): Interrupted tasks pull people back to finish them, but they are not, on average, remembered any better than finished ones. - [Commitment and consistency](/agents/commitment-and-consistency.md): Once people have stated what they intend to do, they are somewhat more likely to actually do it. - [Foot-in-the-door](/agents/foot-in-the-door.md): Splitting one large ask into a sequence of small ones, on the theory that each small yes makes the next easier. - [Door-in-the-face](/agents/door-in-the-face.md): Ask for something large, get refused, then ask for something smaller — the smaller request is accepted more often than if asked alone. - [Temptation bundling](/agents/temptation-bundling.md): Attach something people enjoy to a task they keep putting off, and the task gets done slightly more often. - [Self-determination theory](/agents/self-determination-theory.md): People keep doing things that feel self-chosen for longer than things that feel required. - [Streak effect](/agents/streak-effect.md): Showing a run of consecutive days makes the next session a little likelier; showing a broken run makes it less likely than showing nothing. ## Decision making - [Default effect](/agents/default-effect.md): People tend to stick with the option that is preselected for them. - [Anchoring](/agents/anchoring.md): The first piece of information people see shapes how they judge everything after it. - [Left-digit effect](/agents/left-digit-effect.md): A price of 2.99 feels much cheaper than 3.00, even though the difference is one cent. - [Choice overload](/agents/choice-overload.md): Too many options can stop people from choosing at all, but only under certain conditions. - [Framing effect](/agents/framing-effect.md): The same fact can lead to different choices depending on how it is worded. - [Hick's law](/agents/hicks-law.md): The time it takes to pick one option grows with the number of options, but slowly — with the logarithm, not the count. - [Decoy effect](/agents/decoy-effect.md): A third option that is plainly worse than one of the first two can make that one more popular. - [Status quo bias](/agents/status-quo-bias.md): People leave things as they are — the plan, the setting, the provider they already have — even when changing would cost them nothing. - [Scarcity](/agents/scarcity.md): A product that looks like it is about to run out is wanted more than the same product in plentiful supply. - [Pain of paying](/agents/pain-of-paying.md): Paying stings, and the sting fades when the money is hard to see — so a saved card tends to buy a slightly bigger basket. - [Reactance](/agents/reactance.md): Push hard enough towards a choice and people get angry at you — but on a screen the anger and the compliance usually arrive together. - [Hyperbolic discounting](/agents/hyperbolic-discounting.md): A reward that arrives now outweighs a bigger one later — strongly when the cost is effort, barely at all when it is money. - [Mental accounting](/agents/mental-accounting.md): People keep money in separate mental pots, so the same price feels affordable or extravagant depending on which pot it lands in. - [Compromise effect](/agents/compromise-effect.md): The same option wins more choices from between a cheaper and a dearer alternative than from the edge of the set. - [Partitioned pricing](/agents/partitioned-pricing.md): People remember the base price, not the fees, so hiding the total lifts revenue — but splitting the price on its own does almost nothing. - [Regret aversion](/agents/regret-aversion.md): People avoid the option they expect to feel worst about afterwards, and what they will later find out decides which option that is. ## Social - [Social proof](/agents/social-proof.md): When people are unsure what to do, they look at what others like them are doing. - [Reciprocity](/agents/reciprocity.md): People feel obliged to give something back when they receive something first. - [Authority bias](/agents/authority-bias.md): People give more weight to a claim when it arrives with a credential, and spend less effort checking the claim itself. - [Social comparison](/agents/social-comparison.md): When there is no objective yardstick, people work out how they are doing by looking at how other people are doing. - [Descriptive norms](/agents/descriptive-norms.md): Telling people what others really do nudges their own behaviour towards that number — by one to five percentage points, and sometimes the wrong way. ## Memory - [Peak-end rule](/agents/peak-end-rule.md): People remember an experience by its most intense moment and its ending, not by its length. - [Von Restorff effect](/agents/von-restorff-effect.md): In a uniform set, the one item that looks different is recalled better — and the items around it slightly worse. - [Cognitive load](/agents/cognitive-load.md): Working memory is small, so an interface that makes people hold things in their head loses to one that keeps those things on screen. - [Serial position effect](/agents/serial-position-effect.md): Asked to recall a list, people return the first few items and the last few, and lose the middle. - [Recognition over recall](/agents/recognition-over-recall.md): People recognise the right option in a set far more reliably than they produce it from memory, so put the options on screen. - [Picture superiority effect](/agents/picture-superiority-effect.md): People recall pictures better than the words for the same things, so a concept the user has to carry to a later screen is worth drawing. - [Availability heuristic](/agents/availability-heuristic.md): People judge how common or likely something is by the instances they can bring to mind, so whatever is easy to retrieve looks frequent. ## Perception and attention - [Aesthetic-usability effect](/agents/aesthetic-usability-effect.md): People rate an attractive interface as easier to use than an identical plain one, so polish buys a better verdict, not a better flow. - [Processing fluency](/agents/processing-fluency.md): When something is easy to read and easy to follow, people judge it, and the people behind it, more kindly. - [Banner blindness](/agents/banner-blindness.md): Anything parked in a slot where promotions live gets skipped, so a notice that matters should not sit there. - [Law of proximity](/agents/law-of-proximity.md): Things placed close together are read as belonging together, and a gap is read as a break between them. - [Habituation](/agents/habituation.md): An alert that fires the same way often enough stops being an alert, and the fix is a different one rather than a louder one. ## Interaction - [Fitts's law](/agents/fitts-law.md): The time to hit a target grows with distance and shrinks with size, logarithmically: the first millimetres of size buy far more than the last. - [Progressive disclosure](/agents/progressive-disclosure.md): Showing only what is needed now and keeping the rest one tap away simplifies the first screen, and makes almost everything behind the tap invisible. - [Doherty threshold](/agents/doherty-threshold.md): People work faster when the system answers almost immediately, and delay costs in proportion to milliseconds rather than at one magic number. ## Judgment and bias - [Negativity bias](/agents/negativity-bias.md): A bad moment in your product counts for more than a good one, so fixing what hurts moves more than adding delight. - [Planning fallacy](/agents/planning-fallacy.md): People predict how long a job will take by imagining it going to plan, so long and unfamiliar jobs come out too short. - [Mere exposure effect](/agents/mere-exposure-effect.md): Seeing something a few times makes people like it slightly more — and seeing it far too often undoes that again. ## Pages - [Guide index](/agents/index.md): Every agent guide in one list. - [Principles library](/principles/): All principles, each with a with/without demo of the interface change. - [MCP server](/mcp/): Planned. Will serve every principle as a structured spec. Not live yet. ## Optional - [Home](/): What the project is, for human readers.