psychology-map
Skill19 starsChanged 9 days agoMITView on GitHub
Map 30 behavioral, perceptual, conversion, and trust mechanisms with evidence ratings and failure modes. Sub-skill: load only when routed from the hook-everything entry skill, not for general writing or editing.
In this repository14 other files it gives its agents.
- AGENTS.mdAGENTS.md
- hook-everythingSkill
- anti-slopSkill
- audience-calibrationSkill
- creative-ideationSkill
- hook-compoundingSkill
- marketing-hooksSkill
- market-intelligenceSkill
- product-hooksSkill
- self-reviewSkill
- trust-architectureSkill
- value-checkSkill
- visual-ux-hooksSkill
- writing-hooksSkill
The file
Files
SKILL.md
Psychology Map (30 Mechanisms Across 6 Layers)
Select mechanisms based on the user's exact cognitive state. Always state the Evidence Level (STRONG, MIXED, WEAK) when explaining a design choice.
Evidence Rating Key
STRONGindicates support from independent replications or meta-analyses with consistent directional effects.MIXEDindicates a real theoretical construct or practical framework with contested replication, strong moderators, or indirect transfer to marketing and UX.WEAKindicates failed independent replication in modern meta-analyses, included here to prevent misuse.
Layer 1: Attention, Curiosity & Creative Bisociation
| ID | Mechanism | Evidence | Core Psychology | Invisible Application | Fails / Backfires When |
| M1 | Information Gap (Loewenstein 1994) | STRONG | Curiosity triggers when attention focuses on a specific gap between what we already know and what we want to know. | Anchor a question in facts the audience already half-knows, exposing one missing causal link. | Audience has zero baseline context, the gap is too wide, or the answer is trivial. |
| M2 | Narrative Transportation (Green & Brock 2000) | STRONG | Immersion in a concrete scene with sensory detail and human stakes absorbs cognitive bandwidth and lowers counter-arguing. | Open inside a specific moment of action or failure before stating any abstract thesis. | The story feels invented or disconnected from the core utility. |
| M3 | Cognitive Fluency & Strategic Disfluency (Alter & Oppenheimer 2009) | MIXED | High processing ease (fluency) signals safety and truth; a small anomaly (disfluency) can wake up analytical attention, though disfluency effects replicate inconsistently. | Keep UI and prose effortless to scan, and insert one counter-intuitive metric or contrast to stop skimming. | Everything is disfluent (causes confusion) or everything is overly smooth (induces autopilot skimming). |
| M4 | Schema Violation / Asymmetric Specificity | MIXED | The brain predicts the next token or event; exact, unrounded numbers or unexpected pairings break prediction and lock focus (practitioner heuristic). | Use forensic specifics ("4:55 PM on a Friday", "14 lines of SQL", "$84 left") instead of vague summaries. | Specificity is irrelevant trivia that does not connect to the payoff. |
| M23 | Bisociation (Cross-Matrix Collision) (Koestler 1964) | MIXED | Conceptual creativity model where originality occurs when two previously unrelated, self-consistent frames of reference collide to solve a single problem. | Frame a familiar domain problem through the operational mechanics of an orthogonal field (e.g., ER triage applied to incident response). | The analogy is a superficial cliche rather than a structural mechanism match. |
Layer 2: Perceptual, Spatial & Visual UX Attention (See skills/visual-ux-hooks)
| ID | Mechanism | Evidence | Core Psychology | Invisible Application | Fails / Backfires When |
| M27 | Von Restorff Isolation Effect (Von Restorff 1933) | STRONG | When multiple homogeneous stimuli are presented, the single item that differs in form or spacing is attended to and remembered far better. | Keep 90% of the layout or typography calm and neutral so the single primary action or key metric stands out effortlessly. | Bolding or badging 5 different elements on the same screen (visual noise cancels isolation). |
| M28 | Choice Overload & Hick's Law (Iyengar & Lepper 2000; Hick 1952; Scheibehenne et al. 2010) | MIXED | Decision time increases logarithmically with options (Hick); excessive uncategorized choices can paralyze action (30% vs 3% is the original 2000 jam study result only; Scheibehenne et al. 2010 meta-analysis found a near-zero mean effect across 50 studies unless options are uncategorized). | Cap visible decision points to 1 Primary Action + 1 Escape Hatch, or pre-filter options via a 1-click diagnostic. | Dumping 12 navigation links, 5 pricing tiers, or 4 competing CTAs in the hero section. |
| M29 | Doherty Threshold (<400ms) vs. Labor Perception (>1.5s) (Doherty & Thadani 1982; Buell & Norton 2011) | MIXED | Direct manipulation requires <400ms response to sustain flow; high-stakes background analysis can benefit from 1.5s-3s visible step-by-step transparency. | Make UI toggles and previews instant (<400ms), and show explicit multi-step progress when running deep diagnostics. | Adding artificial delays to basic UI clicks, or returning a complex audit in 10ms with zero explanation of what was checked. |
Layer 3: Action, Momentum & Behavioral Conversion
| ID | Mechanism | Evidence | Core Psychology | Invisible Application | Fails / Backfires When |
| M5 | Fogg Behavior Model (B=MAP) (Fogg 2019) | MIXED | Practical diagnostic model stating action happens when Motivation, Ability (low friction), and a Prompt converge at the exact same moment. | Increase action by cutting steps (Ability) and aligning Prompts with existing habits before trying to pump Motivation. | Spamming more prompts when Ability (friction) is the real blocker. |
| M6 | Endowed Progress (Nunes & Dreze 2006) | MIXED | People given a credible head start toward a goal completed it at higher rates (34% vs 19% in field study). | Credit the user for steps already completed via auto-detection or import, stating the exact reason. | Progress bar is padded with fake steps the user sees through. |
| M7 | Goal-Gradient Hypothesis (Hull 1932; Kivetz et al. 2006) | STRONG | Effort accelerates exponentially as distance to a visible milestone shrinks. | Break long onboarding or long reads into short, visibly closeable milestones ("1 step left to live preview"). | The finish line is hidden or moves backward after action. |
| M8 | Default & Status-Quo Architecture (Johnson & Goldstein 2003) | STRONG | People overwhelmingly stick with pre-selected defaults because switching costs cognitive effort and perceived risk. | Ship working defaults, pre-loaded sample data, and sensible configurations out of the box. | Defaults opt users into unwanted billing or privacy exposure. |
| M24 | Self-Perception & Foot-in-the-Door (Bem 1967; Freedman & Fraser 1966; Burger 1999) | MIXED | People infer their own commitment by observing their voluntary micro-actions, though meta-analytic FITD effects are small and sensitive to perceived pressure. | Invite a tiny, zero-risk, high-value micro-action first (e.g., running a 1-click check or toggling a parameter) before asking for signup or purchase. | The first step feels like a trap or requires personal data before delivering value. |
| M25 | Mental Accounting & Zero-Price Affect (Thaler 1985; Shampanier, Mazar & Ariely 2007) | MIXED | People evaluate costs inside separate mental budgets; a $0 utility item reduces downside risk, though zero-price offers can attract low-intent users without qualification boundaries. | Offer an un-gated, zero-price micro-tool that solves step 1 completely, and anchor paid tiers against an existing "weekly waste budget" rather than "new software spend." | Using "Free" as a bait-and-switch where the free output is locked behind a surprise paywall. |
Layer 4: Memory, Ownership & Return Loops
| ID | Mechanism | Evidence | Core Psychology | Invisible Application | Fails / Backfires When |
| M9 | Ovsiankina Task Resumption (Ovsiankina 1928; Ghibellini & Meier 2025) | STRONG | An interrupted task the user still intends to complete creates an intrusive urge to resume. Quieted by completion or a concrete plan (Masicampo & Baumeister 2011). | Save exact cursor/workspace state and offer a 1-click resume link or a 1-click "schedule for later" button. | Turning resumption into needy guilt notifications ("We miss you"). |
| M10 | Zeigarnik Recall Claim (Zeigarnik 1927; 2025 Meta-Analysis) | WEAK | Often cited as "unfinished tasks are remembered better," but a 2025 meta-analysis found no reliable memory advantage outside original data. | Never cite Zeigarnik as proof that cliffhangers boost memory. Use M9 (Ovsiankina) or M1 (Information Gap). | Used to justify withholding basic answers behind cheap cliffhangers. |
| M11 | Generation Effect (Slamecka & Graf 1978) | STRONG | Information a person actively generates or infers themselves is remembered and valued far higher than passive reading. | Give 90% of a logical chain so the reader connects the final 10% in their own mind; let users tweak an input on screen 1. | Asking the user to generate an answer when they lack the context to succeed. |
| M12 | IKEA Effect & Co-Creation (Norton, Mochon & Ariely 2012) | MIXED | Investing manageable labor into assembling or customizing an artifact increases valuation in lab studies, provided the user succeeds cleanly. | Let users configure 2-3 high-impact building blocks that visibly shape their project before asking for commitment. | The assembly task is too complex and the user fails (ownership drops to zero on failure). |
| M13 | Peak-End Rule & Duration Neglect (Fredrickson & Kahneman 1993) | STRONG | Retrospective judgment of an experience is governed by its single best/worst peak moment and its final moment, not total length. | Engineer one unmistakable "aha" payoff and end every piece of writing or session on a crisp, resonant note. | Ending a piece with a repetitive summary paragraph or ending a trial with a surprise paywall. |
| M14 | Hook Loop & Variable Utility (Eyal 2013; Ferster & Skinner 1957) | MIXED | Trigger -> Action -> Variable Reward -> Stored Investment that loads the next internal trigger (practical framework). | Provide variability through fresh, genuinely useful signal (new anomalies, faster workflows) and compounding user assets. | Using empty extrinsic slot-machine mechanics (badges/streaks) with no utility. |
Layer 5: Defense Bypass, Cognitive Routing & Regret Regulation
| ID | Mechanism | Evidence | Core Psychology | Invisible Application | Fails / Backfires When |
| M15 | Psychological Reactance & Autonomy Framing (Brehm 1966) | STRONG | When people sense their freedom to choose or think is being pushed, they resist; explicitly affirming their freedom to refuse disarms resistance. | Explicitly state who should not use the product or read the guide, and include visible zero-guilt exit options. | Used as manipulative reverse psychology ("Don't click this link"). |
| M16 | Self-Determination & Identity Motivation (Ryan & Deci 2000; Oyserman 2007) | MIXED | Intrinsic action relies on Autonomy / Competence / Relatedness, and sticks when congruent with the person's craft identity (difficulty-as-importance). | Mirror the user's competence and frame manageable effort as a badge of serious craft. | Talking down to the audience or using aspirational lifestyle fluff that clashes with their real identity. |
| M17 | Persuasion Knowledge Model (PKM) (Friestad & Wright 1994) | STRONG | Once an audience recognizes a tactic as a "persuasion attempt," they shift from message processing to defense/skepticism mode. | Strip all marketing signposts (exclamation marks, superlatives, artificial countdowns) so the tactic never triggers PKM recognition. | Using templates or hype formulas the audience has already seen a dozen times. |
| M18 | Loss Aversion on Owned Assets (Kahneman & Tversky 1979; Gal & Rucker 2018) | MIXED | Owned losses weigh heavier than hypothetical gains, though general loss aversion is often confounded with status-quo inertia (Gal & Rucker 2018). | Quantify real leakage in the user's current workflow or show what they have already built in the sandbox before asking to save. | Inventing fake scarcity ("Only 2 digital seats left"). |
| M26 | Regret Regulation & Process Justification (Zeelenberg 1999; Zeelenberg & Pieters 2007) | MIXED | Buyers hesitate when they anticipate Process Regret (self-blame for making an unjustifiable or gullible decision). | Reduce anticipated regret at conversion by providing clear decision justification (exact benchmarks, failure bounds, and 1-click self-serve reversal). | Pressuring the user with "Buy now before you regret missing out" (increases anxiety and process regret). |
| M30 | Elaboration Likelihood (ELM) & Construal Routing (Petty & Cacioppo 1986; Trope & Liberman 2010) | ELM = STRONG; Construal Level Theory = MIXED | High-motivation/ability audiences process via the Central Route (demanding concrete, low-construal forensic proof), while low-ability/skimming audiences enter via Peripheral Cues (note: for CLT, publication-bias-adjusted estimates are smaller than the original literature). | Match hook density and construal level to the reader's current route (see skills/audience-calibration). | Giving fluffy peripheral slogans to a Central-Route engineer, or dumping a 40-line table on a cold mobile scroller. |
Layer 6: Deep Trust Architecture (See skills/trust-architecture)
| ID | Mechanism | Evidence | Core Psychology | Invisible Application | Fails / Backfires When |
| M19 | ABI Trust Triad (Mayer, Davis & Schoorman 1995) | STRONG | Trust is the willingness to accept vulnerability based on Ability (competence), Benevolence (user-first motive), and Integrity (honest principles). | Prove Ability with live demos, Benevolence by warning against over-buying, and Integrity with transparent pricing/limits. | High Ability paired with low Benevolence (smart product that traps users in dark patterns). |
| M20 | Pratfall Effect (Aronson, Willerman & Floyd 1966; Helmreich et al. 1970) | MIXED | A competent entity can become more trusted after admitting a small, real flaw, though Helmreich et al. (1970) showed the effect depends on observer traits and baseline competence. | Pair proof of high competence with an upfront admission of a real limitation or past engineering blunder. | Admitting flaws before competence is established (makes the creator look amateur). |
| M21 | Inoculation Theory (Two-Sided Refutation) (McGuire 1961; Banas & Rains 2010) | STRONG | Exposing an audience to a weakened version of a counter-argument and refuting it builds lasting immunity against future skepticism. | Raise the reader's strongest objection yourself in paragraph 2 ("You might wonder why we didn't just use Postgres...") and answer it with data. | Raising a strawman objection instead of the real doubt in the user's mind. |
| M22 | Operational Transparency & Costly Signaling (Buell & Norton 2011; Spence 1973) | MIXED | Costly signaling (un-fakeable effort) builds credibility, while operational transparency ("labor illusion") increases value in specific waiting tasks. | Expose the exact steps or benchmarks running under the hood instead of hiding behind vague "AI magic" claims. | Faking slow loading spinners when no real computation or verification is happening. |
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.