A laboratory for thinking clearly when the evidence is incomplete. Turn a hunch into a prediction. Draw what might cause what. Let your favourite explanation lose.
Everything runs in your browser. Your entries are not sent to a server. Exports are yours.
A construct is an idea; an operationalization is an observable rule. The distance between them is where many claims break.
A deliberately invented toy model. The coefficients and outputs are not estimated from people, and do not validate a psychological theory.
Try high intention with high friction. A simple model can make that contrast vivid, but cannot tell you whether its assumptions are true.
Add nodes, select two to draw a directional arrow, drag nodes to move them, or select one to remove it. Arrows represent hypotheses, not proof.
Select nodes with click or keyboard.
Accessibility alternative: each node is keyboard-selectable; use the buttons to connect and remove. Diagram relationships are also listed below.
Do not vote for your favourite story. Assign evidence-quality judgments, then see which explanation survives the strongest challenge.
Scenario: people who use a planning tool start more tasks. Why?
Higher rungs can support stronger inference, but quality, relevance, preregistration, sampling and bias matter at every level.
Evaluate multiple relevant studies and their bias, not just a pooled headline.
Randomization helps address confounding; replication tests whether results generalize.
Time ordering helps, but alternative causes remain.
Useful for patterns, usually weak for causal direction.
Rich for mechanisms and generating hypotheses; not automatic population estimates.
Good for asking questions; not independent confirmation.
This is a teaching heuristic, not a universal hierarchy. Different questions call for different methods.
Short enough to use. Substantive enough to disagree with.
If umbrella sales and wet pavements rise together, umbrellas need not cause rain. Ask about reverse causation, common causes, and what intervention would change the outcome.
Try: What would happen if we changed X while holding other causes stable?
A third factor can affect both exposure and outcome. People who voluntarily use a study tool may already be more motivated. Random assignment, careful design and sensitivity analyses help, but each has limits.
Try: Name one variable that predicts both the choice and the result.
“Motivation” cannot be observed directly like temperature. A survey, click count or diary samples only part of the construct. Reliability is consistency; validity asks whether the measure captures the intended meaning.
Try: What could this measure miss, distort or incentivize?
One striking finding may depend on chance, analytic choices or context. Repeating a method and obtaining comparable results strengthens confidence; a failed replication also needs careful interpretation.
Try: Could a stranger reconstruct the analysis from your documentation?
A test with 90% sensitivity and 90% specificity sounds decisive. But if only 1% of cases have the condition, among 1,000 cases you expect about 9 true positives and 99 false positives. Most positive results would be false positives in this illustrative scenario.
Try: Start with prevalence, then apply test accuracy.
A clever design can still be unethical. Consent, privacy, power differences, possible harm and the right to withdraw belong in the method, not an appendix. Do not collect sensitive personal data without a sound purpose and protections.
Try: Who bears the risk, and who gets the benefit?
Who was reachable, who opted in, and who dropped out? A large convenience sample can still be biased. Sampling design matters more than impressive sample size alone when generalizing to a population.
Try: Who is missing from the data, and why?
Original sources, carefully separated from the site's own interpretation.
Behavior is described as the convergence of motivation, ability and a prompt. It is an established published framework, not interchangeable with the toy friction equation above.
Stanford Behavior Design Lab — source ↗Ryan and Deci (2000) examine autonomy, competence and relatedness in motivation and psychological development. This is a research tradition with its own constructs and empirical literature, not proof of an Identity–Action Gap model.
Ryan & Deci (2000), original paper ↗The question of why intention and self-description diverge from action is a developing independent inquiry by Sahil Tayal. Its proposed terms and relationships are hypotheses; this site does not present them as a validated scale, established discovery or clinical assessment.
Explore the public atlas and its research entry ↗The displayed propensity comes from a transparent hand-authored formula. It uses no participants, no fitted coefficients, no observational dataset and no diagnostic interpretation.
2026-10-25 · Initial publication candidate. The model's empirical status is explicitly marked unvalidated. Source descriptions were checked against the original Stanford page and Ryan–Deci publication; site deployment, DNS and real-device testing remain unverified. No earlier corrections are asserted.
Find a mistake? Send a correction with the claim and source. Updates should be dated here when made.
“The point of a method is not to protect an idea from criticism. It is to make criticism informative.”
Questions belong to everyone. Conclusions must earn their confidence.
Return to the central atlas ↗