Security Nudge Lab 2D: Decision Architecture Simulator

Nudge Parameters

Friction (extra confirm step) 20%

Warning salience 30%

Social-proof pressure 40%

Agent spawn rate 6/s

Max concurrent agents 120

Live Behavioral Metrics

Compromise 0%Safe 100%

Click-through (risky) rate

—

Agents processed

Model p(risky) now

—

Avg. decision latency

—

How it works

Each agent that reaches the decision gate (a suspicious attachment / link prompt) makes a binary choice — open it (risky) or decline (safe) — sampled from a logistic behavioral model driven by four security-UX "nudges" from behavioral-economics decision architecture (Thaler & Sunstein):

p(risky) = σ( b0
   − wF · friction
   − wS · salience
   + wD · default_risky
   + wP · social_proof )

σ(z) = 1 / (1 + e^(−z))

Friction — an extra confirmation step before opening reduces risky uptake (loss of momentum / System-2 engagement).
Warning salience — a visually prominent warning (color, icon, size) suppresses risky choices via attention capture.
Default option — whichever action requires zero clicks (open vs decline) is disproportionately chosen — the well-documented default effect.
Social-proof pressure — a claim like "94% of colleagues opened this" pushes choices toward the risky action, illustrating how attacker-side social engineering can hijack the same lever defenders use.

Decision latency is modeled as increasing with friction (more steps = more time) plus gate-vs-decision travel time. Every agent is an independent Bernoulli trial with probability p computed live from the current slider state — this is the same class of model security-UX teams use to A/B test phishing-warning and consent-dialog designs before shipping them.

This 2D build is a genuine top-down reimplementation of the same x/z flow field the 3D original uses (agents only ever move in a ground plane; the 3D version's vertical axis is a cosmetic bob) — drag to pan the flow view and scroll/pinch to zoom, exactly like orbiting the 3D camera. Two extra analytics panels — a live logistic response-curve plot and a decision-latency histogram — make the same math legible without a 3D renderer.

Approaching agent Declined (safe) Opened (compromised)

drag to pan · scroll/pinch to zoom

Logistic response curve

Decision-latency histogram