spreadlab/internal/engine/cascade_test.go
Justin Visser a0635544d9 engine: additive forward probability, behaviour-preserving
Forwarding was a single global ForwardProb; make it a per-student
composite (Config.ForwardChance): baseline propensity raised by the
fake's Novelty, lowered by ambient HarmAwareness, and scaled down for an
educated student by ProgramEffect (1 = today's hard block). RunCascade
now takes a precomputed per-student []float64 chance and has no education
special case. Defaults are behaviour-neutral, so the 82/58/6 golden
tests are unchanged; the model is tuned in a later slice.
2026-06-18 20:03:04 +02:00

113 lines
3.5 KiB
Go

package engine
import (
"slices"
"testing"
)
// The cascade is deterministic once thresholds are fixed, so these tests
// hand-craft a tiny line graph (0-1-2-3) and exact thresholds: no
// randomness, every expectation is exact.
func lineGraph(numNodes int) *Graph {
graph := NewGraph(numNodes)
for node := 0; node < numNodes-1; node++ {
graph.AddEdge(node, node+1)
}
return graph
}
// uniformThresholds gives every directed edge the same threshold.
func uniformThresholds(graph *Graph, value float64) EdgeThresholds {
thresholds := make(EdgeThresholds)
for node := range graph.NumNodes() {
for _, neighbor := range graph.Neighbors(node) {
thresholds[[2]int{node, neighbor}] = value
}
}
return thresholds
}
// uniformChance gives every student the same forwarding chance.
func uniformChance(numNodes int, value float64) []float64 {
chances := make([]float64, numNodes)
for node := range chances {
chances[node] = value
}
return chances
}
func TestRunCascadeSpreadsAlongOpenEdges(t *testing.T) {
graph := lineGraph(4)
thresholds := uniformThresholds(graph, 0.1) // 0.1 < 0.5: every edge forwards
result := RunCascade(graph, 0, uniformChance(4, 0.5), thresholds)
wantRounds := []int{0, 1, 2, 3} // one hop further each round
if !slices.Equal(result.ReachedAtRound, wantRounds) {
t.Errorf("ReachedAtRound = %v, want %v", result.ReachedAtRound, wantRounds)
}
if result.NumReached != 4 {
t.Errorf("NumReached = %d, want 4", result.NumReached)
}
if result.NumRounds != 4 {
t.Errorf("NumRounds = %d, want 4", result.NumRounds)
}
}
func TestRunCascadeThresholdBlocksOneDirection(t *testing.T) {
graph := lineGraph(4)
thresholds := uniformThresholds(graph, 0.1)
// Close the 1 -> 2 direction only. The open 2 -> 1 direction must not
// matter: student 2 never gets the fake, so never forwards anything.
thresholds[[2]int{1, 2}] = 0.9
result := RunCascade(graph, 0, uniformChance(4, 0.5), thresholds)
wantRounds := []int{0, 1, NeverReached, NeverReached}
if !slices.Equal(result.ReachedAtRound, wantRounds) {
t.Errorf("ReachedAtRound = %v, want %v", result.ReachedAtRound, wantRounds)
}
if result.NumReached != 2 {
t.Errorf("NumReached = %d, want 2", result.NumReached)
}
}
func TestRunCascadeEducatedReceivesButDoesNotForward(t *testing.T) {
graph := lineGraph(4)
thresholds := uniformThresholds(graph, 0.1)
// Student 1 is educated by a full-strength program: forwarding chance 0.
chance := uniformChance(4, 0.5)
chance[1] = 0
result := RunCascade(graph, 0, chance, thresholds)
// Student 1 still receives the fake in round 1, but the chain stops there.
wantRounds := []int{0, 1, NeverReached, NeverReached}
if !slices.Equal(result.ReachedAtRound, wantRounds) {
t.Errorf("ReachedAtRound = %v, want %v", result.ReachedAtRound, wantRounds)
}
}
func TestNewEdgeThresholdsDeterministicAndComplete(t *testing.T) {
graph, err := HolmeKim(30, 3, 0.45, newRand(17))
if err != nil {
t.Fatal(err)
}
first := NewEdgeThresholds(graph, newRand(2))
second := NewEdgeThresholds(graph, newRand(2))
if wantSize := 2 * graph.NumEdges(); len(first) != wantSize {
t.Errorf("len(thresholds) = %d, want %d (two directions per edge)", len(first), wantSize)
}
for directedEdge, firstDraw := range first {
if firstDraw < 0 || firstDraw >= 1 {
t.Errorf("threshold %v = %v, want in [0, 1)", directedEdge, firstDraw)
}
if secondDraw := second[directedEdge]; firstDraw != secondDraw {
t.Errorf("threshold %v differs across identical seeds: %v != %v",
directedEdge, firstDraw, secondDraw)
}
}
}