Most feed-ranking algorithms don't just count total likes โ they weight engagement velocity: how much interaction a post gets, discounted by how long ago it happened. A simple model of that recency-weighted score S is:
dS/dt = -kยทS (exponential decay between events)
S โ S + 1 (on each new engagement)
where k is the decay rate slider. A high, fast-arriving burst of engagements keeps S elevated, which the platform reads as "trending" and pushes to more feeds.
Automation exploits this directly: bots connected to the account fire near-instantly on publish (latency slider), injecting an artificial early spike in S before any human has even seen the post. That inflated score then raises the probability that real human followers, seeing an already-"popular" post, engage themselves โ a contagion cascade:
P(human engages) = clamp(pโ + S / S_scale, 0, 0.95)
pโ is the organic-engagement-rate slider โ the baseline chance with no algorithmic boost at all. Each engaged node (bot or human) then propagates the same test to its own unengaged neighbours, so the cascade spreads outward through the follower graph exactly like an epidemic on a contact network.
- Bot ratio โ fraction of the network that is automated; more bots means a bigger, faster initial score spike.
- Bot response latency โ how quickly automation fires after publish; real bot networks are near-instant, which is one of the signals platforms use to detect them.
- Organic engagement rate โ the baseline human interest with zero algorithmic help.
- Decay rate โ how fast the score's "trending" boost fades if new engagement stops arriving.
Real-world relevance: this velocity + decay + contagion structure is why platforms increasingly flag accounts whose early engagement arrives too fast and too uniformly to be organic โ the same visibility mechanism that automation exploits is also the fingerprint that detects it.