alerts-feed mechanic asserted
Alerts feed
A notification layer surfaces discrete events and threshold crossings with severity and a link to their location.
definition
The feed subscribes to simulation events (raid arrived, agent died, contract completed) and to condition predicates over state (food below a threshold, building unpowered, storage full, entity damaged). It owns a list of active alerts, each with type, severity, timestamp, a deduplication key, a location or entity link, and a lifetime. Each tick it adds an alert on the rising edge of a predicate or on an event, merges alerts with the same key, removes condition alerts when the condition clears, and expires event alerts after a timeout or dismissal. Optional rules pause the game or drop speed on chosen types. Canonical forms: RimWorld's severity-colored event letters plus a persistent condition list (low food, colonist needs rescue), Factorio alerts for entities under attack or destroyed, trains out of fuel, and ghosts lacking construction materials, problem icons over buildings and the Chirper feed in Cities: Skylines, Dwarf Fortress announcements, and the Football Manager inbox.
player experience
In a system too large to watch, the feed decides where the player looks. The player triages, jumps the camera to the alert, and learns to dread specific alert types. Auto-pause on alerts turns real-time play into a series of decision points.
design levers
- event alerts versus persistent condition alerts that clear themselves
- severity tiers, color, and sound per alert type
- auto-pause or speed-drop rules per alert type
- aggregation: one alert for many identical problems versus one per entity
- persistence until dismissed versus until the condition clears
- click-through to location and player-side filtering
- in-fiction framing (social posts, letters) versus neutral system text
failure modes
- Alert fatigue: frequent low-value alerts bury the one that matters, and players miss a fatal problem.
- In-fiction feeds read as noise; many Cities: Skylines players hid the Chirper feed with mods.
- Predicates near a threshold, such as power just at capacity, flicker and spam add and remove cycles.
- Alerts that report symptoms (no goods) rather than causes send players to fix the wrong thing.
dynamics
- The alert set encodes what the designer considers a problem, so players optimize to keep icons off rather than toward their own goals.
- With auto-pause, alert density sets the effective pacing of a real-time game.
- Paired with an event director, alerts announcing incident arrival create a recognizable beat structure of calm, alarm, and response.
open questions
- Can a learned salience model rank and merge alerts by predicted consequence so the feed stays useful at megabase or metropolis scale?
- Do natural-language digests of many alerts help players more than icons, or do they slow reaction time?
ratings
12 from named runs; stored scores are their lower medianoriginal score
An alert explicitly names a problem or event, marks severity and links its location. These immediate visible functions match glance-level transparency; complex causes behind an alert belong to the reported systems and do not lower this layer to 4.
original score
The feed deterministically reports events and predicates without owning their causes. Aggregation and auto-pause shape attention and pacing, but they do not themselves generate unscripted simulation variation deserving 2.
original score
Alert types, message templates, severity and lifetimes fit small tables. This is more than rule-only content (1), while rich in-fiction writing is optional rather than required moderate bespoke content for every alert variant (3).
original score
Condition checks, active-alert maintenance and deduplication are low-cost linear work at typical entity scale. This exceeds a negligible single-state read (1), but is not ordinarily a late-game slowdown source (3).
original score
Thresholds, persistence, deduplication and auto-pause interact and need playtesting to prevent flicker or fatigue. This exceeds a few stable parameters (2), without normally producing sensitive runaway strategies at level 4.
original score
Contextual ranking, causal grouping and generated explanations can notably improve relevance and presentation variety beyond incremental alert tuning (2). Event notification remains the form, so substantially new mechanics at 4 are not required.