Do you have questions or comments about this model? Ask them here! (You'll first need to log in.)

---

---

WHAT IS IT?

This simulator models how opinions spread and evolve in a connected population (multi-agent system). Each agent has an opinion (−1 to +1), a prevalence (0–99), an influence (0–1), and social links. The model tracks the co-evolution of opinions, prevalence (depth of internal representations), influence, and the network structure.

3D Representation

• X: opinion (−1 left, +1 right)
• Y: prevalence (0–99)
• Z: influence (0–1) Colors: blue (right), red (left), yellow (meta-influencer). Links: green (same sign), gray (opposite signs).

---

HOW TO USE

1. Choose the population size with pop.
2. Click Setup (creates agents, black background, initializes tick-event to event-init).
3. Click Go to run/pause.

---

GENERAL CONTROLS

• Setup: initialize agents/network; sets tick-event ← event-init.
• Go: start/stop the simulation.
• in_file: load agent states from a file (for data).
• refresh: resets plots after ~200 ticks.
• cumulative: if OFF, resets change/total every tick.
• output: None | Statistics | Values | File.

---

POPULATION & ITERATIONS

• pop: number of agents (e.g., 100–5000).
• nb_try, max_iter, threshold: repetitions, trial length, majority threshold.
• choice_iter: iteration to replay when loading from file.

---

SOCIAL NETWORK (link dynamics)

• link-removal-threshold: opinion distance (in %) above which a link may be cut.
• link-formation-threshold: maximum distance to allow a new link.
• prob: probability applied to deletions/formations.
• linksdown / linksup: caps on links removed/created per tick.
• bridge-prob: chance to create bridges across opposing camps.
• show-links?: toggle link visibility; linktick: visual thickness.

---

META-INFLUENCERS

Agents with fixed high influence (influence = 1).

• meta-influencers-selection: All / Right side / Left side.
• meta-influencers: share of agents promoted to “meta”.
• prev-low / prev-high: prevalence eligibility bounds.
• meta-min / meta-max / meta-links: min/max and current quota of meta links.
• meta-ok: enables dynamic meta linking even if vary-influence = OFF.
• vary-influence: if ON, influence increases after successes and decreases after setbacks.
• metablock: if ON, metas cannot flip sign (veto on polarity changes).

At initialization (setup) and at the start of each new trial, the model sets meta-links ← meta-min if vary-influence = true or meta-ok = true (as in your code).

---

OPINION & PREVALENCE DYNAMICS

• rate-infl: speed of influence updates after adoption.
• noise: probability of additive opinion noise.
• polarization-factor: penalizes adoption across large opinion gaps.
• prevalence-weight: weight of prevalence differences in adoption.
• adoption-floor: minimum adoption probability (avoids strict zero).

Prevalence modulation (renamed in code)

• mod-prev (formerly modulation-prevalence): ON to adapt prevalence to current vs previous opinion.
• Rate-mod (formerly rate-modulation): adjustment intensity.

---

GROUP EFFECT

• group-impact-mode: all (all linked neighbors) or k-nearest.
• group-k: number of neighbors in k-nearest mode.
• group-impact-weight: weight of neighborhood alignment in adoption.

• group-impact-alpha: non-linearity

  • <1: small aligned clusters matter more,
  • =1: linear,
  • >1: only large aligned majorities matter.

---

REWARD MECHANISM

A successful emitter (who convinces a neighbor) receives a temporary bonus (tx-bonus) that boosts future persuasion.

• reward-step: bonus increment per success.
• reward-cap: cap on cumulative bonus.
• reward-scope: both / left-only / right-only.
• reward-prev-delta: increase in the target’s prevalence after adoption (optional).
• reward-decay: bonus decay over ticks.

---

Meme-Based Representation (Weighted Memes + Targeted Injection)

When use-memes? is ON, opinions and prevalence are computed from internal “meme” stocks.
This version distinguishes meme quantity (how many representations an agent holds) from meme weight (how strongly each representation shapes the opinion).

---

1) Two levels: Quantity vs Weight

Meme quantity → Prevalence

• meme-plus, meme-minus store how many pro/anti memes the agent holds.
  
• Prevalence (0–99) is derived from the total quantity:
  meme-plus + meme-minus rescaled to 0–99.

Interpretation: prevalence approximates how “rich” an agent’s internal representation system is (how many arguments/frames are available).

Meme weight → Opinion

• meme-plus-w, meme-minus-w store the cumulative weighted strength of pro/anti memes.
  
• Opinion is computed from the weighted balance:

[ opinion = \frac{meme\text{-}plus\text{-}w - meme\text{-}minus\text{-}w}{meme\text{-}plus\text{-}w + meme\text{-}minus\text{-}w} ]

• If pro-weights dominate → opinion moves toward +1
• If anti-weights dominate → opinion moves toward −1
• If balanced → opinion stays near 0

A tiny denominator safeguard is used to avoid division by zero.

---

2) Weighted transmission during interactions

When an agent is influenced, the receiver gets: - a quantity increment (meme-gain) on the side of the emitter (plus or minus), - and a weight increment proportional to that quantity.

Weight distribution parameters

• meme-weight-mean (typical 0.2–3.0): average strength of newly acquired memes.

  • Low values → many memes are needed to polarize opinions.
    
  • High values → opinions polarize faster, even with small meme quantities.

• meme-weight-sd (typical 0.0–1.0): heterogeneity (memes differ in strength).

  • 0.0 → all memes are equally strong.
    
  • Higher values → mixed populations with “weak” and “strong” memes.

• meme-weight-min / meme-weight-max: hard bounds preventing unrealistic weights.

---

3) Meme anti-leak and decay (optional)

• meme-anti-leak (0–1): when one side grows, a fraction of the opposite stock is reduced.
  High values create “winner-takes-more” dynamics (polarization reinforcement).

• meme-decay (0–0.05 typical): forgetting rate applied each tick to quantities and weights.

---

Meme Injection (Targeted diffusion of new representations)

Beyond “events” that shift opinions directly, the simulator can inject memes into a selected subgroup to simulate the introduction and diffusion of a new narrative, argument, or frame.

1) Targeting bounds (preferred controls)

Agents are eligible for injection if they satisfy:

• inject-low_meme ≤ opinion ≤ inject-high_meme
• inject-low-prev ≤ prevalence ≤ inject-high-prev

This allows injection into: - moderates only (e.g., −0.2 to +0.2), - one camp only (e.g., +0.2 to +1), - low-prevalence agents only (e.g., 0 to 30).

2) Injection strength and reach

• inject-prob-max (0–1): maximum share of eligible agents that actually receive the injected memes.
• inject-sign: "plus" or "minus" (which direction the injected memes support).
• inject-amount (typical 1–10): how many memes are injected (quantity → raises prevalence).
• inject-weight (typical 0.2–5.0): how strong injected memes are (weight → shifts opinion more sharply).

Rule of thumb:
- Increase inject-amount to raise prevalence (more representations).
- Increase inject-weight to raise polarization intensity (stronger conviction shift).

3) Scheduling injection

• auto_inject?: if ON, injection occurs when ticks = inject-tick.
• inject-tick: the tick at which injection happens.
• repeat-inject?: if ON, injection repeats every inject-pace ticks.
• inject-pace: the interval between repeated injections.

Example scenarios

• One-shot targeted campaign:
  auto_inject? = ON, repeat-inject? = OFF, inject-tick = 50, inject-prob-max = 0.2,
  inject-low_meme = -0.2, inject-high_meme = 0.2, inject-sign = "plus",
  inject-amount = 3, inject-weight = 2.0.

• Slow diffusion of a weak but persistent narrative:
  repeat-inject? = ON, inject-pace = 25, inject-prob-max = 0.05,
  inject-amount = 1, inject-weight = 0.5.

• High-impact shock on a small subgroup:
  inject-prob-max = 0.02, inject-amount = 2, inject-weight = 5.0,
  targeting a narrow opinion band.

---

Parameter Ranges — Key Sliders (Overview)

The table below summarizes the recommended value ranges for the nine main sliders governing opinion transmission, meme dynamics, and external shocks.
Ranges are indicative and meant to support exploratory, comparative, and pedagogical simulations rather than strict calibration.

| Slider name | Typical range | What it controls | Low values → expected effects | High values → expected effects | |------------|---------------|------------------|-------------------------------|--------------------------------| | prevalence-weight | 0.0 – 2.0 | Weight of prevalence differences in adoption probability | Opinion change weakly tied to representational depth; more random diffusion | Deeply rooted agents dominate transmission; strong inertia | | adoption-floor | 0.0 – 0.10 | Minimum probability of adoption regardless of distance | Near-impossible cross-camp adoption | Persistent low-level noise and occasional inversions | | polarization-factor | 0.0 – 1.0 | Penalty applied to large opinion gaps | Distance barely matters; smooth convergence | Strong ideological barriers; entrenched camps | | group-impact-weight | 0.0 – 1.0 | Strength of group alignment effect | Individual interactions dominate | Local majorities strongly condition adoption | | group-impact-alpha | 0.2 – 3.0 | Non-linearity of group effect | Small minorities exert strong influence | Only large aligned groups matter | | meme-max | 50 – 200 | Maximum stock of memes per agent (prevalence scale) | Shallow belief systems; fast saturation | Deep ideological accumulation; slow dynamics | | meme-gain | 0.5 – 2.0 | Meme increment per successful transmission | Slow learning; weak reinforcement | Rapid ideological buildup | | meme-anti-leak | 0.0 – 0.5 | Cross-erosion of opposite meme stock | Memes accumulate independently | Strong competition; winner-takes-more dynamics | | meme-decay | 0.0 – 0.05 | Forgetting rate of memes per tick | Stable long-term memory | Rapid erosion; volatile belief systems | | event-prob-max | 0.0 – 1.0 | Proportion of agents affected by an event | Micro-perturbations, local shocks | System-wide shocks |

---

Usage notes

• Low–mid ranges are recommended for exploratory runs and sensitivity analysis.
  
• Extreme values are useful to study boundary cases (lock-in, collapse, polarization).
  
• Parameters interact strongly: e.g., high meme-gain combined with high meme-anti-leak accelerates polarization.

This table is intended as a cheat sheet; empirical calibration should rely on systematic parameter sweeps.

Meme Injection — Slider Ranges, Effects, and Interactions

This table documents the nine sliders governing meme injection.
Together, they define when, how often, to whom, and with what strength new memes are introduced into the population, allowing controlled simulations of campaigns, rumors, or polarizing shocks.

---

Core Injection Parameters

| Slider | Typical range | Role in the model | Low values → effects | High values → effects | |------|---------------|-------------------|----------------------|----------------------| | inject-tick | 1 – max_iter | First tick at which meme injection can occur | Early priming of agents; long-run structural impact | Late shock; short-term perturbation with limited propagation | | inject-pace | 1 – 200 | Interval (ticks) between injections | Continuous or quasi-continuous pressure | Rare, punctuated shocks | | inject-prob-max | 0.0 – 1.0 | Maximum probability that an eligible agent receives a meme at an injection tick | Injection remains marginal or localized | Broad exposure; near-systemic dissemination | | inject-amount | 0 – meme-max | Quantity of meme stock added per injection | Subtle informational nudges | Strong narrative saturation or propaganda burst | | inject-weight | 0.0 – 2.0 (or higher) | Relative impact of injected memes on opinion vs. prevalence | Injected memes mainly increase prevalence (attention/salience) | Injected memes strongly bias opinion formation |

---

Opinion Dynamics Simulator – “Inject Metas Only” Extension

Purpose of the Extension

This version of the model introduces an option controlled by a switch (inject-metas-only) that allows restricting meme injection to meta-agents only, as they exist at the moment of injection.

Meta-agents are defined operationally as agents colored yellow, that is, agents explicitly selected by the influenceurs button.

The goal of this extension is to enable targeted meme injection scenarios without altering the endogenous opinion dynamics or the standard meme mechanisms.

---

General Principle

• Meme injection continues to rely on the inject-memes procedure.
• A new switch, inject-metas-only, optionally filters the target agents.
• When the switch is ON, only yellow agents present at the time of injection can receive memes.
• When the switch is OFF, the injection behavior is strictly identical to the original version of the model.

No additional distinction is introduced between different types of meta-agents: the visual criterion (yellow color) is used as the sole reference.

---

Detailed Operation

1. Selection of Target Agents

The inject-memes procedure first builds a base pool of agents, defined by the existing bounds:

• opinion ∈ [inject-low_meme, inject-high_meme]
• prevalence ∈ [inject-low-prev, inject-high-prev]

This pool corresponds exactly to the historical behavior of the model.

2. Optional “Meta-only” Filtering

If the inject-metas-only switch is activated:

• the pool is restricted to agents colored yellow (color = yellow)
• meme injection applies exclusively to those agents

No additional conditions are imposed.

3. Interaction with vary-influence

• The inject-metas-only switch remains operational regardless of the state of vary-influence.
• There is intentionally no software safeguard preventing both options from being enabled simultaneously.
• The analytical interpretation of such configurations is left to the user.

This design choice ensures maximum experimental flexibility while preserving fully deterministic model behavior.

---

Relevant Code (Conceptual Excerpt)

let base-pool turtles with [
  opinion >= inject-low_meme and opinion <= inject-high_meme and
  prevalence >= inject-low-prev and prevalence <= inject-high-prev
]

let pool base-pool

if inject-metas-only [
  set pool base-pool with [ color = yellow ]
]


### Targeting and Selectivity Parameters

| Slider | Typical range | Role in the model | Low values → effects | High values → effects |
|------|---------------|-------------------|----------------------|----------------------|
| **inject-low_meme** | −1.0 – 1.0 | Lower bound of opinion eligible for injection | Broad targeting across ideological spectrum | Injection limited to one side or a narrow ideological niche |
| **inject-high_meme** | −1.0 – 1.0 | Upper bound of opinion eligible for injection | Narrow ideological window | Broad or opposing-side reach (depending on bounds) |
| **inject-low-prev** | 0 – 99 | Minimum prevalence required to receive injected memes | Inclusion of low-salience or weakly engaged agents | Targeting already attentive or mobilized agents |
| **inject-high-prev** | 0 – 99 | Maximum prevalence eligible for injection | Focus on low-to-mid engagement agents | Restriction to highly engaged elites |

---

### Interaction Effects (Key Dynamics)

- **inject-tick × inject-pace**  
  - Early + short pace → *persistent campaign dynamics*.  
  - Late + long pace → *isolated shock events*.

- **inject-prob-max × inject-amount**  
  - Low prob + high amount → *elite seeding* (few agents, strong impact).  
  - High prob + low amount → *mass diffusion* (many agents, weak signal).

- **inject-weight × inject-amount**  
  - High weight + high amount → *rapid opinion polarization*.  
  - Low weight + high amount → *agenda-setting without strong persuasion*.

- **inject-low_meme / inject-high_meme × inject-low-prev / inject-high-prev**  
  - Narrow opinion + high prevalence → *echo-chamber reinforcement*.  
  - Broad opinion + low prevalence → *grassroots diffusion potential*.

---

### Conceptual Interpretation

- Low values generally model **background noise**, rumors, or weak informational exposure.  
- High values approximate **organized campaigns**, disinformation bursts, or polarizing media events.  
- Intermediate combinations allow exploration of **threshold effects**, diffusion delays, and nonlinear amplification.

These sliders are designed to be orthogonal but **non-independent**: meaningful experiments emerge from their **joint configuration**, not isolated tuning.


## Preset — *Campaign / Rumor / Polarizing shock* (recommended default profile)

This preset is a **ready-to-run parameter profile** designed to reproduce a common empirical pattern in opinion dynamics:

1) a **stable baseline** (moderate homophily, limited cross-camp contact),  
2) a **rumor/campaign phase** (repeated external shocks reaching only part of the population),  
3) a **polarizing outcome** (network segmentation, fewer inversions, stronger within-camp reinforcement).

It is intended as a **starting point** for university-level experimentation (reproducible and interpretable), not as a calibrated model for a specific case.

---

### A. Core idea
- Use **repeat-event + event-pace** to create a *campaign/rumeur* that reappears periodically.
- Use **event-prob-max < 1** so the shock reaches only a subset of eligible agents, generating diffusion rather than an immediate global collapse.
- Keep **bridge-prob low but non-zero** so rare cross-camp bridges exist (possible inversions), while still allowing polarization to emerge.
- Activate **group impact** and a **mild reward system** to reproduce realistic “social proof” and reinforcement mechanisms.

---

### B. Default values (Preset)
*(Values assume your current sliders/switches; ranges remain adjustable by the operator.)*

#### 1) Events: “campaign / rumor” mechanics
- **auto_event**: ON  
- **repeat-event**: ON  
- **event-init**: `50` (first shock after an initial stabilization period)  
- **event-pace**: `25` (shock repeats every 25 iterations)  
- **event-prob-max**: `0.10` (≈10% of eligible agents receive the shock each cycle)

**Bounds & targeting**
- **meme_set**: OFF *(use bounds rather than structural Left/Right set)*  
- **low_meme / high_meme**: `-0.30 / +0.30` *(targets the “convertible middle”)*  
- **low-prev / high-prev**: `10 / 60` *(targets moderate-prevalence agents—reachable but not rigid)*  

**Shock magnitude**
- **event_size**: `0.25` *(clear movement without saturating ±1 too fast)*  
- **prev_change**: `+8` *(moderate strengthening of prevalence in targeted agents; set 0 if you want “pure opinion shock”)*  

**Interpretation**
- This reproduces a campaign/rumeur that repeatedly perturbs the **middle** rather than only extremes, and spreads indirectly through the network.

---

#### 2) Network: controlled homophily + rare bridges
- **network**: ON  
- **link-formation-threshold**: `0.20`  
- **link-removal-threshold**: `0.40`  
- **prob**: `0.30`  
- **linksup / linksdown**: `10 / 10` *(balanced churn)*  
- **bridge-prob**: `0.05` *(rare but non-zero cross-camp bridges)*  
- **show-links?**: optional (OFF for performance; ON for demonstrations)

**Expected effect**
- The network remains mostly homophilous, but with occasional “bridges” allowing limited cross-camp exposure.

---

#### 3) Opinion adoption: prevalence-driven, polarization-aware
- **prevalence-weight**: `1.40`  
- **polarization-factor**: `0.60`  
- **adoption-floor**: `0.03`  
- **noise**: `0.01` *(small background drift)*

**Expected effect**
- Adoption is mostly driven by **prevalence differences**, while strong polarization reduces cross-camp adoption without making it impossible.

---

#### 4) Group impact: social proof (moderate strength)
- **group-impact-mode**: `k-nearest`  
- **group-k**: `8`  
- **group-impact-weight**: `0.60`  
- **group-impact-alpha**: `1.20`

**Expected effect**
- Local neighbourhood alignment matters; **majorities** have slightly more weight than minorities (alpha > 1).

---

#### 5) Reward: modest reinforcement (avoid runaway)
- **reward-step**: `0.03`  
- **reward-cap**: `0.30`  
- **reward-decay**: `0.005`  
- **reward-scope**: `both`  
- **reward-prev-delta**: `0` *(keep prevalence effects attributable to memes/events; set 1–3 if you want “success breeds conviction”)*

**Expected effect**
- Successful influencers become moderately more effective over time, but decay prevents permanent dominance.

---

#### 6) Memes: ON (recommended for this preset)
- **use-memes?**: ON  
- **meme-max**: `120`  
- **meme-gain**: `1.0`  
- **meme-anti-leak**: `0.20`  
- **meme-decay**: `0.01`

**Expected effect**
- Agents gradually accumulate representations; anti-leak creates a mild competitive relation between pro/anti stocks, supporting polarization over repeated shocks.

---

#### 7) Meta-influencers: optional, controlled
- **meta-ok**: ON *(optional; ON reproduces real-world asymmetric reach)*  
- **meta-influencers-selection**: `All`  
- **meta-influencers**: `5%`  
- **prev-low / prev-high**: `20 / 80`  
- **meta-min / meta-max**: `8 / 20`  
- **metablock**: ON *(prevents sign switching for metas)*  
- **vary-influence**: OFF *(recommended OFF to avoid meta inflation; turn ON only if studying endogenous influencer emergence)*

**Expected effect**
- A small set of highly connected agents accelerates diffusion while the metablock prevents metas from oscillating across camps.

---

### C. What you should observe (typical outcomes)
- Early phase (before **event-init**): moderate clustering, limited movement.
- During repeated events: a growing asymmetry in meme stocks and prevalence among the targeted “middle”.
- Over time: stronger within-camp reinforcement, more **gray links** disappearing, fewer inversions.
- If you increase **event-prob-max** toward 0.30–0.50: faster and more global shifts.
- If you increase **bridge-prob** toward 0.10–0.20: more cross-camp exposure and more inversions (polarization weakens).

---

### D. Minimal variant (if you want “shock-only”, no campaign)
- Set **repeat-event = OFF** and keep a single **event-init** (or press **event** once).
- Keep all other values unchanged to compare *one-shot shock* vs *campaign repetition*.


---

## Meme Dynamics — Integrated Monitoring Indicators

This simulator includes **six dedicated monitors** designed to track how *memes* (internal representations) shape opinion formation, prevalence, and polarization over time.  
Each monitor is updated **at every tick** and provides a complementary analytical perspective on the meme–opinion coupling.

---

### 1. Mean Meme Stock
**What it measures**  
The average total number of memes held by agents:

Mean Meme Stock = mean(meme-plus + meme-minus)

**Why it matters**  
This indicator captures the **global cognitive density** of the population.

- Low values indicate weakly structured opinions (few internal representations).  
- High values reflect ideologically “loaded” agents with many arguments.

**Interpretation**  
- Rising values → accumulation of representations (learning, persuasion, repeated events).  
- Falling values → forgetting or erosion (via `meme-decay`).

**Implementation (monitor expression)**  

---

### 2. Meme Polarity Index
**What it measures**  
The **net ideological balance of all memes** in the system.

It compares the total stock of positive memes to negative memes.

**Interpretation**  
- Values close to `+1` → dominance of pro (+) memes.  
- Values close to `-1` → dominance of contra (−) memes.  
- Values near `0` → balanced or plural meme ecology.

**Implementation (monitor expression)**  

---

### 3. Opinion–Meme Gap
**What it measures**  
The average absolute difference between:
- the agent’s expressed opinion, and  
- the opinion implied by its internal meme balance.

**Why it matters**  
This indicator captures **latent cognitive inconsistency**:
agents may *express* an opinion that is not fully supported by their internal representations.

**Interpretation**  
- Low gap → opinions are well grounded in memes.  
- High gap → cognitive tension, instability, or transitional states.

**Implementation (monitor expression)**  

---

### 4. Ideologization Index
**What it measures**  
The degree to which **strong opinions are backed by high prevalence**.

It combines opinion extremity and representational depth.

**Interpretation**  
- High values → polarized and ideologically entrenched population.  
- Low values → pragmatic or weakly structured opinion landscape.

**Implementation (monitor expression)**  

---

### 5. Right Meme Polarization

**What it measures**  
The **internal ideological polarization of meme stocks** among agents holding a **positive opinion** (`opinion >= 0`).

It compares reinforcing memes (`meme-plus`) to counter-memes (`meme-minus`) **within the right-leaning subgroup only**, independently of the left side.

**Interpretation**  
- Values close to `+1` → strong dominance of reinforcing (pro-right) memes.  
- Values near `0` → balanced or internally plural meme structure.  
- Values below `0` → erosion or counter-meme dominance inside the right camp.

This indicator reveals whether right-side opinions are:
- deeply ideologized,
- stable but weakly supported,
- or vulnerable to counter-narratives.

**Implementation (plot and monitor expression)**  


---

### 6. Left Meme Polarization

**What it measures**  
The **internal ideological polarization of meme stocks** among agents holding a **negative opinion** (`opinion < 0`).

It measures the balance between reinforcing and counter-memes **inside the left-leaning subgroup**, independently from right-side dynamics.

**Interpretation**  
- Values close to `-1` → strong dominance of reinforcing (pro-left) memes.  
- Values near `0` → plural or contested internal meme structure.  
- Values above `0` → counter-meme pressure within the left camp.

This indicator is essential to detect:
- asymmetric polarization,
- minority radicalization,
- or differential resistance to meme diffusion.

**Implementation (plot and monitor expression)**

--- 

### 7. Mean Meme-derived Opinion

**What it measures**  
The **average opinion value reconstructed from meme stocks**, independently of the agents’ current explicit opinions.

This measure computes, for each agent, the opinion implied by the balance of its internal memes  
(`meme-plus` vs `meme-minus`), then averages this value across the whole population.

**Interpretation**  
- Values close to `+1` → the meme ecology strongly supports the positive (right/pro) polarity.  
- Values close to `-1` → dominance of negative (left/contra) meme representations.  
- Values near `0` → balanced or internally conflicting meme structures.

Comparing this indicator with the mean explicit opinion reveals **latent ideological tensions**:
- If meme-derived opinion is stronger than explicit opinion → opinions may soon shift.  
- If explicit opinion is stronger than meme-derived opinion → opinions are weakly grounded and fragile.


**Why it matters**  
This indicator captures the **deep cognitive orientation** of the population, beyond surface-level expressed opinions.  
It helps distinguish:
- superficial opinion alignment,
- structurally grounded ideological states,
- and phases where internal representations lag behind observable behavior.

---

**Implementation (monitor expression)**  

---

### 8. Meme Saturation (%)

**What it measures**  
The **overall saturation level of meme stocks** in the population, expressed as a **percentage of the maximum possible meme capacity**.

It compares the total number of memes currently held by all agents to the theoretical maximum (`meme-max × population`).

**Interpretation**  
- Values close to `0%` → agents hold very few internal representations; opinions are shallow and weakly grounded.  
- Intermediate values (`30–60%`) → active circulation of arguments with moderate cognitive load.  
- Values close to `100%` → saturated cognitive environment; agents are highly entrenched and resistant to change.

This indicator captures the **cognitive density** of the system and helps distinguish:
- early diffusion phases,
- mature ideological ecosystems,
- and saturation-driven stabilization or lock-in effects.

**Implementation (monitor expression)**  



*(These indicators are computed only when `use-memes?` is ON.)*

---

# EXOGENOUS EVENTS (bounded & probabilistic)

**Targeting**

* **`meme_set`** + **`to_left`**: if ON, structural targeting by camp (*Left side* / *Right side*).
* Otherwise, use bounds: **`low_meme` / `high_meme`** (opinion window) and **`low-prev` / `high-prev`** (prevalence window).

**Effects on targeted agents**

* **`event_size`**: opinion shift (toward the intended camp).
* **`prev_change`**: prevalence change (clamped to [0,99]).
* **`event-prob-max`** *(0–1)*: **maximum share** of targeted agents that **actually receive** the shock (each agent draws `U(0,1)`).

**Triggering**

* **`event` button**: one-shot shock (manual).
* **`auto_event` + `tick-event`**: scheduled automatic shock at iteration `tick-event`.

### Repeated events (per your code)

* **`event-init`**: **initial offset** of the first event (on `setup` and at each new trial, `tick-event ← event-init`).
* **`repeat-event`** (switch): if ON, **re-schedules** the next event after each occurrence.
* **`event-pace`** (≥ 1 tick recommended): spacing between repeated events.
* **Scheduling logic**:

  * If **`auto_event` = ON** and **`iter = tick-event`** → run `event`.

    * If **`repeat-event` = ON** → `tick-event ← tick-event + event-pace`.
    * Else (OFF) → **no** automatic re-scheduling (you may adjust `tick-event` manually).
  * If **`auto_event` = OFF** → **each tick**, `tick-event ← iter + event-pace` (the next time you switch ON, the event fires ≈ `event-pace` ticks later).

**Quick examples**

* Single calibration shock: `auto_event=ON`, `repeat-event=OFF`, `event-init=2`, `event-prob-max=1.0`.
* Periodic pulses: `auto_event=ON`, `repeat-event=ON`, `event-init=50`, `event-pace=50`, `event-prob-max=0.30`.
* Diffuse perturbations: `repeat-event=ON`, `event-pace=100`, `event-prob-max=0.05`.

---

# OUTPUTS / MONITORS / CSV

* **Monitors**: % left/right, medians (opinion/prevalence/influence), inversions, interactions, fractal dimension, links created/removed.
* **Graph**: time trajectories of key variables.
* **CSV**: if **`csv-export`** = ON, per-trial export with a standard header (`basename-try.csv`).

---

# THINGS TO WATCH

* Polarization, convergence, fragmentation.
* Roles of **meta-influencers** (and `metablock`), **group effect**, and **reward**.
* Impact of **memes** (memory, cross-leak, decay).
* How **repeated events** and **`event-prob-max`** shape the global dynamics.

---

# QUICK CHEAT SHEET — TYPICAL VALUES

| Parameter               | Useful range | Tendency                                                    |
| ----------------------- | ------------ | ----------------------------------------------------------- |
| `prevalence-weight`     | 0–2          | ↑ makes prevalence gaps dominate adoption                   |
| `adoption-floor`        | 0–0.1        | ↑ allows more “noisy” cross-camp adoptions                  |
| `bridge-prob`           | 0–0.3        | ↑ creates more cross-camp bridges & inversions              |
| `group-impact-weight`   | 0–1          | ↑ strengthens neighborhood alignment effect                 |
| `group-impact-alpha`    | 0.2–3        | <1 favors small aligned clusters; >1 needs large majorities |
| `reward-step`           | 0.01–0.1     | ↑ faster reinforcement of persuasive agents                 |
| `reward-decay`          | 0–0.05       | ↑ bonus fades faster                                        |
| `meme-anti-leak`        | 0–0.5        | ↑ growth erodes the opposite stock more                     |
| `event-prob-max`        | 0–1          | ↑ more “massive” shocks per occurrence                      |
| `event-pace`            | ≥1           | ↓ means more frequent events (if repetition ON)             |
| `mod-prev` & `Rate-mod` | —            | adapt prevalence to opinion changes                         |

---

# CREDITS

* Original concept: *Public Opinion Research Group* (GROP.CA)
* NetLogo implementation & enhancements: **Pierre-Alain Cotnoir** (2015–2026)
* AI-assisted design: GPT-4 & GPT-5 (2024–2026)
* Contact: [pacotnoir@gmail.com](mailto:pacotnoir@gmail.com)

There is only one version of this model, created 5 days ago by Pierre-Alain Cotnoir.

Attached files

No files

Parent: Child of Version_20260115-2_Simulator v6_EN

Children:

• Child of Child of Child of Version_20260115-2_Simulator v6_EN