SRM(s,t) — Symbolic Resonance Score for symbol s at time t.
Integrates four measurable dimensions of political symbol diffusion, penalizes semantic instability
via exponential decay, and normalizes against total symbolic field activity.
V = Viral Velocity ·
A = Affective Weight ·
D = Semantic Drift ·
N = Network Reach
Epistemological Status
This model is a theoretical proposal grounded in established literature: information diffusion
(Bass, 1969; Watts & Dodds, 2007), affective computing (Russell, 1980), computational semantics
(Mikolov et al., 2013). Mathematically coherent and operationalizable.
Not yet validated on a full empirical dataset.
The calculator below is a conceptual demonstration, not a calibrated predictive instrument.
Variables & Operationalization
Symbol
Name
Definition
Measurement
Data Source
V(s,t)
Viral Velocity
Rate of change in symbol adoption — dN/dt per platform per hour. Normalized to [0,1].
Time-derivative of cumulative mention count. Rolling 6h window.
Twitter/X API v2 · Reddit Pushshift · Telegram
A(s,t)
Affective Weight
Composite of emotional valence (−1 to +1) and arousal (0 to 1). A = (|valence|+arousal)/2.
RoBERTa-sentiment · XLM-RoBERTa · Russell Circumplex mapping.
Platform text corpus · multilingual NLP models
D(s,t)
Semantic Drift
Cosine distance between word embedding at baseline t₀ and current t. D > 0.3 = co-optation alert.
Normalized betweenness centrality across ideological communities. Cross-community bridges ×1.5.
Retweet graph → Louvain community detection → betweenness centrality.
NetworkX · Gephi · platform diffusion graphs
Z(t)
Normalization Factor
Sum of raw scores across all active symbols at time t. Dynamic — not a constant.
Z(t) = Σ V·A·e^(−λD)·N across all tracked symbols simultaneously.
Computed dynamically from full symbol corpus.
Parameter Justification
λ — Decay Rate: Why 2.0 as Initial Estimate?
λ = 2.0 is a working estimate, not empirically calibrated. Derived from:
Bass Diffusion Model (1969) — imitation coefficients suggest constants in 1.5–2.5 range;
Hamilton et al. (2016) — political symbols show cosine drift 0.3–0.6 within 30 days;
Epidemiological analogy — SIR model recovery rates suggest half-life constants of 1.8–2.2.
Calibration required — λ must be estimated via maximum likelihood on a labeled historical dataset
(minimum 15 symbol-event pairs) before making predictive claims.
Z(t) — Why Not a Constant?
Z(t) is a partition function borrowed from statistical mechanics — the sum of all unnormalized scores across the active symbol corpus at time t.
Without dynamic Z(t), scores during high-noise events are artificially inflated. In the single-symbol simulator, Z(t) reduces to the raw score.
Measurement Pipeline
01
Symbol Corpus Collection
Identify candidate symbols via automated trend detection and expert seeding. Collect timestamped instances. Each symbol assigned a baseline embedding at t₀.
Twitter API v2Reddit PushshiftTelegram MTProtopHash detection
02
Affective Encoding — A(s,t)
VADER for English baseline, fine-tuned RoBERTa for valence, XLM-RoBERTa for non-English. Map to Russell Circumplex. Aggregate per 6h window.
Conceptual demonstration · Not a calibrated predictive instrument
SRM(s,t) = Σ [ V · A · e−λD · N ] / Z(t)
0.60
Moderate spread
0.75
High emotional charge
0.20
Meaning mostly stable
0.80
Wide cross-community reach
2.00
Low λ: drift penalizes weakly (tolerates co-optation). High λ: requires semantic purity for high resonance.
—
SRM Score
—
—
Velocity V
—
Affective A
—
Drift e^−λD
—
Network N
—
⚠ Single-symbol demo: Z(t) = raw score. In real deployment, Z(t) computed across all tracked symbols. λ = 2.00 (working estimate — requires calibration).
Illustrative Historical Cases
⚠ Methodological transparency: Values below are estimated approximations from published sources — not the result of running the full SRM pipeline on raw data.
Umbrella Symbol — HK Protests
2019–2020 · Hong Kong
Estimated
V — Viral Velocity~0.88
A — Affective Weight~0.85
D — Semantic Drift~0.10 (very stable)
N — Network Reach~0.90
Est. SRM (λ=2.0)~0.84
Lee & Chan (2020); HKPD dataset; Twitter/Weibo archives (Molina et al., 2021).
MAGA Cap — US Elections
2016–2024 · United States
Estimated
V — Viral Velocity~0.82
A — Affective Weight~0.91
D — Semantic Drift~0.48 (significant)
N — Network Reach~0.86
Est. SRM (λ=2.0)~0.62
Mutz (2018); Sides et al. (2019); ANES affective polarization data.
Sunflower Movement
2014 · Taiwan
Estimated
V — Viral Velocity~0.75
A — Affective Weight~0.80
D — Semantic Drift~0.07 (extremely stable)
N — Network Reach~0.68
Est. SRM (λ=2.0)~0.78
Ho (2015); Tang & Chen (2014); PTT archive data.
Known Limitations — v1.1
V–N Collinearity
Viral velocity V and network reach N are not fully independent — rapid spread tends to correlate with high network reach.
→ Fix: Factor analysis or PCA on real dataset.
λ Not Yet Calibrated
λ = 2.0 is an initial estimate. Without labeled historical data, the optimal λ is unknown.
→ Fix: Maximum likelihood estimation on 15+ labeled pairs.
Platform Heterogeneity
Aggregating V and N across Twitter, Reddit, Telegram ignores platform-specific amplification mechanics.