Spectral Entropy

Description

Shannon entropy of a normalised power spectrum, quantifying whether behaviour is dominated by a few rhythms (routine) or spread across many (exploration or overload).

Spectral entropy normalises the power spectrum into a probability distribution and computes its Shannon entropy. Low values mean energy concentrated in a few frequencies; high values mean energy spread broadly. The Behavioral Spectral Framework uses it to quantify exploration versus routine: stable low-frequency dominance indicates habit formation, while rising entropy indicates curiosity or cognitive overload, which only context can separate. In the fintech onboarding case, spectral entropy decreased from 0.89 to 0.62 across three sessions, signalling an emergent routine, and served with wavelet burst counts and mean frequency as predictors in a Random Forest that classified sessions as stable or volatile with 82% accuracy. Entropy drift plots are a recommended visual.

Attributes

Loss type
Unmigrated
Stage
Post-launch
Effort
Days

Steps

  1. Compute the power spectrum for each session or user.
  2. Normalise spectral power to sum to 1.
  3. Compute Shannon entropy of the normalised spectrum, scaled to 0–1 if comparing different resolutions.
  4. Track entropy across sessions and plot entropy drift.
  5. Use entropy as a feature in interpretable models alongside burst counts and mean frequency.
  6. Disambiguate high entropy (curiosity versus confusion) with qualitative evidence.

Sources

  • Vivek Gopinath (2025), From Stars to Signals: Translating Matteo Paz's Fourier-Wavelet Model into Behavioral and UX Research Methodology

Also known as

  • Power spectral entropy
  • Rhythm entropy

Categories

  • Information Theory, Physics & Mathematics