Description
A five-stage mixed-methods workflow treating behavioural telemetry as light-curves: decompose into frequency features, annotate them qualitatively, model and visualise rhythm, then validate meaning.
Inspired by Matteo Paz's 2024 Fourier and wavelet pipeline for detecting variable objects in the NEOWISE survey, the framework treats engagement, sentiment and physiological traces as 'light-curves of cognition' with periodic behaviour (habit cycles), transient events (frustration bursts) and drift (learning, fatigue). The five stages are Signal Framing, Spectral Decomposition, Contextual Annotation, Modeling and Visualization, and Validation and Iteration. It explicitly names where translation fails (stationarity, irregular sampling, semantic noise, label scarcity, non-equivalence of energy and meaning, intentionality, ethics) and redesigns around them: recurrence over strict periodicity, Lomb-Scargle for uneven events, noise retained as emotion, wavelets as a behavioural microscope, metadata context anchors and a human-in-the-loop quant → qual → quant loop. The integration equation states: meaningful behavioural insight = spectral precision × contextual anchoring × interpretive dialogue. It is presented as a conceptual playbook; the case is hypothetical.
Attributes
- Discipline
- Community-Proposed Frameworks
- Loss type
- Community-proposed
- Tested
- Not yet tested
- Stage
- Post-launch
- Data
- Mixed
- Effort
- Weeks
Steps
- Signal Framing: choose the behavioural signal (e.g. interaction density per minute), align time resolution to the hypothesised rhythm, capture context variables and secure informed consent.
- Pre-process: normalise timestamps, interpolate small gaps (< 30 seconds) and retain longer gaps as disengagement indicators.
- Spectral Decomposition: apply Fourier for global periodicity, CWT or STFT for temporal localisation and Lomb-Scargle for irregular events; extract dominant frequencies, power distribution, spectral entropy and phase coherence.
- Contextual Annotation: align spectral peaks and bursts with transcripts, logs or video, open-code each event (frustration, curiosity, learning, boredom) and build a Spectral Annotation Matrix.
- Modeling and Visualization: train interpretable models (Random Forest, XGBoost, LSTM) on spectral features and produce spectrograms, entropy drift plots and coherence maps.
- Validation and Iteration: correlate spectral metrics with SUS, NPS or TLX, verify interpretations in post-analysis participant interviews, and re-analyse with updated codes.
- Run an ethical check: use insights to improve design, not to infer private cognition without consent or to manipulate behaviour.
Sources
- Vivek Gopinath (2025), From Stars to Signals: Translating Matteo Paz's Fourier-Wavelet Model into Behavioral and UX Research Methodology
- Matteo Paz (2024), A Sub-Millisecond Fourier and Wavelet Based Model to Extract Variable Candidates from the NEOWISE Single-Exposure Database
- wavelet.pdf
- uxguy.in
Also known as
- Behavioral Fourier-Wavelet Framework
- Fourier-Wavelet model for UX research
- From Stars to Signals
- BSF
- Behavioural spectral analysis
- Behavioral Spectral Framework (BSF)
Categories
- Community-Proposed Frameworks