The Invisible Burden — EdTech Dashboard

Methodology

Data collection, scoring framework, and inter-rater approach across 12 spectra

Study Overview

The Invisible Burden: Teacher Beliefs and EdTech Adoption in Singapore Primary Schools — a qualitative study using conversational AI interviews to explore how teachers experience EdTech as a work-burden or work-saver.

Findings serve ICT Key Personnels (KPs) in developing strategies for meaningful EdTech-infused lesson delivery.

Data Collection

Platform: Edcafe chatbot (SAM3) — async text & voice
Participants: 61 teachers, Singapore primary schools
Period: 27 March – 3 April 2026
Duration: 20–25 min per session
Sampling: Pseudonymous; two pairs disambiguated (Joyce ×2, Sam ×2)

Scoring & Reliability

12 spectra scored 1–6 by analyst, post-interview
Tone of voice coded independently (P/U/N per dimension)
Two-pass review: initial scoring → delta check against full corpus
53 changes applied: 25 human-confirmed, 14 delta-triggered, S11/S12 recalibration
Cohen (1988) thresholds used for correlation strength

Cohen (1988) R² Thresholds — N=61

Very Strong — R² ≥ 0.49
Strong — R² 0.25–0.49
Moderate — R² 0.09–0.25
Weak — R² 0.01–0.09
Negligible — R² <0.01

Pearson r visualised; Spearman ρ more defensible for ordinal scales in formal reporting.

Scoring Rubric — All 12 Spectra

Descriptors for each score level. Scroll horizontally to see all rank columns. Hover score chips in Responses for quick rubric reference.

Scatter Plot

All 66 Pairwise Correlations

Sorted by R² descending. Cohen (1988) thresholds. Hover dots for names.

Tone coded independently of content — captures emotional register, not what is being reported. S1 excluded (factual; no affective dimension). Positive tone (n=25) tracks strongly with higher S2, S4, S11 scores. Negative tone (n=9) mean S4=2.33 vs Positive mean S4=4.16.

Dimension	P %	U %	N %	Note
S2 Work Burden	28%	56%	16%	Neutral dominant; burden rarely expresses as positive emotion
S3 Facilitating Cond.	30%	39%	31%	Highest N% of any dimension — facilities friction is felt acutely
S4 Volition	49%	41%	10%	Highest P% of non-overall dimensions — conviction is energising
S5 Resilience	28%	56%	16%	Mirrors S2; resilience talk is rarely positive in affect
S6 Psych Safety	33%	49%	18%	Moderate N; isolation is felt but not always named
S7 TechConf	26%	57%	16%	Mostly neutral; troubleshooting discussed matter-of-factly
S8 Collegiality	33%	54%	13%	Mostly neutral; sharing culture described, not celebrated
S9 Pedagogy	48%	44%	8%	Second-highest P% — pedagogy talk is where joy lives
S10 Blended	33%	51%	16%	Moderate; blending discussed pragmatically
S11 Std Impact	46%	43%	11%	Strong P%; teachers feel good about outcomes when they occur
S12 Std Ready	18%	64%	18%	Highest U% — readiness discussed clinically, not emotionally

Structural Patterns

Re-analysis on full 12-spectra dataset. Click member chips to jump to their response.

Key finding: S10 (Blended Efficacy) is the single strongest predictor of S11 (Perceived Student Impact): r=0.733, R²=0.537. S9×S10 (r=0.75) shows pedagogy and practice are inseparable. S5×S6 (Resilience × Psychological Safety) at r=0.885 — school culture is the hidden variable behind both. S4 Volition/Agency is the bridge variable linking dispositional and practice clusters.

Correlation Network Summary

Two structural clusters emerge from the 45 pairwise correlations

Cluster A — Dispositional

Internal attitudes and interpersonal dynamics

Cluster B — Practice

Observable teaching behaviours and outcomes

Bridge Variable: S4 Volition/Agency

S4 correlates strongly with both clusters: with S5 Resilience (r=0.769), S7 Troubleshooting (r=0.769), S8 Collegiality (r=0.751) and with S9 Pedagogy (r=0.696), S10 Blended Efficacy (r=0.674). This means personal pedagogical conviction is the fulcrum. Dispositional support (safety, collegiality) enables agency; agency enables blended practice; blended practice produces impact.