Hypothesis testing
H8 – Digital Displacement, Built–Natural Gradients, SDI Ordering, and Education Moderation Across Maslow Need Levels
H8a (Digital Belonging↑ / Self-Actualisation↓ with TDM)
H1: Higher Technology & Digital Media (TDM) weight is positively associated with Love & Belonging (LB) and negatively associated with Self-Actualisation (SA), controlling for Built and Humans.
H0: TDM has no association with LB or SA after controls.
H8b (Built gradient across need levels (downward slope))
H1: The association of Built Structures & Indoor/Urban Spaces with need salience declines as you move up the Maslow ladder.
H0: No monotone change across levels.
H8c (Natural gradient across need levels (upward slope))
H1: The association of Natural Landscapes & Phenomena increases with need level.
H0: No monotone change across levels.
H8d (Education moderation)
H1: Educational contexts amplify the TDM shift (stronger TDM×Edu negative for SA) and flatten Natural’s upward trend.
H0:No moderation by education.
Question:
Do Technology & Digital Media, Built Structures, and Natural Landscapes differ in how they relate to Maslow need levels, and are these patterns moderated by educational contexts?
Data
Data subset:
All 10 Developmental Affordance (DA) groups and all 6 Maslow need levels from the full corpus. No temporal filters. Inclusion rule: images were retained if they had at least one DA or at least one Maslow category present after merge and cleaning (deduplicated at image × category).
Variables:
-
- Predictors (DA groups): Technology & Digital Media (TDM), Built Structures & Indoor/Urban Spaces (Built), Natural Landscapes & Phenomena (Natural), Humans.
- Outcomes (Maslow): Love & Belonging (LB), Self-Actualisation (SA); levelwise analyses use all six Maslow levels (L1–L6).
- Controls: Category fixed effects (FEs); Humans included alongside TDM/Built/Natural in relevant models.
Normalisation & Weighting: See WWeighted Calculation Methods reference page.
N: 6,7890
(Acronyms: LB = Love & Belonging; SA = Self-Actualisation; TDM = Technology & Digital Media; FDR = false discovery rate–adjusted p; “β” = regression coefficient.)
Methods
- H8a (Digital displacement): Multiple regression using Weighted Least Squares (WLS) on proportional outcomes, Binary GLM (logit) on presence/absence, and Quantile Regression (QR) at q = 0.25/0.50/0.75 with bootstrap CIs. All models include category FEs; WLS uses cluster-robust SEs by category.
- H8b/H8c (Gradients across levels): Levelwise WLS: for each Maslow level, regress the level outcome on Built (H8b) or Natural (H8c) controlling for Humans + FEs. (Trend lines across levels are estimated in the scripts; results below focus on the levelwise coefficients you surfaced.)
- H8d (Education moderation): WLS with interactions (TDM×Edu, Natural×Edu), plus simple slopes for TDM at Edu=0 vs Edu=1.
- Multiple testing: Benjamini–Hochberg FDR within model families; only FDR-significant effects are reported in the main tables.
Results — H8a Digital Belonging↑ / Self-Actualisation↓ with TDM
Outcome: Love & Belonging (LB)
| Model | Predictor | β [95% CI] | FDR-p | Notes |
|---|---|---|---|---|
| WLS (cluster) | TDM | 0.210 [0.160, 0.260] | 3.06e-16 | Positive association |
| WLS (cluster) | Built | 0.052 [0.021, 0.083] | 0.00102 | Positive |
| WLS (cluster) | Humans | 0.189 [0.161, 0.217] | 4.02e-39 | Positive |
| Binary (logit) | Built | 0.827 [0.578, 1.075] | 1.44e-10 | Presence ↑ odds of LB |
| Binary (logit) | Humans | 2.818 [2.527, 3.109] | 2.91e-79 | Presence ↑↑ odds of LB |
| QR (q=0.75) | TDM | 0.385 [0.330, 0.442] | — | Strongest at upper LB |
| QR (q=0.50) | Built | 0.100 [0.056, 0.117] | — | Median LB |
| QR (q=0.75) | Built | 0.118 [0.074, 0.158] | — | Upper LB |
| QR (q=0.25/0.50/0.75) | Humans | 0.167 / 0.267 / 0.327 ([0.154, 0.179] / [0.250, 0.290] / [0.291, 0.346]) | — | Positive at all quantiles |
TDM, Built, and Humans are each positively associated with LB; TDM’s effect is largest at the upper part of the LB distribution (q=0.75).
Outcome: Self-Actualisation (SA)
| Model | Predictor | β [95% CI] | FDR-p | Notes |
|---|---|---|---|---|
| WLS (cluster) | TDM | −0.182 [−0.246, −0.118] | 3.22e-08 | Negative association |
| WLS (cluster) | Built | −0.089 [−0.147, −0.030] | 0.00322 | Negative |
| WLS (cluster) | Humans | −0.295 [−0.350, −0.239] | 5.94e-25 | Negative |
| Binary (logit) | TDM | −0.709 [−0.939, −0.479] | 2.40e-09 | Presence ↓ odds of SA |
| Binary (logit) | Built | +0.154 [0.004, 0.303] | 0.0498 | Small positive (presence) |
| Binary (logit) | Humans | −0.269 [−0.523, −0.015] | 0.0498 | Negative |
| QR (q=0.50) | TDM / Built / Humans | −0.416 / −0.117 / −0.331 (CIs all negative) | — | Median SA |
| QR (q=0.75) | TDM / Built / Humans | −0.218 / −0.187 / −0.464 (CIs all negative) | — | Upper SA |
TDM is negatively associated with SA across WLS, Binary, and QR. Built is largely negative for SA in WLS/QR, with a small positive effect in the presence model a common divergence between proportional vs. binary outcomes.
Results – H8b — Built gradient across need levels (downward)
Levelwise WLS β for Built, controlling Humans + FEs. Only FDR-significant levels are shown.
| Maslow Level | β [95% CI] | FDR-p | Direction |
|---|---|---|---|
| L1 Physiological | +0.121 [0.072, 0.170] | 4.28e-06 | Over-represented |
| L2 Safety | −0.046 [−0.066, −0.025] | 2.18e-05 | Under-represented |
| L6 Self-Transcendence | −0.068 [−0.087, −0.049] | 5.44e-12 | Under-represented |
Built is high at lower needs (L1) and suppressed at the top (L6), consistent with a downward gradient across levels.
H8c — Natural gradient across need levels (upward)
Levelwise WLS β for Natural, controlling Humans + FEs. All levels were FDR-significant.
| Maslow Level | β [95% CI] | FDR-p | Direction |
|---|---|---|---|
| L1 Physiological | −0.163 [−0.213, −0.113] | 3.16e-10 | Under-represented |
| L2 Safety | −0.090 [−0.122, −0.058] | 5.20e-08 | Under-represented |
| L3 Love & Belonging | −0.157 [−0.195, −0.119] | 1.41e-15 | Under-represented |
| L4 Esteem | −0.101 [−0.140, −0.061] | 6.39e-07 | Under-represented |
| L5 Self-Actualisation | +0.335 [0.263, 0.408] | 7.22e-19 | Over-represented |
| L6 Self-Transcendence | +0.175 [0.127, 0.223] | 1.89e-12 | Over-represented |
Natural is suppressed at lower levels (L1–L4) and elevated at higher levels (L5–L6), i.e., a strong upward gradient.
H8d — Education moderation (amplified TDM shift; altered Natural pattern)
WLS with interactions; cluster-robust by category.
| Outcome | Term | β [95% CI] | FDR-p | Interpretation |
|---|---|---|---|---|
| LB | TDM×Edu | +0.225 [0.170, 0.280] | 1.34e-15 | TDM’s pro-LB effect is stronger in education images |
| SA | TDM×Edu | −0.150 [−0.217, −0.082] | 1.73e-05 | TDM’s anti-SA effect is stronger in education images |
| LB | Natural×Edu | −0.112 [−0.132, −0.092] | 6.06e-28 | Natural’s LB association is dampened in education |
| SA | Natural×Edu | +0.557 [0.512, 0.602] | 1.92e-130 | Natural’s SA association is enhanced in education |
Simple slopes for TDM (derived from main + interaction):
- LB: TDM at Edu=0 → +0.204; at Edu=1 → +0.204 + 0.225 = +0.429.
- SA: TDM at Edu=0 → −0.175; at Edu=1 → −0.175 − 0.150 = −0.325.
Education amplifies the digital displacement pattern (LB↑, SA↓ with TDM). Natural’s role in education flips by outcome LB is dampened while SA is strengthened.
Combined Interpretation
Across 6,789 images (category-balanced), we see a coherent “digital displacement” pattern and complementary built–natural gradients:
Digital displacement (H8a):
-
-
- Technology & Digital Media (TDM) is pro-belonging and anti-self-actualisation even after controls.
- In WLS, TDM → LB↑ (β≈0.21, 95% CI [0.16, 0.26]) and SA↓ (β≈−0.18, [−0.25, −0.12]); Binary and QR agree directionally (e.g., QR q=.75 for LB: β≈0.39; QR q=.50 for SA: β≈−0.42).
- Built and Humans show a similar split—LB positive, SA negative—though magnitudes differ by model.
-
This indicates that screens/indoor cues and people cues are associated with belonging needs rather than self-actualisation.
Built–natural gradients (H8b/H8c):
-
-
- Built is over-represented at low needs and under-represented at high needs (e.g., L1 Physiological β≈+0.12; L6 Self-Transcendence β≈−0.07), consistent with a downward slope across the Maslow ladder.
- Natural flips that pattern: under-represented at lower levels (L1–L4 all negative) but over-represented at higher levels (L5 SA β≈+0.34; L6 ST β≈+0.18), i.e., a clear upward slope.
-
These complementary gradients fit a narrative where indoor/urban cues cluster with more immediate, social or security needs, while nature cues align with growth/realisation needs.
Education moderation (H8d):
-
-
- Educational contexts amplify the TDM shift and alter the role of Natural:
- TDM×Edu: stronger LB↑ (interaction β≈+0.23) and stronger SA↓ (β≈−0.15). Simple slopes: TDM→LB ≈ +0.20 (non-Edu) vs +0.43 (Edu); TDM→SA ≈ −0.18 (non-Edu) vs −0.33 (Edu).
- Natural×Edu: LB↓ (β≈−0.11) and SA↑ (β≈+0.56) in education imagery.
-
“Digital” signals in education strengthen the belonging-over-self-actualisation tilt, while “natural” signals in education suppress belonging and accentuate self-actualisation.
Together, these results tell a consistent story: screen/indoor contexts and people emphasise connectedness and social inclusion, especially in education, whereas nature tracks higher-order growth needs—most clearly at Self-Actualisation and Self-Transcendence.
Concluison
H8a — Reject H₀. TDM is pro-Belonging and anti-Self-Actualisation: LB β≈0.21 (FDR p≈3e-16); SA β≈−0.18 (FDR p≈3e-08).
H8b — Reject H₀. Built shows a downward gradient: positive at L1 (β≈+0.12, FDR p≈4e-06) and negative by L6 (β≈−0.07, FDR p≈5e-12).
H8c — Reject H₀. Natural shows an upward gradient: negative at lower levels (e.g., L3 β≈−0.16, FDR p≈1e-15) but positive at higher levels (L5 β≈+0.34, FDR p≈7e-19).
H8d — Reject H₀. Education strengthens the TDM pattern and shifts Nature by
-
-
- TDM×Edu → LB β≈+0.23 (FDR p≈1e-15), TDM×Edu → SA β≈−0.15 (FDR p≈2e-05).
- Simple slopes for TDM: LB +0.20 → +0.43, SA −0.18 → −0.33 (non-Edu → Edu).
-