EcoAI Framework

Research & Findings

Empirical validation of the EcoAI Index across 601 upper-secondary students in southern Italy — a mixed-methods sequential design.

Reading mode

General public Researcher
1

Research Design

This research set out to answer a concrete question: do young people in Italian secondary schools have the competences to navigate a world shaped by both ecological crisis and artificial intelligence? To find out, we built and tested a measurement tool — the EcoAI Index — with over 600 students.

1

Exploratory Phase — 403 students

Ariano Irpino, May–June 2025. We administered a first questionnaire to understand how students relate to sustainability and AI. We then grouped them into clusters based on their responses.

2

Confirmatory Phase — 198 students

A refined questionnaire was administered to a second sample. This phase statistically confirmed the five dimensions of the EcoAI Index and their relationships.

The study adopts a sequential mixed-methods design with two administrations serving distinct analytic functions: an exploratory phase for instrument development and cluster profiling, followed by a confirmatory phase for structural validation.

Survey Q1 — Exploratory N = 403

Students were grouped by how they approach sustainability and AI. Three distinct profiles emerged for each theme.

Method

K-means cluster analysis on Likert batteries. Two independent cluster solutions: sustainability orientation (3 groups) and AI stance (3 groups).

Sustainability clusters

Distant from sustainability30.5%
Sustainability-oriented36.0%
Eco-aware and engaged33.5%

AI clusters

Pragmatic enthusiasts47.7%
Detached27.1%
Critical and ambivalent25.1%
Survey Q2 — Confirmatory N = 198

The five-dimension model was statistically confirmed. The structure holds consistently across genders and school years.

CFA — DWLS/WLSMV estimator (polychoric matrix, JASP)

0.978

CFI

0.975

TLI

0.063

RMSEA

0.082

SRMR

✓ Configural invariance — gender & school year
✓ Metric invariance confirmed
✓ Structural chain confirmed: INFO → VAL/AATI → AGY → EPAS SEM chain: INFO → VAL/AATI → AGY → EPAS
2

The Five EcoAI Index Dimensions

The EcoAI Index is a measurement tool that captures five distinct aspects of eco-digital competence, from how well students evaluate online information to whether they feel part of a broader human-technology-nature network. Each dimension is scored on a 1–5 scale.
Five latent constructs measured via Likert items (1–5), estimated through CFA with DWLS estimator. Reliability reported as McDonald's ω; convergent validity via AVE. Weights are SEM path coefficients normalized to sum = 1.00.
INFO

Eco-Digital Information Literacy

How well students search for, evaluate, and critically use information on environmental and civic issues through digital sources.

Competences in retrieval, critical evaluation, and purposive use of eco-digital information across digital platforms. Prerequisite node in the SEM chain.

ω 0.655 AVE 0.394 SEM weight 0.22
AATI

AI Awareness & Critical Attitudes

Understanding how algorithms work, recognizing AI bias, and developing a critical but informed stance toward AI systems.

Algorithmic awareness, bias perception, and ethical orientation toward AI governance. Highest discriminant power in Eco-Index classification (η²=0.528). Highest inter-factor correlation with EPAS (r=0.822) and VAL (r=0.851).

ω 0.806 AVE 0.418 SEM weight 0.18
VAL

Eco-Civic Values

A personal commitment to sustainability, collective environmental responsibility, and civic justice — values that guide eco-digital behavior.

Value orientation toward sustainability, collective environmental responsibility, and civic justice. Ceiling effect documented in the Low Eco-Index class (M=3.094), attenuating effective discriminant power relative to η² value.

ω 0.844 AVE 0.525 SEM weight 0.20
AGY

Eco-Digital Civic Agency

The actual willingness and capacity to act as a responsible citizen in eco-digital contexts — feeling capable of making a difference.

Perceived self-efficacy and behavioral readiness for eco-digital civic action. Mediated dimension in the SEM: receives effects from both VAL and AATI, mediates toward EPAS. Normalized weight 0.24.

ω 0.763 AVE 0.370 SEM weight 0.24
EPAS

Eco-Posthuman Agency Scale

A sense of belonging to wider networks — recognizing oneself as part of an interconnected human-technology-nature system, not just an individual user.

Subjective disposition to recognize oneself as node in hybrid human-technological-ecological networks; relational integration openness. Terminal node in SEM; highest normalized weight (0.26). Integrative function confirmed by elevated correlations with AATI (r=0.822), VAL (r=0.819), AGY (r=0.814).

ω 0.535 AVE 0.423 SEM weight 0.26
3

Structural Model

The five dimensions are connected to form a chain. Information literacy builds the foundation, shaping how students think about AI and their values, which in turn drives real civic action, and ultimately a deeper sense of ecological identity.
Structural Equation Model (SEM) with DWLS estimator. All paths p<.001 except VAL→EPAS (p=.002). Indirect effects: INFO→AGY β=0.29; INFO→EPAS β=0.34; AATI→EPAS β=0.14. AATI↔VAL correlation r=0.851 (highest in factor matrix).
INFO Information
Literacy
β=0.55
AATI AI Attitudes
VAL Civic Values
β=0.42
β=0.30
AGY Civic Agency
β=0.45
EPAS Eco-Posthuman
Agency

All coefficients p<.001 · VAL→EPAS p=.002 · r(AATI↔VAL)=0.851

Reading: information skills (INFO) feed into critical AI awareness and civic values, which together build agency, which culminates in eco-posthuman identity.

4

Eco-Index Profiles

Three student profiles: Based on their total EcoAI Index score, students were divided into three groups. The chart below lets you explore how each group scores on each dimension — and spot where the gaps are.
Percentile-based classification (P33/P66), n≈66 per class, N=198. Composite Eco-Index: M=3.357, SD=0.559, range 2.058–4.974 (scale 1–5). Weights derived from normalized SEM path coefficients. Discriminant power reported as η².
M 3.357 SD 0.559 N 198
Low class (n≈66)
Medium class (n≈66)
High class (n≈66)
⚠ VAL ceiling effect: Even students in the Low class score above 3.0 on Eco-Civic Values — suggesting that values are widely shared, but not yet translated into action or critical awareness. VAL Low class M=3.094 exceeds scale midpoint. Ceiling effect attenuates effective discriminant contribution despite η²=0.416. Discussed in §6.3.2.
Discriminant power η²

0.528

AATI

0.474

EPAS

0.416

VAL

0.406

AGY

AATI is the primary discriminant despite lowest SEM weight (0.18) — theoretically relevant asymmetry.

5

Inter-factor Correlation Matrix

Latent factor correlations from CFA. Bold values >0.80. No correlation exceeds the redundancy threshold (0.90), confirming empirical distinctiveness of the five constructs.

Cite this research

Rubino, D. (2025). Eco-Digital Posthumanism: Eco-Digital Co-Responsible Agency, Genealogy and Research Agenda. SocArXiv. ResearchGate