Bridging theory and practice in just energy transitions: A multi-method analysis integrating AI-driven computational frameworks and qualitative field research in post-lignite Greek regions

The European Union's just transition framework allocates over €150 billion to ensure that decarbonization does not disproportionately burden fossil-fuel-dependent communities, with Greece receiving approximately €1.6 billion for its post-lignite regions of Western Macedonia and Peloponnese. The stakes are considerable: these regions face cascading socio-economic disruption, mass displacement of lignite workers, youth out-migration rates exceeding 10% per decade, and agricultural decline driven by aging workforces, structural constraints and the impacts of extreme weather events. Compounding these challenges is a collective crisis of belief among populations subjected to successive economic shocks since 2009. Existing research has examined just transition financing, governance design, and stakeholder perceptions in isolation, and has established that place-based approaches, inclusive participation, and capacity building are essential to equitable outcomes. However, no study has systematically diagnosed why transitions that appear to possess all conventionally recognized success factors, including strong political commitment, substantial funding, and technically available solutions, nonetheless struggle to deliver results at the community level. The mechanisms linking policy ambition to implementation shortfall remain poorly understood. This study addresses that gap through a convergent multi-method design integrating rule-based and machine-learning computational text analysis with independent qualitative field research. The evidence base comprises two independent studies conducted in 2025: a computational study of 20 semi-structured stakeholder interviews and two focus groups in Western Macedonia and Megalopolis (22 sessions, approximately 55,000 words, collected February to May 2025), and a qualitative field study of 25 interviews, two focus groups, and site observations across Western Macedonia, the Peloponnese, Laconia, and Athens (June to August 2025). The computational component applies 13 analytical frameworks simultaneously, using NLP and ML techniques including TF-IDF vectorization, Latent Dirichlet Allocation topic modeling, VADER sentiment analysis, K-Means clustering, spaCy named entity recognition, and network centrality algorithms, producing quantified diagnostics across multiple dimensions that single-framework approaches cannot capture. The qualitative fieldwork comprises 25 interviews, two farmers groups (farmers; national-level decision makers focus groups, and makers), with site observations across four Greek regions including the non-lignite agricultural district of Laconia as a within-region contrast case, providing the environmental, cultural, and experiential depth that computational methods alone cannot reach. All reported scores are researcher-defined, discourse-based indicators; their construction, decision rules, and uncertainty are documented in the supplementary materials, and they should be read as exploratory diagnostics rather than validated measurements. The converging results indicate a paradox of plenty: expressed leadership commitment and resource-related discourse register high (8.0/10 and 10.0/10 on framework-specific indices), yet corpus-level awareness (1.7/10), demonstrated capability (1.3/10), and expressed will (1.0–1.9/10 across categories) register critically low; because these indices derive from different frameworks and analytical units, they are reported side by side rather than combined into a single validated scale. This gap coexists with strikingly low discourse presence of affected populations (workers 4.8% of coded discourse, women 0.5%), a concentration of resource-related discourse on technology (53% of resource-related segments; 48.6% normalized) despite critically low readiness, a generational sentiment divide (75.6% positive among youth-focused passages versus 50.0% in aging-population passages) that plausibly accelerates the demographic erosion necessary for adoption, and unaddressed water scarcity that undermines the agricultural modernization upon which diversification strategies depend. These findings are formalized as Implementation Gap Theory, which identifies four interacting gap types (vertical, horizontal, temporal, and cognitive) proposed as testable mechanisms of implementation shortfall despite favorable preconditions. The framework is offered as a candidate diagnostic for coal-dependent transition regions across the European Union, pending independent validation, and demonstrates that funding adequacy and policy commitment are necessary but insufficient conditions for just outcomes.