Asuntouutisoinnin sentimentin kriittinen kerrosanalyysi koneoppimisen ja agenttipohjaisen tekoälyn avulla: Moniagenttijärjestelmään perustuva lähestymistapa Suomen asuntomarkkinoihin
Year: 2026
Type: dissertation
Abstract: This study explores Finnish housing related news discourse and prices in Statistics Finland's index series. The literature review investigates macrofinancial fundamentals and the narratives dominating media. Discursive patterns and the futures-oriented interpretation of housing market discourse are considered from additional perspectives, e.g., megatrends, geopolitics, weak signals, dragon kings, and the psychology of bubbles. The focus then shifts to financialisation, the voices of the primary definers, media discourse, media as a filter, and Finnish media pluralism. The research found a gap: there was no indexed study in Google Scholar or Scopus at the time of writing combining corpus-scale sentiment analysis with causal layered analysis (CLA) in a housing or real estate market context. To fill this gap, a comprehensive raw corpus of news headlines was scraped with 46 Boolean keywords from the national public broadcaster Yle's 1 031 013 news articles between 2006 and 2025. The author used artificial intelligence (AI) tools and large language models (LLMs) to isolate housing news sentiment, gating 697 679 news headlines to 30 460 relevant articles and, at the last stage, to a labelled corpus of 8 180. This thesis proves with inter-rater reliability (IRR) checks that LLMs can label news sentiment adequately. The author vs Mistral LLM IRR check returned a very convincing quadratic-weighted Cohen's kappa of 0.665, author vs advisor of 0.697, and yearly bin Pearson r of 0.870. CLA: "litany; systemic; worldview; myth and metaphor" were layered in a multi-agent system (MAS) panel, creating a holistic picture: home ideology, the two-dwelling trap, "buy now" and "postponed recovery" genres, which made sentiment fluctuate. "Cut your losses" is missing from the media discourse, as are megatrends: climate change and biodiversity loss as drivers of prices. Weaponising oil and gas against the EU, and Russia and the US attacking democracy are dragon king events. Extreme yet retrospectively foreseeable through weak signals, they may have had the greatest influence on old dwelling prices in Finland between 2022 and 2025.
Source: Aaltodoc (Aalto University)
Author: Laitinen, Rauli
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Funder: European Commission
OpenAlex ID: W7203958743