AI-based Scenario Generator for the 3LandSpiel - Concept and Recommendation for a Serious Game Deployment in a Cross-Border Educational Context

The 3LandSpiel teaches adolescents in the Upper Rhine about their trinational region. This thesis recommends how an AI-based generator can produce new game scenarios in German and French, hosted in Switzerland and affordable beyond the funding period.

Müller, William, 2026

Type of Thesis Bachelor Thesis
Client Fachhochschule Nordwestschweiz FHNW
Supervisor Grieder, Hermann
Views: 1
The 3LandSpiel is a trinational educational board game funded under Interreg Oberrhein. Part of the project is an AI-based scenario generator that should produce new game content in German and French, so that teachers can adapt the game to their class without writing everything by hand. Its users are pupils, and it must remain operable after the funding ends. This constrains the choice of language model, hosting provider and architecture at once: bilingual quality, Swiss and European data protection, verifiable data residency, and long-term cost. No such choice had been made.
Requirements were derived from the project documentation, two prior FHNW student projects and consultation with the client, then prioritised using the MoSCoW scheme. A literature review established the conceptual and regulatory basis. Five language-model families and six Swiss hosting providers were compared against six weighted criteria in a decision matrix. A sensitivity analysis then tested how far the resulting ranking depends on the weighting applied.
Nineteen requirements were derived, twelve of them mandatory, which shows how tightly regulation and funding constrain the solution. Of four viable combinations, Mistral Small 3.1 on Infomaniak AI Tools scored highest at 4.70 of 5 and stayed top under all four alternative weightings. Apertus 8B, the Swiss model from ETH Zurich and EPFL, on the PHOENIQS platform reached 4.50 and draws level whenever long-term independence weighs more heavily or hosting is sponsored. Both are recommended in a hybrid architecture: inference is bought as a managed Swiss service, while the knowledge base and content moderation stay under project control. The client receives a decision-ready package: a validated requirement set, a documented comparison of the Swiss market, a target architecture, and a four-phase roadmap with responsibilities. The sensitivity analysis proved decisive, since it turned a single apparent winner into two defensible options and made explicit which priority decides between them. The choice therefore remains reversible after the pilot.
Studyprogram: Business Information Technology (Bachelor)
Keywords 3LandSpiel, AI, Scenario Generator, Serious Game
Confidentiality: vertraulich
Type of Thesis
Bachelor Thesis
Client
Fachhochschule Nordwestschweiz FHNW, Basel
Authors
Müller, William
Supervisor
Grieder, Hermann
Publication Year
2026
Thesis Language
English
Confidentiality
Confidential
Studyprogram
Business Information Technology (Bachelor)
Location
Basel
Keywords
3LandSpiel, AI, Scenario Generator, Serious Game