Compatibility Model for Multi-Party Collaborative AI Systems Development
2026 (English)Licentiate thesis, comprehensive summary (Other academic)
Abstract [en]
Multi-party collaborative development of AI systems promises efficiency gains and cost reduction. AI marketplaces support these collaborations as platforms where AI assets like data and models can be traded or otherwise exchanged. The parties collaborating through such platforms can be individuals or organisations. They operate in distinct contexts with different technical, economic, and regulatory requirements. Thus, collaboration requires more than technical compatibility. The literature lacks a holistic view of compatibility requirements and their relationships.
Therefore, the thesis aims to design and operationalise a compatibility model for AI systems development on an AI marketplace. The methodology for developing the model was Design Science Research. The empirical context was the Bonseyes Marketplace (BMP).
The designed model contains four compatibility levels: legal, economic, governance, and technical. Beyond identifying these levels, the conceptual contribution lies in demonstrating the interdependence of their requirements and in explicitly distinguishing compatibility from interoperability. While compatibility enables exchange, interoperability enables the use of AI assets. Compatibility is thus a precondition for interoperability.
The compatibility model was operationalised through discussions of pricing alternatives at the economic level and licensing rules and terms at the governance level. The technical level was partially operationalised through a system architecture for realising compatibility between an AI marketplace and an AI pipeline comprising AI assets.
The model’s usefulness was assessed by analysing BMP validation data in the context of the compatibility levels. Most observations could be associated with the levels, demonstrating the model’s usefulness for analysing marketplace-supported collaborations.
The results can support collaborators and marketplace developers in identifying and aligning compatibility requirements. Researchers can focus on limitations in the completeness of both the model and its operationalisation, and on validation in additional marketplaces.
Place, publisher, year, edition, pages
Karlskrona: Blekinge Tekniska Högskola, 2026. , p. 169
Series
Blekinge Institute of Technology Licentiate Dissertation Series, ISSN 1650-2140 ; 2026:09
Keywords [en]
Compatibility, AI, Marketplace, Pricing, Licencing, Licences, Requirements
National Category
Engineering and Technology Computer Sciences
Research subject
Computer Science
Identifiers
URN: urn:nbn:se:bth-30575ISBN: 978-91-7295-536-3 (print)OAI: oai:DiVA.org:bth-30575DiVA, id: diva2:2109376
Supervisors
Projects
European Union's Horizon 2020 grant agreement No 732204 (Bonseyes)
Funder
EU, Horizon 2020, 7322042026-10-152026-10-092026-11-06Bibliographically approved
List of papers