Ecosystem research has developed rich taxonomies and typologies of multi-organizational arrangements, but it offers limited guidance on how such ecosystems can be intentionally constructed and aligned. This dissertation addresses that gap by introducing a prescriptive, role-based design logic — Ecosystem by Design — which operationalizes how organizations can configure responsibilities and manage interdependencies in settings where contributors pursue joint objectives. It shifts the discourse from ecosystem governance and orchestration to ecosystem design, focusing on artificial intelligence (AI) as a practically demanding and relevant context. As modular, rapidly evolving, and distributed solutions across multiple organizations, AI-driven systems introduce a variety of interdependencies and are subject to co-evolutionary dynamics. The Ecosystem by Design approach integrates four components, each developed in a dedicated study. Study 1 introduces the ECCO framework (Ecosystem Conceptualization and Characterization with Ontological Dimensions) for systematically defining ecosystem boundaries, central elements, locality, actors, and links. Study 2 develops the CAPS framework (Characterizing AI Phenomena and Solutions), a profiling tool that captures the nature, use, and strategic dimensions of AI solutions. Study 3 provides functional role templates and an Ecosystem Activity Canvas to map activities and allocate responsibilities without prematurely fixing organizational assignments. Study 4 identifies and categorizes activity- and component-level interdependencies, operationalized as Interdependency Cards and an Interdependency Matrix for scoring and mitigation. Methodologically, the thesis follows a phenomenon-based, abductive–iterative logic, combining literature synthesis, workshops, and multiple case studies to derive transferable design artifacts. It makes use of the very technologies it investigates by analyzing large text corpora in Studies 1 and 2 utilizing three different topic modeling techniques. With the Ecosystem by Design approach, the dissertation contributes to three domains. For ecosystem research, it moves beyond descriptive typologies by offering a role-based design logic grounded in the construction of ecosystem phenomena. For AI research, it introduces CAPS to translate solution-level technical characteristics into ecosystem requirements. It demonstrates how AI-specific dependencies, such as retrieval/versioning, managed model services, or provenance, shape design choices. For phenomenon-based research, it offers repeatable templates that render researchers’ pre-understandings explicit and comparable across cases. Practically, the thesis presents a modular toolbox — comprising ECCO, CAPS, Role Cards, Activity Canvas, and Interdependency Cards/Matrix — designed for workshop application and governance integration, with artifacts made available in OSF repositories. While pre-filled for AI scenarios, the systematics extend to other technology-centric ecosystems. In summary, this dissertation reframes ecosystems as designable systems, in which achieving alignment requires explicit characterization of phenomena, role-based responsibility design, and the deliberate management of interdependencies. The Ecosystem by Design approach makes these practices actionable for both researchers and practitioners.
Julius Kirschbaum (Thu,) studied this question.