Human-centered Hybrid Intelligence: Provision of human input and its utilization by AI in creative work and problem-solving

Abstract

Motivation Artificial Intelligence (AI) applications are rapidly moving from futuristic concepts to becoming indispensable instruments for organizations navigating volatile markets, geopolitical shocks, and accelerating technological change. AI tools are now on the agenda for the majority of organizations, yet relatively few have embedded them deeply enough to influence processes that drive innovation and competitive advantage, such as creative work and complex problem-solving. While 70% of organizations experiment with generative AI or predictive analytics, only a few have scaled these systems beyond isolated pilots. Most teams still struggle with cognitive overload, unstructured workflows, and collaboration frictions - symptoms that current point solutions cannot address. Closing this gap requires deliberately designed human-AI systems that weave algorithmic assistance into the way diverse groups think, decide, and act together. Hybrid Intelligence Systems (HIS) hold great promise for addressing these challenges by combining the complementary strengths of humans and AI. Humans bring empathy, common sense, and contextual awareness, while AI excels at processing large amounts of data, identifying patterns, and handling repetitive tasks. This collaboration can create opportunities for AI not only to support human efforts but also to act as a creative partner in both individual and team-based problem-solving. Realizing the potential of HIS requires careful consideration of how humans and AI can best work together. A key limitation in current AI development is the lack of a socio-technical perspective. Much of the focus has been on technological capabilities, often neglecting the human and organizational factors critical for effective integration. A human-centered approach to HIS is needed to address questions such as what types of input (e.g., tasks and information) humans are willing to provide to AI, how AI can effectively utilize this input and transform it into outputs that practitioners perceive as relevant, credible, and actionable, and how systems can be designed to align with human needs and goals. By addressing these questions, it is possible to create systems that not only augment human capabilities but also foster trust, agency, and collaboration. This dissertation takes a human-centered approach to designing HIS for creative work and complex problem-solving. By investigating the interplay between human input and AI output, it develops validated design principles and prototypes to bridge the gap between technical feasibility and socio-technical integration. The research contributes to the advancement of human-centered AI by ensuring that HIS empower individuals and teams in ethically sound and practical ways, ultimately supporting innovation and decision-making in modern creative work and complex problem-solving.

Research Design This cumulative dissertation is anchored in the Design-Science Research (DSR) paradigm and unfolds along its three intertwined cycles—relevance, rigor, and design. Each cycle is linked to the core research questions, ensuring that theoretical inquiry and practical problem-solving continually inform each other. Motivated by the epistemological interests in understanding the socio-technical interplay and human-centered design of HIS, this dissertation adopts an interdisciplinary, mixed-method approach to define the problem space and generate solutions. Literature reviews were conducted to capture existing scientific knowledge and identify research gaps. These reviews informed both descriptive and prescriptive contributions to the cumulative generation of design knowledge. Qualitative methods, including semi-structured interviews, were used to identify real-world challenges and elicit user requirements. These methods allowed the exploration of the types of human input and the corresponding AI output necessary for an effective HIS design. A combination of qualitative and quantitative methods was used to assess design knowledge, ensuring qualitative depth and quantitative breadth. Qualitative approaches, including interviews and content analysis, provided in-depth insights into user perceptions and the feasibility of the proposed solutions. Quantitative methods, such as surveys and structural equation modelling, were employed to validate theoretical constructs and measure the broader impact of HIS on creative work and complex problem-solving. In addition, primary and secondary data stemming from artefact evaluations served as the basis for further complementary statistical analyses. By integrating these diverse research methods, this dissertation ensures a rigorous and comprehensive approach to addressing the challenges of HIS. Iterating between literature reviews, empirical discovery, artefact construction, and multi-method evaluation keeps this research firmly rooted in rigorous scholarship while driving toward actionable, human-centered solutions. In doing so, the research not only advances theoretical understanding of HIS but also delivers validated design knowledge that practitioners can apply in practice.

Results This cumulative dissertation provides validated design knowledge for human-centered HIS and their integration in creative and complex problem-solving contexts. The findings are divided into descriptive and prescriptive contributions, addressing theoretical and practical challenges of designing HIS for creative work and problem-solving. The former focus on understanding the prerequisites and opportunities for human input provision and its utilization by AI. These contributions include identifying the conditions under which individuals are inclined to delegate tasks or share information with AI, such as trust, motivation, and task complexity. For instance, a task delegability framework was refined and validated within a specific knowledge work context, providing empirical evidence of factors influencing the decision to delegate tasks to AI. Additionally, a taxonomy of user interactions with large language models (LLMs) was developed, capturing archetypes of user behaviors and highlighting the diverse ways in which AI can support human creativity and problem-solving. The prescriptive knowledge contributions emphasize actionable design knowledge for the development of HIS. Design principles have been identified to optimize task allocation in HIS, structure creative work in teams, and enable effective human-AI co-creation, inter alia. These principles were instantiated in various artifacts, including conversational agents and a hybrid facilitation system. Evaluations demonstrated how HIS can enhance individual and group creative processes, offering scalable solutions for diverse working settings. Furthermore, developed prompt templates for LLMs demonstrated helpfulness, quality, and relevance of LLM-generated outputs, highlighting the effectiveness of prompt engineering in supporting creative work and problem-solving.

Contribution This dissertation advances the fields of hybrid intelligence, creativity support systems, and human-centered AI by distilling ten publications into a coherent body of validated design knowledge. By synthesizing the findings, it addresses the theoretical and practical challenges of human-centered HIS in creative work and problem-solving contexts. In the context of hybrid intelligence research, the dissertation identifies the prerequisites for human input provision by investigating which tasks and types of information people prefer to delegate or share with AI. The findings show that humans choose to offload routine, time-consuming, and less motivating tasks to AI, while retaining responsibilities involving common sense, human contact, or moral considerations. These insights about what people delegate and why they do so form the basis of a refined task delegability framework, illustrating how task complexity, motivation, and trust influence collaboration preferences. Ethical considerations are also addressed through research on privacy preferences and the conditions under which users are comfortable sharing sensitive personal data with AI systems. These findings highlight the importance of designing HIS that prioritize user trust and transparency while respecting privacy concerns. Building on these descriptive insights, the dissertation highlights opportunities for HIS. It does so by examining different human inputs and corresponding AI outputs that can maximize collaboration efficiency. Additionally, a taxonomy of user interaction patterns with LLMs sheds light on the diverse ways users engage with AI systems, offering an understanding of user behaviors and preferences that can guide the development of adaptive HIS. The dissertation also advances research on creativity support and problem-solving by offering prescriptive design principles and frameworks that guide the development of HIS in these contexts. These include actionable guidelines for designing systems that allocate tasks in HIS according to complementary strengths and support co-creation processes. The presented design knowledge enables the development of HIS that are adaptable to diverse user needs and organizational contexts, enhancing both individual and team-based work processes. Finally, the dissertation introduces overarching design principles that synthesize its central findings. The results of this dissertation also have practical implications for organizations seeking to integrate HIS into creative work and problem-solving contexts. By adopting the design principles and frameworks presented, organizations can develop systems that empower users to contribute meaningful input to AI while leveraging AI’s strengths in data processing and performing routine tasks. Furthermore, the dissertation demonstrates how HIS can improve facilitation by combining complementary human and AI capabilities to address complex tasks and challenges. The contributions provide a robust foundation for practical implementation, helping organizations understand what types of human-AI configurations they should strive for and thereby advancing the design and integration of HIS to support creative work and problem-solving in diverse organizational settings.

Limitations This cumulative dissertation is subject to several limitations relating to the definition of the problem space, the production of the solution space, and the evaluation of the generated design knowledge. While efforts were made to ensure methodological rigor and practical relevance, these limitations reflect the inherent constraints of the research context. When defining the problem space, the dissertation relied on qualitative methods, such as semi-structured interviews, to capture real-world challenges and requirements for HIS. While these methods provided valuable insights, the findings are subject to the interviewees' subjective interpretations. Although established coding and analysis techniques were employed to minimize bias and ensure consistency, complete objectivity cannot be guaranteed. Moreover, the selection of participants and scenarios may limit the generalizability of the identified challenges and user needs. Although diverse perspectives were sought to mitigate this risk, the findings remain context-specific and may not apply to other domains or use cases. A notable limitation is the lack of organizational context in some aspects of the research. While the studies focused on understanding user perceptions and HIS outputs, the integration of HIS within specific organizational structures and workflows was not fully addressed. This restricts the projectability of produced design knowledge, i.e., the practical applicability of the findings for organizations looking to deploy HIS in real-world environments, where factors such as team dynamics, organizational culture, and resource constraints play a critical role. Design knowledge was generated by synthesizing findings from the literature to produce the solution space. The selection of relevant literature and research contexts was guided by established methodologies, but it is possible that certain perspectives or emerging insights were overlooked. This selection process could influence the comprehensiveness of the generated design principles and frameworks. The evaluation of the design knowledge also presents limitations. While qualitative methods provided depth, quantitative methods ensured breadth, with approaches ranging from small-scale interviews to large-scale surveys. However, some studies had a limited number of participants, and some evaluations were conducted in controlled or semi-naturalistic settings rather than real-world environments. These constraints may affect the ecological validity and scalability of the findings. Finally, the cumulative nature of the dissertation poses inherent challenges in maintaining coherence across diverse studies. While each paper contributes to the overarching research goal, integrating the findings may result in certain nuances being overlooked. This is particularly relevant for interdisciplinary topics, where balancing technical, ethical, and socio-technical perspectives is challenging.

Future Research This dissertation sets out several areas for future research into the design and integration of HIS in creative work and problem-solving contexts. While it advances understanding of HIS and generates actionable design knowledge, further research is required to address the identified limitations and expand the applicability of the findings. In the context of socio-technical integration, future research should focus on incorporating organizational contexts more deeply. This would involve studying how HIS can be adapted to suit different organizational structures, workflows, and team dynamics. For example, research could examine the influence of cultural and resource-related factors on the deployment and effectiveness of HIS. Furthermore, longitudinal studies in real-world organizational settings could shed light on the long-term impact of HIS on creative work and problem-solving processes. Further exploration is also required to validate and refine the design principles and frameworks presented in this dissertation. These contributions should be evaluated with larger and more diverse user groups in naturalistic environments to enhance their ecological validity. Furthermore, interdisciplinary approaches combining perspectives from psychology, sociology, and organizational science could offer a more comprehensive understanding of HIS. Research into ethical considerations and user trust is a crucial area for future work. While this dissertation addresses trust in AI and privacy preferences, further studies could examine how HIS can be developed to adapt dynamically to users' evolving ethical concerns and trust levels. Investigating how HIS can incorporate user feedback in real time to align with their expectations would also be valuable. From a methodological perspective, expanding the range of evaluation methods — for example, by using real-time interaction data or observational studies — could provide richer insights into the dynamics of HIS. Additionally, the role of LLMs in supporting creative work requires further investigation, particularly with regard to understanding how their outputs can be fine-tuned to meet specific user needs. Lastly, the taxonomy of user interactions with LLMs developed in this dissertation provides a foundation for adaptive HIS design. Future studies could expand this taxonomy to encompass a broader range of use cases and interaction styles. They could also examine how adaptive systems can dynamically respond to different user archetypes to ensure inclusivity and personalisation in HIS. Addressing these areas will enable future research to build on the contributions of this dissertation and further advance the development of human-centered HIS that empower users and align with the complexities of modern creative work and complex problem-solving.