Controlling the "Uncontrollable": Tensions, Agency and Control in AI-Augmentation

Artificial Intelligence (AI) is present in many aspects of our lives. Organisations often adopt AI to increase speed or productivity, or to reduce costs. However, many still struggle to achieve the desired outcomes as the introduction of AI impacts not only the business processes, but more importantly, the individuals and their tasks. This is due to AI’s unique nature as a technology: it can be perceived as another agent that processes data independently, learns, makes inferences, acts, and changes its behaviour based on cues from the environment. Furthermore, AI shows an inherent lack of transparency on how it functions, making it difficult to monitor, fully anticipate, understand, and control its behaviour and outputs. The “black-box” and agentic nature of AI can challenge our feeling of agency and feeling of control, which we try to protect as they are inherent to how we perceive ourselves (i.e., our identities). This can create tensions for humans, particularly when working closely with AI, as in AI-Augmentation. As AI is here to stay, it is important that we understand these topics in-depth to support a successful human-AI collaboration in the future. Therefore, this thesis investigates the tensions humans face when interacting with AI and how humans navigate the challenges to their feeling of agency and feeling of control when sharing tasks with AI in AI-Augmentation. First, this thesis conceptualises potential underlying reasons for the tensions we face when interacting with AI: our inability to establish a “Theory of Mind” of AI and our lack of control mechanisms for AI. Second, the thesis investigates how humans deal with the tensions in practice when sharing tasks with AI in augmentation. This is achieved via one qualitative case study on Knowledge work & Analytical AI and one on Creatives & Generative AI. This thesis contributes to our understanding on the underlying issues we face when working with AI and provides insights into AI-Augmentation in practice. It introduces the conceptualisation of AI-Augmentation as a spectrum, which links to different forms of human-AI collaboration. Furthermore, it shows how tasks in work processes are affected differently by AI, from no effect, emergence of new sub-tasks to automation. Furthermore, this thesis contributes to our understanding on how individuals deal with the tensions faced in AI-Augmentation by uncovering that - depending on the perceived agency of AI - people react at a task level (no/little perceived AI agency), or beyond the task level (higher perceived AI agency) to retain their feeling of agency and feeling of control. The conceptualisation of a layered identity is introduced to understand these different types of reactions, which link to AI agency being perceived on a spectrum as a relational phenomenon. The thesis further extends our understanding of reactions to AI and enabling factors that help people work with AI successfully.

Name dc:type.qualificationname

Doctor of Philosophy (PhD)

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

University of Cambridge

Year dc:date.issued

2025

Author dc:creator

• Leeb, Constanze

Advisor dc:contributor.advisor

• Mortara, Letizia

• Agency
• AI-Augmentation
• Artificial Intelligence
• Control
• Identity

Licence

• https://www.repository.cam.ac.uk/bitstreams/14176d0d-8018-41a4-a8b5-53ee7a1db33a/download
• http://purl.org/NET/rdflicense/allrightsreserved

Language dc:language

eng

Author Identifier

0000-0003-2524-9358

OAI identifier oai:identifier

oai:www.repository.cam.ac.uk:1810/396696