Work

Mental integration with generative AI through computational thinking

Year: 2024
Type: dissertation
Abstract: The rapid development of artificial intelligence (AI) has been received with both fear and excitement. The future of the field is clouded by uncertainty, inviting speculation of its direction. This thesis views the future as a path of collaboration, wherein humans and AI systems are working together, utilizing their respective strengths. While the collaborative path is not assured, it represents a balanced approach where humans retain a degree of control over AI. As AI systems settle in societies, we are faced with the question: how can we best leverage their potential? To narrow down the question, this thesis focuses on generative AI systems. These systems are often built as user friendly interfaces that appeal for the general populace. Behind the seemingly simple interfaces lie numerous directions towards enhanced collaboration. These directions can be broadly categorized into two groups: (1) enhancing the knowledge of AI systems and (2) refining the AI systems themselves. This thesis attempts to provide guidance by bridging the gap in mutual understanding between the user and the generative AI system. The analysis is approached through computational thinking. According to the results obtained, computational thinking is an integral tool for understanding AI. The relationship with computational thinking and human-AI collaboration is further explored by focusing on the four characteristics of computational thinking. The characteristics include problem decomposition, algorithmic thinking, abstraction, and automation. Each characteristic provides additional techniques for enhanced collaboration. Further findings underline the importance of several prerequisites for efficient collaborative work. Explainable AI, mutual trust, and mutual understanding lay the foundation for collaboration with generative AI systems. As of now, opaque methods such as deep learning present challenges for explainability and mutual understanding. Working with generative AI requires understanding when and how to utilize it. Allocating tasks according to the competencies of both the user and AI system highlights their complementary benefits. Using generative AI can be approached as a problem of finding the right abstractions through prompt engineering. Providing specific prompts that align with the semantics of the generative AI system promotes suitable outputs. Evaluating the output and giving feedback enables iterative improvement, as the AI system learns from its prior outputs.
Author: Hoven, Tuomas
Institution: University of Turku
Language: English
Open Access status: green
OpenAlex ID: W6992741536
Cites: 29
Cited by:
Related to: 10