Granular AI risk modelling from tasks to occupations and workforce implications

Xu, Dawei

Granular AI risk modelling from tasks to occupations and workforce implications

Xu, Dawei

Permalink

Publication Type:: Thesis
Issue Date:: 2025

Open Access

Copyright Clearance Process

Recently Added
In Progress
Open Access

This item is open access.

Adobe PDF

Download thesisAdobe PDF (6.15 MB)

View statistics

Full metadata record

Field	Value	Language
dc.contributor.author	Xu, Dawei
dc.date.accessioned	2026-06-15T21:50:57Z
dc.date.available	2026-06-15T21:50:57Z
dc.date.issued	2025
dc.identifier.uri	http://hdl.handle.net/10453/195382
dc.description	University of Technology Sydney. Faculty of Engineering and Information Technology.	en_US.UTF-8
dc.description.abstract	The rise of artificial intelligence (AI) is reshaping the nature of work at an unprecedented pace. Understanding the risk of this transformation is critical, yet most existing assessments rely on coarse occupation-level measures or static proxies of technological risk, failing to capture the evolving and heterogeneous relationships between technical progress and human labor. These limitations create blind spots in identifying which tasks and occupations are most exposed to change at a granular level. To address this gap, this thesis develops a dynamic, multi-layered framework for modeling AI risk across tasks, occupations, and career trajectories by integrating fine-grained job task data, longitudinal measures of AI performance, and advanced modeling techniques including graph neural networks (GNNs) and large language models (LLMs). Specifically, the thesis first constructs a graph-based occupation–skill network, leveraging GNNs to model how machine risk diffuses across interdependent jobs. Second, it applies LLM-driven classification to over 13,000 job tasks, distinguishing substitution, complementarity, and negligible effects, thereby uncovering substantial within-occupation heterogeneity. Third, it introduces an ontology-anchored AI Exposure Index that aligns human tasks with a curated, benchmark-centered knowledge graph of AI capabilities, enhanced with momentum-weighted measures of technical progress and research attention. Finally, the thesis proposes an interpretable career-mobility model as a foundational extension of its overarching objective to examine AI risk at increasing levels of granularity, linking exposure patterns to individual job transitions and longer-term career development. Through comprehensive experiments on large-scale labor datasets, the thesis demonstrates that AI-related risk is both more pervasive and uneven than previously recognized, with significant implications for individual career trajectories, workforce development, and public policy. Overall, the findings advance measurement of AI’s labor-market impact by bridging job content, technological dynamics, and mobility processes, providing a foundation for task-aware and capability-aligned labor analytics in an era of rapid AI innovation.	en_US.UTF-8
dc.format	Thesis (PhD)
dc.language.iso	en_US	en_US.UTF-8
dc.relation	https://opus.lib.uts.edu.au/bitstream/10453/195382/1/thesis.pdf
dc.rights	info:eu-repo/semantics/openAccess
dc.rights	The author owns the copyright in this thesis including all reproduction and reuse rights for the work. The work may not be altered without the permission of the copyright owner. Attribution is essential when quoting or paraphrasing from this thesis.
dc.rights	© 2025 Dawei Xu
dc.rights	au.edu.uts.lib/cph
dc.title	Granular AI risk modelling from tasks to occupations and workforce implications	en_US.UTF-8
dc.type	Thesis
utslib.copyright.status	open_access	*

Abstract:

The rise of artificial intelligence (AI) is reshaping the nature of work at an unprecedented pace. Understanding the risk of this transformation is critical, yet most existing assessments rely on coarse occupation-level measures or static proxies of technological risk, failing to capture the evolving and heterogeneous relationships between technical progress and human labor. These limitations create blind spots in identifying which tasks and occupations are most exposed to change at a granular level. To address this gap, this thesis develops a dynamic, multi-layered framework for modeling AI risk across tasks, occupations, and career trajectories by integrating fine-grained job task data, longitudinal measures of AI performance, and advanced modeling techniques including graph neural networks (GNNs) and large language models (LLMs). Specifically, the thesis first constructs a graph-based occupation–skill network, leveraging GNNs to model how machine risk diffuses across interdependent jobs. Second, it applies LLM-driven classification to over 13,000 job tasks, distinguishing substitution, complementarity, and negligible effects, thereby uncovering substantial within-occupation heterogeneity. Third, it introduces an ontology-anchored AI Exposure Index that aligns human tasks with a curated, benchmark-centered knowledge graph of AI capabilities, enhanced with momentum-weighted measures of technical progress and research attention. Finally, the thesis proposes an interpretable career-mobility model as a foundational extension of its overarching objective to examine AI risk at increasing levels of granularity, linking exposure patterns to individual job transitions and longer-term career development. Through comprehensive experiments on large-scale labor datasets, the thesis demonstrates that AI-related risk is both more pervasive and uneven than previously recognized, with significant implications for individual career trajectories, workforce development, and public policy. Overall, the findings advance measurement of AI’s labor-market impact by bridging job content, technological dynamics, and mobility processes, providing a foundation for task-aware and capability-aligned labor analytics in an era of rapid AI innovation.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/195382