Back to the building blocks: Making forecasting more context aware through human and data-centric practices

The rise of artificial intelligence (AI) and machine learning (ML) presents substantial opportunities for high-stakes fields like health and public policy. While model-centric AI, focusing on algorithmic improvements and architectural innovations, has historically dominated the field, growing calls for more responsible AI have led to the emergence of two complementary paradigms:(1) human-centred AI (HCAI), which emphasizes user-centred design and collaboration with domain experts, and (2) data-centric AI, which focuses on data quality, selection, and curation. This thesis examines the interdependence of these approaches through the DelphAI framework, a sociotechnical methodology that translates expert engagement into data-centric techniques to improve forecasting and address challenges related to responsibility and impact. The rise of artificial intelligence (AI) and machine learning (ML) presents substantial opportunities for high-stakes fields like health and public policy. In response to calls for more responsible AI, three major development paradigms have gained prominence: (1) model-centric AI, which focuses on optimizing architectures, algorithms, and training procedures; (2) data-centric AI, which emphasizes the quality, relevance, and curation of training data; and (3) human-centred AI (HCAI), which promotes collaboration with domain experts and alignment with user needs. This thesis examines the interdependence of these paradigms through the DelphAI framework, a sociotechnical methodology that translates expert engagement into data-centric and model-informed forecasting strategies aimed at improving impact, trust, and deployment in real-world systems.

Grantor dc:publisher

University of Guelph

Author dc:creator

• Kupferschmidt, Kristina L.

Advisor dc:contributor.advisor

• Taylor, Graham

• Human-centered AI
• machine learning
• artificial intelligence
• responsible AI
• time-series forecasting

Statement dc:rights

• Attribution 4.0 International

Licence dc:rights.uri

• http://creativecommons.org/licenses/by/4.0/

Language dc:language.iso

en

Handle dc:identifier.uri

https://hdl.handle.net/10214/29044

OAI record oai:identifier

https://atrium.lib.uoguelph.ca/server/oai/request?verb=GetRecord&metadataPrefix=dim&identifier=oai%3Aatrium.lib.uoguelph.ca%3A10214%2F29044