ÌÇÐÄVlogÆƽâ°æ

SPEAKERS

DESCRIPTION

Artificial intelligence now sits at the centre of the policy agenda: governments must learn to regulate powerful models even as they enlist those same models to craft more responsive interventions (Abbey, 2023) . Expert synthesis shows that AI could accelerate progress toward most UN Sustainable Development Goals while posing fewer¡ªbut real¡ªrisks (Vinuesa et al., 2020). One crucial contribution is the use and combination of novel unstructured data to address complex research questions that cannot be addressed using existing and costly household surveys and administrative registries. Machine-learning pipelines that repurpose satellite imagery, call-detail records, legislative text, knowledge and technology metadata (patents, publications, and other repositories), policy documents, can shrink data gaps quickly and cheaply, supporting investigation of emerging technologies and their impacts, such as the twin green¨Cdigital transition and giving policymakers non-linear views of complex social¨Cecological systems (Exploring Machine-Learning Trends in Poverty Mapping, 2025) . The session therefore tackles AI not merely as an object of governance but as an analytic partner that can map dynamic social processes onto the SDG landscape¡ªprovided its deployment remains transparent and locally accountable.

At the method level, Athey¡¯s overview of machine-learning techniques for economists highlights how causal forests and related estimators simulate counterfactual reforms at scale (Athey, 2019; Wager & Athey, 2018; Chernozhukov et al., 2018) . In data-scarce settings, computer-vision models trained on open imagery now predict wealth, infrastructure quality and disaster impact with survey-comparable accuracy (Jean et al., 2016; Burke et al., 2021) , and a recent scoping review catalogues when such models generalise across time and space (Sakamoto, Jerzak, & Daoud, 2024) . On the language side, the Text as Data framework formalises best practices for turning bills, court rulings, publications, policy documents, and social-media streams into structured variables, while systems like POPmine show that real-time sentiment extraction can feed directly into governance dashboards (Grimmer, Roberts, & Stewart, 2022; Saleiro et al., 2015) . Collectively these advances illustrate how AI¡ªwhen paired with domain expertise and ethical guardrails¡ªcan help design policies that are adaptive, inclusive and empirically grounded.

This session is part of the ÌÇÐÄVlogÆƽâ°æ Macau AI Conference 2025, which brings together thought leaders, innovators, and stakeholders under the theme AI for Humanity: Building an Equitable Digital Future. To view the full program, please visit the program page. To join the Conference, please visit the .

Celebrating #UN80 and #ÌÇÐÄVlogÆƽâ°æat50 | Oct. 23-25, 2025 | Macau SAR, China
ÌÇÐÄVlogÆƽâ°æ Macau AI Conference 2025 | AI Education Day | AI for SDGs: Global Youth AI Future Innovation Competition