Conference Agenda

Problem Statement
Multilateral bodies like UNESCO and ASEAN (Association of Southeast Asian Nations) have invested heavily in voluntary AI governance frameworks. However, a basic question has received less attention: do these frameworks actually produce coherent AI policy outcomes among member states? The existing literature reports broad normative convergence around AI governance principles but relies predominantly on qualitative methods [1, 2].

Research Questions

This research addresses two primary questions:

1. How can policy coherence between national AI policies and regional governance frameworks be quantitatively measured?
2. What patterns of coherence and divergence emerge across diverse national contexts within ASEAN?

Research Methodology
This study develops a novel quantitative AI policy coherence index to evaluate the alignment of eleven ASEAN member states with the 2024 ASEAN Guide on AI Governance and Ethics [3]. The methodology involves:

1. Corpus Construction: The study is built on a cross-national corpus of 59 policy documents (2015–2025), drawn from official government sources across all 11 member states, including national AI strategies, legislation, and government-issued guidelines.
2. Scoring Rubrics: The coding framework operationalizes the ASEAN Guide's seven core governance principles into 38 observable sub-principles, each scored on a weighted ordinal scale (0–3), which reflects the progression from "no policy reference," "aspirational statements," "specified governance mechanisms," and "comprehensive implementation frameworks with enforcement provisions." Scores are aggregated and normalized at both the principle and country levels. Coding reliability was validated through temporal re-coding of 20% of documents, with a Cohen's kappa of 0.89.
3. Statistical Analysis Beyond descriptive statistics: The analysis employs (1) cosine similarity to measure directional alignment between each member state's governance profile and the regional reference vector, (2) Pearson correlation to assess whether development across principles is interdependent, and (3) principal component analysis (PCA) to identify the underlying structural dimensions of cross-country variation.

Disciplinary Fields
This research draws on policy coherence theory from UN and OECD scholars [4, 5], multilateral governance theories in international relations, and quantitative policy analysis methods. The study contributes to the emerging field of multilateral AI governance.

Novelty and Policy Relevance
Existing research on AI governance predominantly relies on qualitative content analysis [6] or quantitative topic modeling [7] but falls short on measuring the degree to which national policies align with the principles under a given multilateral framework. This study addresses that gap by developing the first weighted scoring index to quantify national AI policy coherence against a regional multilateral AI governance framework, applied comprehensively across all member states of a soft-law regional organization.
The research is directly relevant to three contemporary debates in communications policy. First, it empirically tests whether voluntary AI governance frameworks foster policy convergence without binding enforcement. Second, it offers a replicable diagnostic tool to identify which AI governance dimensions are lagging. Third, it provides a blueprint for multilateral bodies such as the ITU and UNESCO on designing coordination mechanisms that remain inclusive despite sharp variations in institutional maturity [8].

Results and Conclusions

The analysis reveals variation in AI policy coherence across ASEAN member states (scores ranging from 0.03 to 0.87). This region is divided into a three-regime structure. Five countries (Indonesia, Malaysia, Philippines, Singapore, and Vietnam) demonstrate both high AI policy coherence scores and high directional alignment between each other (cosine similarity > 0.90). In the middle, Brunei, Cambodia, and Thailand display moderate overall AI policy coherence scores but similarly high directional alignment with the first group, indicating their advancement along the same trajectory but at earlier stages of policy development. The third group (Laos, Timor-Leste, and Myanmar) shows near-absent policy articulation across all principles.
PCA reveals that 92.4% of cross-country differences are attributable to a single dimension of overall governance capacity, rather than to different combinations of principle emphasis. Multilateral AI governance in this region operates as a "bundled construct" in which countries advance across all dimensions simultaneously, suggesting that targeted interventions are unlikely to succeed where broader institutional infrastructure is absent.
The broader implication is that soft-law AI governance frameworks appear to work in this region, but only where member states have sufficient institutional foundations to act on them. For ASEAN to deepen this convergence, the harder task is not refining governance principles but helping the region's least-developed members build the capacity to implement them.