Analysis & Policy Observatory

Large Language Models (LLMs) such as GPT and similar architectures have revolutionized artificial intelligence by enabling machines to understand and generate human-like text. However, these models are inherently statistical predictors rather than real-time reasoning systems, leading to fundamental limitations in accessing up-to-date information and verifying factual accuracy. This issue is particularly critical in high-stakes domains such as cryptocurrency markets, decentralized finance (DeFi), and autonomous AI agents, where real-time, verifiable, and tamper-proof information is essential for decision-making.In this paper, we introduce AI Oracle, a novel framework that integrates blockchain-powered oracles with LLMs and autonomous agents to ensure real-time access to cryptographically verified knowledge. We compare AI Oracle with both standalone LLMs and retrieval-based systems using the Model Context Protocol (MCP), highlighting significant advantages in factual reliability, adversarial robustness, and interpretability. AI Oracle combines decentralized consensus, immutable storage, and cryptographic attestation to equip AI agents with enhanced resistance to manipulation, hallucination, and misinformation.Beyond architectural improvements, we explore the broader applicability of AI Oracle across domains that require provable correctness and trust—ranging from real-world asset (RWA) tokenization to autonomous agent coordination and decentralized governance. By positioning AI Oracle as a trust-minimized epistemic infrastructure, we propose a new paradigm in AI systems: the fusion of decentralized trust with autonomous reasoning, enabling agents to operate with resilience, transparency, and embedded verifiability across dynamic environments.

ABSTRACT At the core of practice for many non-Indigenous Australian social workers is learning how to work in deeply respectful, culturally appropriate ways alongside First Nations Peoples. Larrakia and Warumungu woman, and Northern Territory social work scholar, Dr Christine/Karen King’s (2011) writings provided the impetus for this inquiry. King (2011) invites social workers to critically reflect on their practice and worldviews. Responding to this invitation, 10 non-Indigenous social workers with policy, academic, and/or practice experience conducted a co-operative inquiry into the question “What have we learnt about practising social work in the Northern Territory, Australia?” We explored the uniqueness of the Northern Territory and identified three interwoven themes. The first theme grounds our practice to Place—on Country. Here, we identified the importance of engaging with the diverse histories of the Peoples we work alongside. Second, we reflected on our vulnerability as social workers and explored the limitations of our professional knowledge base for working with First Nations Peoples. The third theme relates to the importance of engaging with First Nations Peoples’ worldviews. We conclude by affirming the importance of social work practice being led by First Nations People on Country. IMPLICATIONS Social workers working in the Northern Territory alongside First Nations Peoples will benefit from engaging with vulnerability, reflexivity, and critical reflection on their values, understandings, language, and unexamined aspects of practice and cultural self. Social workers in the Northern Territory need to commit to ongoing learning and ground their practice with First Nations’ worldviews and wisdom.

Abstract This paper documents the development of a programme component within an alcohol and other drugs residential service. Designed to improve relational competence, this pilot component was designed to have multiple levels of function: from the specific, for example, stimulating residents to formulate and monitor relationally oriented goals, to the more pacific, for example, embedding relationally focused learning as an ongoing rehabilitative concern. Development of the component began with a literature review. After finalising a set of relationship types (friends, siblings, etc.), this review identified three domains: ‘values’, ‘knowledge’ and ‘skills’. To create an initial draft, these categories were populated by items selected from a multi‐stage inspection of the relevant literature. This draft was then reviewed by multiple stakeholder groups. Feedback from these consultations resulted in substantial modifications to the initial draft. In addition to the component having a broad‐spectrum interventive purpose, the authors propose that this component could play a role in (i) pre‐admission assessment and (ii) programme evaluation insofar as data concerning individual residents are aggregated and analysed.

Abstract This is an introductory article to the AJPA symposium on 'Generating and demonstrating implementation impact'. It provides an overview and synthesis of key themes canvassed in the papers, which includes a focus on implementation for impact regarding measurement, partnership approaches, and storytelling.

Impact—what does it mean and how do we know what “counts”? We all want to do work that has an impact, and this is true of all sectors, whether that be government, public, private, not-for-profit, university, education, or community stakeholders. However, understandings of what it means in practice, what it takes to achieve, and how it can be tracked and calculated remain largely unclear and contested. While the rhetoric of “impact” and the “impact agenda” has become popular in the last decade or so, our practice and research appear to be lagging. In this introductory paper to the special issue on Impact into practice: Demonstrating applied public administration and policy improvement we outline how systems thinking approach can aid understanding of research and education impact on government practice. A systems approach reveals where reliance exists, where responsibility falls, and where new and deepened relationships are needed. While more needs to be done by all parties to acknowledge the collective nature of impact and the necessary reliance on one another, we argue that redistribution of responsibility is needed, including the government’s significant role. Without collective recognition of reliance, responsibility, and relationships in the system of impact, our respective endeavors can only be expected to go so far. By thinking about impact as a system, we can end the “blame game” between university and government sectors, and encourage action within and across sectors, in the pursuit of better outcomes for citizens and society.

Background: Tracking population trends in childhood obesity and identifying target areas for prevention requires accurate prevalence data. This study quantified the magnitude of non-participation bias for mean Body Mass Index-z scores and overweight/obesity prevalence associated with low (opt-in) compared to high (opt-out) participation consent methodologies. Methods: Data arose from all Local Government Areas (LGAs) participating in the Healthy Together Victoria Childhood Obesity Study, Australia. Primary schools were randomly selected in 2013 and 2014 and all Grades 4 and 6 students (aged approx. 9–12 years) were invited to participate via opt-in consent (2013) and opt-out consent (2014). For the opt-in wave N = 38 schools (recruitment rate (RR) 24.3%) and N = 856 students participated (RR 36.3%). For the opt-out wave N = 47 schools (RR 32%) and N = 2557 students participated (RR 86.4%). Outcomes: differences between opt-in and opt-out sample estimates (bias) for mean BMI-z, prevalence of overweight/obesity and obesity (alone). Standardized bias (Std bias) estimates defined as bias/standard error are reported for BMI-z. Results: The results demonstrate strong evidence of non-participation bias for mean BMI-z overall (Std bias = −4.5, p < 0.0001) and for girls (Std bias = −5.4, p < 0.0001), but not for boys (Std bias = −1.1, p = 0.15). The opt-in strategy underestimated the overall population prevalence of overweight/obesity and obesity by −5.4 and −4.5 percentage points respectively (p < 0.001 for both). Significant underestimation was seen in girls, but not for boys. Conclusions: Opt-in consent underestimated prevalence of childhood obesity, particularly in girls. Prevalence, monitoring and community intervention studies on childhood obesity should move to opt-out consent processes for better scientific outcomes.

Background and challenges to implementation Almost 16 million Vietnamese adults currently smoke tobacco, including 45.3% of men and 1.1% of women. In Vietnam more women die from tobacco-related disease (9.5%) than smoke cigarettes, suggesting that women suffer disproportionately greater illness and premature death from exposure to secondhand smoke (SHS). Intervention or response The Vietnam Women's Union developed the “Women create smoke-free homes” national initiative to help protect women and children from exposure to SHS throughout Vietnam. Supported by Vital Strategies and Danson Media, a campaign was designed for the initiative, featuring the personal story of a 41 year-old, non-smoking victim of lung cancer, Ms Nguyen Thi Huong, with a call to action to “Protect yourselves and your loved ones”. Pre-testing research was conducted on initial testimonial edits to strengthen communication of Ms Huong's story. The resulting three 30-second ads were launched by the Women´s Union and disseminated from December 2016 to February 2017 through Facebook, with accompanying Facebook posts, radio spots and public relations activities in print, online, radio and television media. Results and lessons learnt The testimonial ads received 4,152,937 views, with Facebook comments mostly from women tagging male partners. A national tobacco campaign evaluation study (n=2022) found 8% of respondents, including 10% of female non-smoker respondents, recalled the testimonial ads unprompted (representing approximately 4.75 million people aged 15-55 years). Of these, 74% of female non-smokers (and 75% of male smokers) reported trying to make their home 'smokefree' and 77% reported trying to persuade others to quit smoking as a result of seeing the ads, while 67% of male smokers reported making a quit attempt after seeing the ads. Conclusions and key recommendations The campaign demonstrates effectiveness of a national Women's Organization communicating personal stories of tobacco-related harm to women. This supports previous evidence of the power of testimonial approaches and points to further potential opportunities for tobacco control communication in Vietnam and other countries.

Fiber reinforced composites containing a ZrB2 matrix is prepared by reactive melt infiltration, using a Zr based intermetallic compound. Different precursors are evaluated in order to manufacture fiber reinforced preforms and enable a melt infiltration process. In order to infiltrate the porous preform, capillary forces must be high enough to pull the melt into the preform. This is dependent on the pore size, wetting angle between melt and preform, atmosphere during infiltration, dynamic viscosity and surface tension. Additionally it is important to keep the capillary system open during infiltration. The sample preforms contain boron powder, which forms ZrB2 during Zr2Cu melt infiltration. Depending on the precursors used, composition of the infiltrated sample varies in the amount of ZrC and ZrB2. Interactions of fiber and the Zr2Cu melt are evaluated and the microstructure of the infiltrated samples is investigated by means of X-ray diffraction (XRD) and scanning electron microscopy (SEM).