More Self-Reflection in Research Can Lead to Better Science
Four new Nature papers assess the reproducibility, replicability, and robustness of social and behavioural science research, drawing on a database of 3,900 papers compiled by the DARPA‑funded SCORE programme. The analysis, involving over 850 researchers, finds that only about half of the 164 examined effects replicate, with effect sizes shrinking to less than half of the original reports. These results highlight a persistent “decline effect” and underscore the need for stronger methodological standards. The work also showcases global “replication games” as a novel, collaborative validation model.
SciSciGPT: Advancing Human–AI Collaboration in the Science of Science
SciSciGPT is an open‑source, large‑language‑model‑powered AI collaborator built for the science‑of‑science domain. It automates complex research workflows, speeds prototyping, and enhances reproducibility across empirical studies. The authors showcase case studies where the tool streamlines data collection, analysis, and reporting. They...
What’s New in PeaceTech? 10 Notable Developments From 2025
In 2025 ACLED logged over 185,000 violent events, nearly twice the 2021 total, as drones, AI and smartphones become integral to modern warfare. Civilian smartphones now transmit real‑time battlefield data, raising death and injury risks for non‑combatants. Online disinformation and...
Building a Human Resilience Infrastructure for the Age of AI
A new report by Janna Anderson and Lee Rainie gathers hundreds of global tech experts who warn that AI will become an invisible operating system shaping daily life and societal structures within the next decade. Eighty‑two percent predict a significantly...
Beyond Common Ground: How Everyday Places Solve Big Social Challenges
Daniel P. Aldrich’s new book, *Beyond Common Ground*, argues that social infrastructure—parks, libraries, community centers, and even radio—offers a more adaptable solution to systemic challenges than traditional “gray” infrastructure. Drawing on qualitative and quantitative data from nine countries, the author...
A Shakeup Is Coming for the Nation-State
On June 13, 2025 Israel deployed AI‑filtered quadcopter swarms from inside Iranian territory, disabling Iran’s radar and missile sites before a massive bombing campaign. Earlier, on June 1, Ukraine concealed AI‑trained drones in cargo trucks to infiltrate Russian airspace and...
Crisis Engineering: Time-Tested Tools for Turning Chaos Into Clarity
Crisis Engineering, authored by Marina Nitze, Matthew Weaver, and Mikey Dickerson, offers a hands‑on playbook for leading through system‑wide meltdowns. Drawing on the authors’ experience at Layer Aleph, the book outlines a framework that turns chaotic events into opportunities for lasting...
Designing a New Digital Regulator
Academics, advocates, and policymakers have long advocated for a new digital regulator (NDR) to oversee AI and technology markets. On February 25, 2026, GWU’s Institute for Data, Democracy and Politics and Vanderbilt Policy Accelerator convened a summit of leading experts to examine...
‘StravaLeaks’: How Le Monde Located 18,000 French Military Personnel with a Fitness App
Le Monde’s investigation, dubbed “StravaLeaks,” identified roughly 18,000 French military personnel who publicly shared workout data on the Strava app. The disclosed routes pinpointed high‑value assets, including the Charles de Gaulle carrier strike group, nuclear‑submarine base Île Longue, and even the movements of...
Could Retrieval-Augmented Generation with Large Language Models Help Make Local Zoning Codes Easier to Navigate?
Researchers at Urban tested retrieval‑augmented generation (RAG) with large language models on Minneapolis' 467‑page zoning code to see if AI can simplify permit queries. The benchmark showed that RAG‑enhanced models returned more accurate, context‑aware answers than baseline LLMs. City officials...
Why AI Sandboxes Matter for Responsible Innovation and Public Trust
AI regulatory sandboxes are emerging worldwide as structured testbeds for emerging technologies. Three primary models—regulatory, operational, and hybrid—offer varying degrees of oversight and infrastructure. These sandboxes intervene at different stages of policy development, often using waivers to permit experimentation before...
Improving Connecticut’s Public Health Through Cross-Sector Data-Sharing
Connecticut’s Prevention Data Portal, launched in 2018, showcases how cross‑sector data sharing can improve public‑health outcomes without massive new infrastructure investments. The portal aggregates local, state, and federal datasets, delivering free epidemiological profiles, data stories, and infographics on mental health,...
Orchestrating and Designing Data Collaboratives: What Governance Model Is Fit for Purpose?
Stefaan Verhulst’s paper surveys the surge of data‑governance models—data trusts, commons, cooperatives, intermediaries, unions, sandboxes and data spaces—and argues they are not competing solutions but purpose‑driven responses to distinct coordination challenges. He proposes a typology of seven governance archetypes, each...
AI, AGI and the Future
Geoff Mulgan argues that intelligence is the ability to choose, not raw processing speed, and that AI should be viewed as a hybrid ecosystem of tools rather than a single monolithic system. He uses autonomous vehicles and workplace AI stacks...
Steering Technological Progress
In a new NBER working paper, Anton Korinek and Joseph E. Stiglitz explore how policymakers can steer rapid AI‑driven technological progress to expand labor demand and generate higher‑paying jobs. They introduce a theoretical framework that rates innovations by labor‑complementarity, the income level of...
The Economics of Openness: Funding Earth Observation as a Public Good
Earth observation (EO) data are now widely accessible through open archives, cloud platforms and shared tools, yet true public use remains limited. The article argues that openness is more than data availability; it requires institutional capacity, sustained funding, and clear...
Advancing Patient-Centred AI with Adaptive Machine Learning
The paper highlights that fragmented, inequitable health‑care data hampers patient‑centred AI, limiting personalization and equity. It critiques three existing AI pathways and proposes Adaptive Machine Learning (AML) as a fourth, continuously updating models with real‑world, context‑sensitive data. AML rests on...
AI Hazards and Guard Rails
Since its emergence in late 2022, generative AI has accelerated municipal efficiency but also exposed governments to costly hallucinations and factual errors. High‑profile incidents—including a New York City chatbot that gave illegal advice and Deloitte’s $290,000 refund to Australia—highlight the...
The Need to Rename Tech
“The Need to Rename Tech”, edited by Crystal Chokshi and Robin Mansell, gathers scholars who argue that the language used by Big Tech sanitizes the social and political harms of digital tools. The book dissects popular metaphors such as “cloud” and...
Recent Developments in Data Access Policy
The Open Data Policy Lab added 11 new policy developments this quarter across Africa, Europe, Asia, and North America. These entries illustrate five emerging approaches governments use to structure data access, from mandatory public data release to controlled sharing of...
Leveraging Digital Public Goods: Designing Digital Wallets to Unlock Opportunities for Human Security
Digital wallets are evolving from simple payment tools into digital public goods that can verify eligibility for social benefits, share health records, and certify documents. The UNDP Digital X 3.0 webinar, co‑hosted with Japan, highlighted how integrated wallets can streamline access to...
Can AI Strengthen Democracy? Italy’s Parliament Offers a Test Case
Italy’s parliament is launching a pilot program to embed artificial intelligence across its legislative processes, joining a growing but fragmented global trend. The initiative mirrors diverse international experiments, from Chile’s bill‑drafting assistance to Brazil’s citizen‑participation platforms. The Inter‑Parliamentary Union warns...
Human/AI Collective Intelligence for Deliberative Democracy: A Human-Centred Design Approach
The authors introduce Collective Intelligence for Deliberative Democracy (CI4DD), a framework that leverages AI to augment citizen deliberation. They argue that a human‑centred design approach is essential to ensure trustworthy, inclusive processes. The paper outlines a co‑design methodology that maps...
AI in Science
The paper by Agrawal, McHale and Oettl frames artificial intelligence as an augmentation tool that expands scientists' ability to search combinatorial spaces, rather than fully automating research. By dissecting the knowledge‑production process into stages, the authors reveal a “jagged frontier”...
Why Data Matters for Shipbuilding Industrial Policy
The OECD policy brief highlights shipbuilding as a cornerstone of national competitiveness, economic security, and energy efficiency. It warns that many industrial‑policy decisions in leading shipbuilding nations rely on fragmented or outdated evidence. Four core data shortcomings—misaligned statistical definitions, opaque...
Understanding and Improving Data Repurposing
The authors introduce data repurposing as the practice of applying existing datasets to tasks that were not envisioned at collection time. They differentiate repurposing from traditional data reuse, emphasizing new analytical goals and contextual shifts. A structured framework is presented,...
When AI Is Fluent in Data but Illiterate in Context
A generative AI tool analyzed 191 Congolese survey responses on Fourth Industrial Revolution technologies and automatically generated a theme called “Misinformation Resistance.” In reality, the responses reflected a historically rooted, politically informed distrust of institutions, not a cognitive bias. The...
Governing Real-World Health Data as a Public Utility
The article proposes governing real‑world health data as a public utility, using federated, standards‑based, community‑driven models to overcome fragmentation, proprietary control, and weak oversight. It cites ARPA‑H’s interest in economic models and highlights existing distributed networks and research enclaves as...
The Irrational Decision: How We Gave Computers the Power to Choose for Us
Benjamin Recht’s new book, *The Irrational Decision*, chronicles how 1940s mathematicians forged a narrow definition of rationality—treating every choice as a statistical risk. This quantitative framework underpins modern optimization, game theory, statistical testing, and machine learning, accelerating sectors from pharmaceuticals...
From Questions to Institutionalisation – How to Embed Women’s Health Priorities in EU Research and Policy
The Governance Lab and CEPS used the 100 Questions Initiative to move EU women’s‑health research from priority‑setting to implementation. They propose institutionalising question‑driven research, creating a public catalogue of women’s‑health questions, and framing the field as a competitiveness priority. Embedding...
A Large-Language-Model Framework for Automated Humanitarian Situation Reporting
Researchers introduced an end‑to‑end large‑language‑model (LLM) framework that automatically converts heterogeneous humanitarian documents into structured, evidence‑grounded situation reports. The system combines semantic clustering, question generation, retrieval‑augmented answer extraction with citations, and multi‑level summarization, and was tested on over 1,100 documents...
Data Systems at a Crossroads: Official Statistics for a New Era
A new paper by Open Data Watch and Paris 21 warns that deep cuts in development financing, mounting legitimacy concerns, rapid AI advances, and rising expectations for inclusive data are converging into a systemic crisis for national statistical offices, especially in...
General Social Agents
Researchers Manning and Horton propose AI agents that apply social science theories to novel settings without extensive modifications. By feeding theory‑grounded natural language instructions and limited seed‑game data, the agents simulate human behavior across a heterogeneous population of 883,320 generated...
Sovereign Data Supply Chain: Functional and Operational Framework
The Sovereign Data Supply Chain: Functional and Operational Framework version 1.0 proposes a structured governance model for data originating from indigenous and local territories. It aims to replace extractive data practices with sovereign, rights‑based chains across Latin America and the Caribbean....
The AI Lab Next Door
Local colleges and universities across the U.S. are rapidly building AI capabilities, yet their expertise remains largely untapped by city governments and nonprofits. While higher‑education institutions are governing AI internally and creating a pipeline of AI‑trained talent, public sector entities...
AI Agents Are Coming for Government. How One Big City Is Letting Them In
AI agents capable of querying databases and completing transactions are flooding government websites, mixing benign searches with potentially harmful automated actions. Existing public portals, built for human users, lack safeguards against large‑scale machine traffic, exposing agencies to fraud, service hoarding,...
It’s on You
Behavioral economics promised that nudging individuals could solve major societal problems. In *It’s on You*, Nick Chater and George Loewenstein argue that nudges rarely work and serve as a distraction from needed systemic reforms. They claim elites use behavioral science...
Sorting Methods for Online Deliberation: Towards a Principled Approach
The paper by Nicolien Janssens and Frederik van de Putte examines how online deliberation platforms should order citizen proposals. It introduces a conceptual framework that classifies sorting methods by purpose and the variables they consider. The authors critique the prevalent...
Artificial Intelligence and Government
The newly released book *Artificial Intelligence and Government* surveys how AI is reshaping public institutions worldwide, from climate resilience and urban planning to justice and service delivery. It details adoption strategies, readiness frameworks, and real‑world case studies that show governments...
From Data Ambition to Public Value
Governments have moved past debating data use and now face the challenge of governing data responsibly in an AI‑driven era. The article argues that traditional, technocratic data strategies fall short because they prioritize compliance over legitimacy, privacy, and public trust....
He Studied Cognitive Science at Stanford. Then He Wrote a Startling Play About A.I. Authoritarianism.
The Off‑Broadway play “Data” dramatizes a tech firm’s secret project to build a government‑contracted immigration database, exposing the persuasive language tech leaders use to justify authoritarian‑leaning AI. Its protagonist creates a hyper‑accurate predictive algorithm, echoing real‑world advances where startups like...
The Dead Law Theory: The Perils of Simulated Interpretation
Zachary Catanzaro argues that judges consulting ChatGPT for statutory meaning face a fundamental flaw, not merely a reliability issue. Large language models predict token sequences without true semantic comprehension, making computational legal interpretation a category error. He links this flaw to...
Purpose Drives Design: Functions of a Statewide Longitudinal Data System
Statewide longitudinal data systems (SLDS) can boost education and workforce outcomes, but designs vary based on intended functions—public reporting, research analytics, and individual support. The brief by Stefaan Verhulst explains how policymakers can align system architecture, governance, and legal frameworks...
Local Strategies for Engaging Youth with Data
Local organizations receiving grants from the Local Data for Equitable Communities program are training teenagers to collect, analyze, and present data on pressing neighborhood issues such as displacement, air pollution, extreme heat, and limited public spaces. The initiative, highlighted in...
Using LLMs to Enhance Democracy
The paper by Seth Lazar and Lorenzo Manuali evaluates whether large language models (LLMs) can improve democratic deliberation. It examines LLM‑driven summarization, opinion aggregation, and preference prediction, finding mixed outcomes. While AI tools can make political texts more accessible, they...
The Day Europe’s Data Stops Flowing
Europe’s digital economy is increasingly dependent on a complex data infrastructure that remains vulnerable to prolonged outages. The authors model how a systemic failure could evolve from brief inconveniences to widespread power loss, overwhelmed emergency services, and financial disruption within...
Center for Regulatory Ingenuity
The FAS Center for Regulatory Ingenuity (CRI) is launching a transpartisan effort to modernize stagnant government institutions, beginning with climate policy. It creates high‑trust brainstorming environments and a "network of networks" to help policymakers update outdated laws for the clean‑technology...
Seeing in the Dark: Towards a Broad Construction of the Access to Data Provisions of the DSA
The Digital Services Act’s Article 40 gives vetted researchers EU‑wide data access to study systemic risks, but its “necessary and proportionate” test may limit that access. Past denials on privacy grounds show researchers often lack prior knowledge of what data they...
A Digital Omnibus: Identifying Interlinks and Possible Overlaps Between Different Legal Acts in the Field of Digital Legislation to Streamline...
The European Parliament commissioned a study to dissect the European Commission’s Digital Omnibus package released on 19 November 2025. The report separates administrative simplification from substantive changes to safeguards in data protection, privacy, cybersecurity and artificial intelligence. It flags three hot‑button issues...
Predicted: How AI Is Restructuring Social Life
Mona Sloane’s new book *Predicted* argues that artificial intelligence has moved beyond a technological breakthrough to become a core social infrastructure shaping daily interactions and institutional processes. The work frames AI as a co‑produced arrangement built on prediction, classification, and...
