Stefaan Verhulst
Article by Samuel Greengard: “The generosity of strangers has long powered online communities. Spaces like Stack Overflow, GitHub, curl, OpenSSL, Wikipedia, and OpenStreetMap have a rich history of volunteers sharing knowledge and maintaining code, encyclopedias, and maps.
Now, AI is having a say. Trained on content from these sites, it is changing how volunteers contribute and how users consume content. Stack Overflow, for example, has seen monthly questions fall from a peak of roughly 200,000 in 2014 to about 1,400 in July 2026. Meanwhile, a deluge of contributions—often fake and misleading—has overrun these projects.
This has created unease about the future of online knowledge sharing. “We are starting to replace social communities and their underlying processes with a piece of software,” said Gordon Burtch, Allen and Kelli Questrom Professor in Information Systems at Boston University. “As fewer people go to the community with questions, the incentive for experts to be there falls in turn.”
He isn’t the only one sounding alarms. “There is a huge fear, with growing evidence, that AI companies are kicking the legs out from underneath the table they are eating at,” said Benjamin Mako Hill, an associate professor of communication at the University of Washington and a longtime contributor to peer-production projects.
The stakes extend beyond site traffic and human connections. Fundamental questions arise about the quality of AI-provided information, who owns and profits from it, and what role humans play in vetting it. The cycle can also be recursive: models trained on public knowledge inhibit the production of new content. The question now is: will these communities evolve or disappear?..(More)”.
Article by Barbara Wheeler: “…As the Trump administration has taken down or altered hundreds of federal datasets that don’t align with their political or ideological priorities, it has become clear that the old order of our federal data system has effectively collapsed. The data systems that state and local government agencies, researchers, and community organizations once considered trusted, fixed features of the landscape are no longer reliable, have been weaponized against communities, and in some cases no longer exist.
The ongoing federal data collapse now threatens one of the nation’s most essential data sources—the US Census. Conducted every 10 years, census data is used to allocate federal funding to communities, determine political representation in Congress, and inform programs related to housing, transportation, emergency preparedness, and more. But in September 2026, the Trump administration proposed unprecedented changes to the US Census, including a prohibition on collecting data related to race, ethnicity, and sexual orientation, as well as the exclusion of many noncitizen residents. These changes, if adopted, would infuse discrimination, exclusion, and inaccuracy into a critical national data source. Excluding noncitizens would also likely violate the 14th Amendment’s requirement that census results include “the whole number of persons in each State.”
While the destruction of so many federal data sources must be grieved, it also reveals a moment ripe for reimagination. We are living in a Phoenix moment: a volatile, high-potential period for transformation where we can create new and better approaches to federal data. This moment offers us an opportunity to change, and ultimately improve, how our country produces, protects, and uses civic data—information a society gathers about itself to influence governance and determine where resources should flow. Like the mythical bird of the same name, we must first contend with the burning ashes of the old data order and assess the cost of its limitations and its loss. Only then can we rise anew from the ashes with better approaches to federal data—approaches that are more inclusive, more insightful, and more actionable for communities across the nation…(More)”.
Paper by Terence Chau et al: “Program impact estimates that are not statistically significant are often treated as providing little guidance for policy choice, even when economically important effects remain plausible. In such cases, the decision-relevant task is not simply to test whether the effect differs from zero, but to estimate the probability distribution of programmatic effects. The importance of going beyond hypothesis testing is especially clear when evaluations are underpowered relative to the effect needed for benefits to exceed costs. We illustrate this point using Chicago’s contract for ShotSpotter, an acoustical gunshot detection system that reduces police response times and may therefore have improved survival among shooting victims. Using a geographic RD design, we estimate that ShotSpotter reduced the shooting fatality rate by 2.5 percentage points, equivalent to roughly 49 fewer deaths per year. Although the estimate is not statistically significant, its implied benefit far exceeds the cost of the service. To have enough statistical power under the hypothesis testing approach to detect an effect of this size would require 87 years’ worth of Chicago shooting data. We instead implement a decision-relevant perspective using Bayesian analysis, treating the life-saving effect as a random variable. We find positive expected net benefits, with posterior probabilities that benefits exceed costs ranging from 70% under a flat prior and a $0.5 million value of statistical life to 99.6% under a no-harm prior and a $13.6 million value. The results illustrate how reliance on statistical significance can discard decision-relevant evidence when economically meaningful effects are difficult to detect precisely…(More)”.
Paper by Paul Keller: “…Our proposal starts from the observation that individual licensing cannot resolve this. A licensing market, no matter how well it functions, cannot reach most of the content on which AI models are trained. Where licensing does happen, it happens between the largest AI developers and the largest rightholders. It leaves out the long tail of individual creators and small rightholders, and it leaves out smaller AI developers. The way out is to stop trying to reconcile these objectives transaction by transaction, and to do it collectively instead…
The obligation is triggered by the commercial supply, on the EU market, of access to AI models trained on publicly available content. It applies wherever the company is established and wherever the model was trained. Companies pay a percentage of the EU revenue from the supply of model access into a single EU-level fund. In exchange, training on publicly available content is no longer subject to individual reservations under Article 4 of the CDSM Directive.
The remuneration takes the form of a statutory remuneration claim that cannot be waived and is subject to mandatory collective management. Research, and the development and release of models under open-source licences, fall outside the obligation; it arises only when such a model is commercially supplied.
Rightholders keep control, but they exercise it through access rather than through reservations. Content behind paywalls or other access controls stays outside the mechanism and remains available for licensing. Uses at the point of inference, such as retrieval-augmented generation and grounding, are unaffected by the proposal.
The fund splits the proceeds between two destinations. The larger part goes to creators and rightholders through collective management organisations, with a guaranteed share reserved for authors and performers. A significant, fixed share goes to the stewards of publicly available content: cultural heritage institutions such as libraries, archives and museums, open knowledge projects and public service media…(More)”.
Article by Valerie Volcovici: “…Reuters interviewed 15 major environmental, conservation and development organizations about their AI use. All said they used the technology in some form and expressed concern about the environmental impact of the data centers powering it. The centers, filled with servers, can require vast amounts of electricity, along with water for cooling.
It’s a global dilemma for individuals, governments and industries balancing AI’s benefits with its enormous risks. The tradeoffs likely will grab attention at the UN’s COP31 climate conference this month, when host Turkey will call on governments, businesses and other institutions to sign a pledge outlining principles to ensure AI is developed and deployed sustainably…
Many of the organizations that spoke to Reuters believe AI’s benefits currently outweigh its costs because the technology is enabling work that was previously difficult, expensive or impossible.
Kenya’s Save the Elephants, a partner of the Wildlife Conservation Network, said it is using machine learning to improve population surveys and study elephant communication.
Frank Pope, the group’s chief executive, said an aerial survey of Kenya’s largest protected area in 2017 generated roughly 200,000 images that took nine people nine months to review manually. AI models trained on Nvidia’s GeForce graphics platforms can now complete similar analyses in hours.
“The results have been eye-opening,” Pope said. “We found that human surveys were missing about a third of the elephants, roughly half of zebras and giraffes, and three-quarters of oryx populations.”
More accurate wildlife data can help secure financing for conservation, he said.
“Ultimately, these tools make nature measurable and therefore investable,” he said. “That’s a critical transformation, if we’re going to conserve biodiversity at scale.”
At the institute founded by the late conservation icon Jane Goodall, researchers are using AI to digitize decades of field notes about chimpanzee behavior in Tanzania’s Gombe National Park.
The project was once expected to take 13-15 years because the records were written by hand using the local Kiha language and scientific shorthand.
Thanks to a tool developed with Amazon Web Services, the notes can now be scanned, translated and structured in minutes, slashing the time needed to about two years, said Lilian Pintea, the Jane Goodall Institute’s vice president of conservation science.
During her lifetime, Pintea said, Goodall supported the use of AI to advance conservation science, provided people remained firmly in control of how it was used.
“AI technology should be seen as tools which are human-centered,” Pintea said…(More)”.
Paper by Riel Miller: “War recurs not because leaders fail but because the human capacity to imagine the future has been systematically narrowed to a single mode: anticipation-for-the-future (AfF), the colonising drive to design, control, and win tomorrow. This article argues that this narrowing – what the Futures Literacy Framework calls futures illiteracy – is not a deficit correctable by better planning or wider access to foresight tools, but an ontological condition: a way of being in time that reproduces the delusional agency at the root of war. Drawing on the Theory of Anticipation, Whitehead’s concept of concrescence, Bergson’s proposition that the real creates the possible, and feminist critiques of the master’s tools from Lorde and Feukeu, the article develops the distinction between AfF and anticipation-for-emergence (AfE) as a practical way of distinguishing the simultaneity of the closed and open present. It proposes that cultivating futures literacy – understood as the meta-cognitive capacity to hold both anticipatory frames simultaneously – could change the conditions of change: not a remedy for war, but as fundamental shift in the anticipatory contexts from which war draws its so far inescapable necessity…(More)”.
Article by Simone Haarbosch, Tamara Metze and Bas van Vliet: “The inclusion of local and experiential knowledge, interests, and values in the development of just sustainability transitions remains challenging. Deliberative mini-publics have been conceptualized and organized to address these challenges. However, these participatory processes are often instrumental and designed within experts’ and policymakers’ problem framings. Moreover, they often rely primarily on argumentative modes of deliberation. The use of artifacts has been proposed to go beyond speech. This paper explores the opportunities and challenges for the use of visual artifacts in mini-publics. We conducted a thematic analysis of transcripts of 11 mini-publics that we organized with diverse citizen groups (N = 42) in the Province of Overijssel (the Netherlands). We used four types of visual materials: selected images, drawings, geographical maps, and postcards. We explored (1) whether visualized information facilitates more open-ended and inclusive deliberation on energy justice dilemmas, and (2) whether this leads to new insights into various forms of justice in the regional energy transition. The results show that, depending on the type of visual artifact, it facilitates citizens in addressing their local expertise, interests, and values in less cognitive and more affective and relational ways. Selected images and drawings enriched the depth and breadth of deliberations on recognition justice, although they remained within experts’ problem framings, while geographical maps facilitated discussions on distributional fairness in energy systems…(More)”.
Article by Marianne Gros: “The European Commission is now using artificial intelligence to draft new laws and simplify existing ones, it said in an annual review of progress on its ongoing simplification drive.
As part of a major plan to cut red tape and administrative costs for businesses, the EU is making deep reforms to how it creates and enforces rules on everything from tech, agriculture and migration to environmental protection and taxation.
To do that, the Commission is deploying new digital tools internally to “modernise” lawmaking. The tools aim to help its staff “identify similarities across legal acts, pointing at potential overlaps, inconsistencies or duplicates, as well as opportunities for simplification,” it said in an annual simplification report presented on Wednesday.
According to the document, the Commission is also beefing up legal training and staff to make sure new rules are not overly complex.
“Our simplification agenda is more than figures in a report. It is a story unfolding every day in the offices, factories and farms across Europe,” said EU Simplification Commissioner Valdis Dombrovskis at a press conference. Small business owners are spending “less time filling in forms and more time focused on innovation and expansion,” he added.
The EU’s rulebook has been widely criticized for being overly complex and, at times, inconsistent. The Commission’s simplification drive is designed to fix this, making it easier for businesses and governments to comply with its laws by reviewing old legislation and adopting digital tools…(More)”.
Article by John Palfrey: “Sometime in the 18th century, the Enlightenment settled a contentious, centuries-old question: Where does knowledge come from? For millennia, many people had pointed to God. But the early modern period brought a new consensus, namely that knowledge comes from human scientists working at the edge of what is known, armed with observation, mathematics, experimentation, and logic.
As the historian Steven Shapin has noted, this was a social as much as an intellectual revolution. Science started rewiring our shared value system around the idea that research and rigor are virtuous. We held ourselves and our highest institutions to the standard of scientific inquiry.
No political system reflects this shift better than democracy. The ideal of popular sovereignty rests on the principle that each of us is capable of comparing candidates or policy proposals and reaching our own informed conclusion.
Today, we are on the cusp of a similar revolution. In September, one of OpenAI’s models took less than a week to solve a math problem that had stumped scientists for decades; and AI systems are already identifying novel antibiotic candidates by searching chemical spaces too vast for any research team to explore. In a growing number of fields, machines are autonomously producing new information. Humans are still posing the questions and evaluating the outputs, but the generative act—the creative leap across the unknown—is increasingly sparked by systems that do not think the way we do or really think at all.
As we saw with the Scientific Revolution, new information systems often lead to new epistemological and ethical systems. AI implicitly teaches us to value ease, speed, and outsourcing, and these values do have their place in certain contexts. No human can parse as much data as quickly as an AI can, and in sectors like medicine, speed often saves lives.
But as anyone who has watched the internet evolve in recent years knows, moving fast can mean breaking things. We cannot allow our democratic institutions to become one of those things…(More)”.
Article by Julie Morita and Richard Besser: “In 2015, Chicago’s public health department identified three cases of invasive meningococcal disease, a highly dangerous pathogen that can kill within hours. To determine whether these cases signaled a larger outbreak, the department—which one of us, Dr. Julie Morita, led at the time—needed to know not just total cases, but the total population and demographics of affected areas. The team turned to a pillar of our nation’s public health system: the U.S. Census.
The data showed that men who have sex with men—including gay and bisexual men—among the city’s Black population were disproportionately affected. That finding drove the health department’s response, which included working with community partners to surge thousands of free vaccines into neighborhoods across Chicago. Instead of a larger and far deadlier outbreak, the outbreak ended with 11 cases in Chicago and one death.
Census data informs every aspect of our nation’s public health system. As former leaders of local and federal health agencies, we have seen this firsthand. Unfortunately, the Trump Administration has proposed radical changes to the census that would make our country less healthy, less safe, and less fair…(More)”.