Wikipedia’s ‘neutrality’ has always been complicated. New rules will make questioning it harder

Source: The Conversation – Global Perspectives – By Heather Ford, Professor, Communications, University of Technology Sydney

Franckreporter / Getty Images

Last month, the American non-profit organisation behind Wikipedia issued draft guidelines for researchers studying how neutral Wikipedia really is. But instead of supporting open inquiry, the guidelines reveal just how unaware the Wikimedia Foundation is of its own influence.

These new rules tell researchers – some based in universities, some at non-profit organisations or elsewhere – not just how to study Wikipedia’s neutrality, but what they should study and how to interpret their results. That’s a worrying move.

As someone who has researched Wikipedia for more than 15 years – and served on the Wikimedia Foundation’s own Advisory Board before that – I’m concerned these guidelines could discourage truly independent research into one of the world’s most powerful repositories of knowledge.

Telling researchers what to do

The new guidelines come at a time when Wikipedia is under pressure.

Tech billionaire Elon Musk, who was until recently also a senior adviser to US President Donald Trump, has repeatedly accused Wikipedia of being biased against American conservatives. On X (formerly Twitter), he told users to “stop donating to Wokepedia”.

In another case, a conservative think tank in the United States was caught planning to “target” Wikipedia volunteers it claimed were pushing antisemitic content.

Until now, the Wikimedia Foundation has mostly avoided interfering in how people research or write about the platform. It has limited its guidance to issues such as privacy and ethics, and has stayed out of the editorial decisions made by Wikipedia’s global community of volunteers.

But that’s changing.

In March this year, the foundation established a working group to standardise Wikipedia’s famous “neutral point of view” policies across all 342 versions in different languages. And now the foundation has chosen to involve itself directly in research.

Its “guidance” directly instructs researchers on both how to carry out neutrality research and how to interpret it. It also defines what it believes are open and closed research questions for people studying Wikipedia.

In universities, researchers are already guided by rules set by their institutions and fields. So why do the new guidelines matter?

Because the Wikimedia Foundation has lots of control over research on Wikipedia. It decides who it will work with, who gets funding, whose work to promote, and who gets access to internal data. That means it can quietly influence which research gets done – and which doesn’t.

Now the foundation is setting the terms for how neutrality should be studied.

What’s not neutral about the new guidelines

The guidelines fall short in at least three ways.

1. They assume Wikipedia’s definition of neutrality is the only valid one. The rules of English Wikipedia say neutrality can be achieved when an article fairly and proportionally represents all significant viewpoints published by reliable sources.

But researchers such as Nathaniel Tkacz have shown this idea isn’t perfect or universal. There are always different ways to represent a topic. What constitutes a “reliable source”, for example, is often up for debate. So too is what constitutes consensus in those sources.

2. They treat ongoing debates about neutrality as settled. The guidelines say some factors – such as which language Wikipedia is written in, or the type of article – are the main things shaping neutrality. They even claim Wikipedia gets more neutral over time.

But this view of steady improvement doesn’t hold up. Articles can become less neutral, especially when they become the focus of political fights or coordinated attacks. For example, the Gamergate controversy and nationalist editing have both created serious problems with neutrality.

The guidelines also leave out important factors such as politics, culture, and state influence.

3. They restrict where researchers should direct their research. The guidelines say researchers must share results with the Wikipedia community and “communicate in ways that strengthen Wikipedia”. Any criticism should come with suggestions for improvement.

That’s a narrow view of what research should be. In our wikihistories project, for example, we focus on educating the public about bias in the Australian context. We support editors who want to improve the site, but we believe researchers should be free to share their findings with the public, even if they are uncomfortable.

Neutrality is in the spotlight

Most of Wikipedia’s critics aren’t pushing for better neutrality. They just don’t like what Wikipedia says.

The reason Wikipedia has become a target is because it is so powerful. Its content shapes search engines, AI chatbot answers, and educational materials.

The Wikimedia Foundation may see independent and critical research as a threat. But in fact, this research is an important part of keeping Wikipedia honest and effective.

Critical research can show where Wikipedians strive to be neutral but don’t quite succeed. It doesn’t require de-funding Wikipedia or hunting down its editors. It doesn’t mean there aren’t better and worse ways of representing reality.

Nor does it mean we should discard objectivity or neutrality as ideals. Instead, it means understanding that neutrality isn’t automatic or perfect.

Neutrality is something to be worked towards. That work should involve more transparency and self-awareness, not less – and it must leave space for independent voices.

The Conversation

Heather Ford receives funding from the Australian Research Council. She was previously a member of the Wikimedia Foundation Advisory Board.

– ref. Wikipedia’s ‘neutrality’ has always been complicated. New rules will make questioning it harder – https://theconversation.com/wikipedias-neutrality-has-always-been-complicated-new-rules-will-make-questioning-it-harder-262706

Often parents and schools disagree about whether something is ‘bullying’: what happens next?

Source: The Conversation – Global Perspectives – By Karyn Healy, Honorary Principal Research Fellow in Psychology, The University of Queensland

Monkey Business Images/ Shutterstock

Bullying in schools can can have a devastating impact on victims. Research shows it can lead to reduced academic performance depression, anxiety and even suicidal behaviour. So, preventing and reducing bullying is an urgent priority for governments as well as families and schools.

However, a common obstacle to addressing bullying is that parents and schools often disagree about whether a particular situation constitutes bullying.

A study in Norwegian schools found that when parents think their child is being bullied, around two-thirds of the time, the school does not agree. There are also cases in which the school says a child is bullying others, but the child’s parents don’t agree.

Why is it so complicated? How can parents approach this situation?

What does ‘bullying’ mean?

When we look at the definition of bullying, it is not surprising disagreements occur. Identifying bullying is not clear-cut.

The definition used in Australian schools captures the key elements defined by international research. Bullying is a form of aggression that:

  • is hurtful for the victim

  • happens repeatedly over time

  • involves an intent to harm

  • involves a power imbalance, with victims feeling unable to stop the problem.




Read more:
With a government review underway, we have to ask why children bully other kids


After a report of bullying, what does the school do?

When a student or parent reports bullying, usually the first thing a school does is talk with students, teachers and parents, and observe interactions between students.

However, there are many challenges in working out whether behaviour is bullying.

First, bullying often occurs when adults are not around and students often don’t tell teachers, so direct observation is not always possible.

Second, even if a teacher is present, social forms of bullying can be very subtle, such as turning away to exclude someone, or using a mocking facial expression, so it can be easily overlooked.

Third, determining whether there is “intent to harm” can be difficult as students accused of bullying may claim (rightly or wrongly) they were “only joking” or not intending to hurt or upset.

Fourth, the issue of power is not easy to determine. If the student is older or physically bigger, or if multiple students are involved in bullying, a power difference may seem apparent. But when power is based on popularity, a power difference may not be clear. There are also cases in which students may deliberately accuse others of bullying to get them into trouble (which may in itself constitute bullying).

Finally, not all aggressive behaviour is bullying. For example, conflict that involves arguments or fights between equals is not bullying, as there is no power imbalance. However, this situation can still be upsetting.

A more difficult situation occurs when the victim of bullying reacts aggressively – such as when they lash out angrily to taunts. The aggressive response of the victim may be more visible to teachers than the bullying that provoked the outburst, and this can make the direction of bullying difficult for schools to ascertain.

What if the school and parents disagree?

A school may not prioritise limited resources to resolve cases they do not see as bullying. This can leave the student languishing and can be very distressing for families.

However, research shows parents’ reports that their child has been bullied predict an increased risk of later child anxiety and depression, regardless of whether school staff concur or were even asked if the child was bullied.

So whether or not the school initially agrees a child is being bullied, it is important to improve the situation.

What can be done?

Sometimes, by taking steps to address the situation, the school can find out if bullying is occurring.

For example, sometimes children are upset by behaviours that may seem innocuous – such as humming, tapping or standing close. If this behaviour is not intended to hurt, we would expect children to reduce this when made aware it is upsetting. However, if the behaviour increases or continues, even with reminders, there would be more reason to believe it is deliberately intended to provoke (and is bullying).

One helpful strategy for parents is to keep a careful record of the child’s experiences – exactly what the child experiences and how it impacts them. This can help establish a pattern of hurtful behaviours over time.

It’s important for parents to maintain a good relationship and ongoing communication with the school (however difficult). As bullying can be a complex and evolving issue, good communication can help ensure issues are promptly managed.

The parent can coach the child to manage the situation – for example, to ask in a friendly and confident way for other students to stop when they are doing things they don’t like. The parent can also help the child plan when they would ask a teacher for help.

By working together, and understanding the problem better over time, schools and families can address behaviour that is hurtful – whether or not there is initial agreement it is “bullying”.


If this article has raised issues for you, or if you’re concerned about someone you know, call Lifeline on 13 11 14 or Kids Helpline on 1800 55 1800.

The Conversation

Karyn Healy has received funding from QIMR Berghofer Medical Research Institute, the Australian Research Council and Australian government Emerging Priorities Program and is an honorary Principal Research Fellow with The University of Queensland. Karyn is a co-author of the Resilience Triple P parenting program. Resilience Triple P and all Triple P programs are owned by the University of Queensland. The university has licensed Triple P International Pty Ltd to publish and disseminate Triple P programs worldwide. Royalties stemming from published Triple P resources are distributed to the Parenting and Family Support Centre, School of Psychology, Faculty of Health and Behavioural Sciences and contributory authors. No author has any share or ownership in Triple P International Pty Ltd.

– ref. Often parents and schools disagree about whether something is ‘bullying’: what happens next? – https://theconversation.com/often-parents-and-schools-disagree-about-whether-something-is-bullying-what-happens-next-261474

56 million years ago, Earth underwent rapid global warming. Here’s what it did to pollinators

Source: The Conversation – Global Perspectives – By Vera Korasidis, Lecturer in Environmental Geoscience, The University of Melbourne

Pollinators play a vital role in fertilising flowers, which grow into seeds and fruits and underpin our agriculture. But climate change can cause a mismatch between plants and their pollinators, affecting where they live and what time of year they’re active. This has happened before.

When Earth went through rapid global warming 56 million years ago, plants from dry tropical areas expanded to new areas – and so did their animal pollinators. Our new study, published in Paleobiology today, shows this major change happened in a remarkably short timespan of just thousands of years.

Can we turn to the past to learn more about how interactions between plants and pollinators changed during climate change? That’s what we set out to learn.

A major warming event 56 million years ago

In the last 150 years, humans have raised atmospheric carbon dioxide concentrations by more than 40%. This increase in carbon dioxide has already warmed the planet by more than 1.3°C.

Current greenhouse gas concentrations and global temperature are not only unprecedented in human history but exceed anything known in the last 2.5 million years.

To understand how giant carbon emission events like ours could affect climate and life on Earth, we’ve had to go deeper into our planet’s history.

Fifty-six million years ago there was a major, sudden warming event caused by the release of a gigantic amount of carbon into the atmosphere and ocean. This event is known as the Paleocene-Eocene Thermal Maximum.

For about 5,000 years, huge amounts of carbon entered the atmosphere, likely from a combination of volcanic activity and methane release from ocean sediments. This caused Earth’s global temperature to rise by about 6°C and it stayed elevated for more than 100,000 years.

Although the initial carbon release and climate change were perhaps ten times slower than what’s happening today, they had enormous effects on Earth.

Earlier studies have shown plants and animals changed a lot during this time, especially through major shifts in where they lived. We wanted to know if pollination might also have changed during this rapid climate change.

Paleobotanist Scott Wing, palynologist Vera Korasidis and colleagues searching for new pollen samples in Wyoming from 56 million-year-old rocks.
Richard Barclay

Hunting for pollen fossils in the badlands

We looked at fossil pollen from the Bighorn Basin, Wyoming – a deep and wide valley in the northern Rocky Mountains in the United States, full of sedimentary rocks deposited 50 to 60 million years ago.

The widespread badlands of the modern Bighorn Basin expose remarkably fossil-rich sediments. These were laid down by ancient rivers eroding the surrounding mountains.

We studied fossil pollen because we wanted to understand changes in pollination. Pollen is invaluable for this because it is abundant, widely dispersed in air and water, and resistant to decay – easily preserved in ancient rocks.

We used three lines of evidence to investigate pollination in the fossil record:

  • fossil pollen preserved in clumps
  • how living plants related to the fossils are pollinated today, and
  • the total variety of pollen shapes.
56 million-year-old fossil pollen clumps collected from Wyoming and photographed on the National Museum of Natural History’s scanning electron microscope.
Vera Korasidis

What did we discover?

Our findings show pollination by animals became more common during this interval of elevated temperature and carbon dioxide. Meanwhile, pollination by wind decreased.

The wind-pollinated plants included many related to deciduous broad-leaved trees still common in moist northern hemisphere temperate regions today.

By contrast, the plants pollinated by animals were related to subtropical palms, silk-cotton trees and other plants that typically grow in dry tropical climates.

The decline in wind pollination was likely due to the local extinction of populations of wind-pollinated plants that grew in the Bighorn Basin.

Distant photo of a tall tree with a symmetrical canopy and amber trunk.
A silk-cotton tree (Ceiba pentandra) relies on the wind for pollination.
Klaus Schönitzer/Wikimedia Commons, CC BY-SA

The increase in animal-pollinated plants means that plants from regions with warmer, drier climates had spread poleward and moved into the Bighorn Basin.

Earlier studies have shown these changes in the plants of the Bighorn Basin were related to the climate being hotter and more seasonally dry than before – or after – this interval of rapid climate change.

Pollinating insects and other animals likely moved 56 million years ago along with the plants they pollinated. Their presence in the landscape helped new plant communities establish in the hot, dry climate. It may have provided invaluable resources to animals such as the earliest primates, small marsupials, and other small mammals.

A lesson for our future

What lessons does this ancient climate change event have to offer when we think about our own future?

The large carbon release at the beginning of the Paleocene-Eocene Thermal Maximum clearly resulted in major global warming. It dramatically altered ecosystems on land and in the sea.

In spite of these dramatic changes, most land species and ecological interactions seem to have survived. This is likely because the event occurred at about one-tenth the rate of current anthropogenic climate change.

The forests that returned to the region after more than 100,000 years of hot, dry climate were very similar to those that existed before. This suggests that in the absence of major extinction, forest ecosystems and their pollinators could reestablish into very similar communities even after a very long period of altered climate.

The key for the future may be keeping rates of environmental change slow enough to avoid extinctions.

The Conversation

Vera Korasidis received funding from the University of Melbourne Elizabeth and Vernon Puzey Fellowship Award.

Scott Wing’s fieldwork was supported by the Roland W. Brown fund of the Department of Paleobiology, and by the MacMillan Fund of the National Museum of Natural History.

– ref. 56 million years ago, Earth underwent rapid global warming. Here’s what it did to pollinators – https://theconversation.com/56-million-years-ago-earth-underwent-rapid-global-warming-heres-what-it-did-to-pollinators-260297

Trump administration cuts to terrorism prevention departments could leave Americans exposed

Source: The Conversation – USA – By Kris Inman, Professor of African Studies and Security Studies, Georgetown University

Ghanaian special forces take part in U.S. military-led counterterrorism training near Jacqueville, Ivory Coast, on Feb. 16, 2022. AP Photo/Sylvain Cherkaoui

Staff at the State Department’s Office of Countering Violent Extremism and Bureau of Conflict and Stabilization Operations, which led U.S. anti-violent extremism efforts, were laid off, the units shuttered, on July 11, 2025.

This dismantling of the country’s terrorism and extremism prevention programs began in February 2025. That’s when staff of USAID’s Bureau of Conflict Prevention and Stabilization were put on leave.

In March, the Center for Prevention Programs and Partnerships at the Department of Homeland Security, which worked during the Biden administration to prevent terrorism with a staff of about 80 employees, laid off about 30% of its staff. Additional cuts to the center’s staff were made in June.

And on July 11, the countering violent extremism team at the U.S. Institute of Peace, a nonpartisan organization established by Congress, was laid off. The fate of the institute is pending legal cases and congressional funding.

President Donald Trump in February had called for nonstatutory components and functions of certain government entities, including the U.S. Institute of Peace, to “be eliminated to the maximum extent consistent with applicable law.”

These cuts have drastically limited the U.S. government’s terrorism prevention work. What remains of the U.S. capability to respond to terrorism rests in its military and law enforcement, which do not work on prevention. They react to terrorist events after they happen.

As a political scientist who has worked on prevention programs for USAID, the U.S. Institute for Peace, and as an evaluator of the U.S. strategy that implemented the Global Fragility Act, I believe recent Trump administration cuts to terrorism prevention programs risk setting America’s counterterrorism work back into a reactive, military approach that has proven ineffective in reducing terrorism.

The US war against terrorism

Between 9/11 and 2021, the cost of the U.S. war on terrorism was $8 trillion and 900,000 deaths, according to a Brown University study. Nonetheless, terrorism has continued to expanded in geographic reach, diversity and deadliness.

Though it was territorially defeated in Syria in 2019, the Islamic State – designated a foreign terrorist organization by the U.S. government – has expanded globally, especially in Africa. Its nine affiliates on the continent have joined several al-Qaida-linked groups such as al-Shabab.

The Islamic State has expanded through a decentralized model of operations. It has networks of affiliates that operate semi-autonomously and exploit areas of weak governance in places such as Mali and Burkina Faso. That makes them difficult to defeat militarily.

A building with an arched rood and multiple windows.
The U.S. Institute of Peace in Washington, D.C., on March 18, 2025.
Roberto Schmidt/AFP via Getty Images

These terrorist organizations threaten the U.S. through direct attacks, such as the ISIS-linked attack in New Orleans on Jan. 1, 2025, that killed 14 people.

These groups also disrupt the global economy, such as Houthi attacks on trade routes in the Red Sea.

To understand why terrorism and extremism continue to grow, and to examine what could be done, Congress charged the U.S. Institute of Peace in 2017 to convene the Task Force on Extremism in Fragile States.

This bipartisan task force found that while the U.S. military had battlefield successes, “after each supposed defeat, extremist groups return having grown increasingly ambitious, innovative, and deadly.”

The task force recommended prioritizing and investing in prevention efforts. Those include strengthening the ability of governments to provide social services and helping communities identify signs of conflict – and helping to provide tools to effectively respond when they see the signs.

The report contributed to the Global Fragility Act, which Trump signed in 2019 to fund $1.5 billion over five years of prevention work in places such as Libya, Mozambique and coastal West Africa.

Programs funded by the Global Fragility Act included USAID’s Research for Peace, which monitored signs of terrorism recruitment, trained residents in Côte d’Ivoire on community dialogue to resolve disputes, and worked with local leaders and media to promote peace. All programming under the act has shut down due to the elimination of prevention offices and bureaus.

What the US has lost

The State Department issued a call for funding in July 2025 for a contractor to work on preventing terrorists from recruiting young people online. It stated: “In 2024, teenagers accounted for up to two-thirds of ISIS-linked arrests in Europe, with children as young as 11 involved in recent terrorist plots.”

In the same month, the department canceled the program due to a loss of funding.

It’s the kind of program that the now defunct Office of Countering Violent Extremism would have overseen. The government evidently recognizes the need for prevention work. But it dismantled the expertise and infrastructure required to design and manage such responses.

Lost expertise

The work done within the prevention infrastructure wasn’t perfect. But it was highly specialized, with expertise built over 2½ decades.

Chris Bosley, a former interim director of the violence and extremism program at the U.S. Institute of Peace who was laid off in July, told me recently, “Adequate investment in prevention programs isn’t cheap, but it’s a hell of a lot cheaper than the decades of failed military action, and more effective than barbed wire – tools that come too late, cost too much, and add fuel to the very conditions that perpetuate the threats they’re meant to address.”

For now, the U.S. has lost a trove of counterterrorism expertise. And it has removed the guardrails – community engagement protocols and conflict prevention programs – that helped avoid the unintended consequences of U.S. military responses.

Without prevention efforts, we risk repeating some of the harmful outcomes of the past. Those include military abuses against civilians, prisoner radicalization in detention facilities and the loss of public trust, such as what happened in Guantanamo Bay, in Bagram, Afghanistan, and at various CIA black sites during the George W. Bush administration.

Counterterrorism prevention experts expect terrorism to worsen. Dexter Ingram, the former director of the State Department’s Office of Countering Violent Extremism who was laid off in July, told me: “It seems like we’re now going to try shooting our way out of this problem again, and it’s going to make the problem worse.”

Four men dressed in military gear walk along a city street.
Federal agents patrol New Orleans, La., following a terrorist attack on Jan. 1, 2025.
Andrew Caballero-Reynolds/AFP via Getty Images

What can be done?

Rebuilding a prevention-focused approach with expertise will require political will and bipartisan support.

U.S. Reps. Sara Jacobs, a Democrat from California, and Mike McCaul, a Texas Republican, have introduced a bill that would reauthorize the Global Fragility Act, extending it until 2030. It would allow the U.S. government to continue preventing conflicts, radicalization and helping unstable countries. The measure would also improve the way various government agencies collaborate to achieve these goals.

But its success hinges on securing funding and restoring or creating new offices with expert staff that can address the issues that lead to terrorism.

This analysis was developed with research contributions from Saroy Rakotoson and Liam Painter at Georgetown University.

The Conversation

Kris Inman does not work for, consult, own shares in or receive funding from any company or organization that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.

– ref. Trump administration cuts to terrorism prevention departments could leave Americans exposed – https://theconversation.com/trump-administration-cuts-to-terrorism-prevention-departments-could-leave-americans-exposed-261630

Eddington ends with a dark joke about disability – but its punchline is centuries old

Source: The Conversation – Canada – By Billie Anderson, Ph.D. Candidate, Media Studies, Western University

Joaquin Phoenix, left, who plays small-town sheriff Joe Cross, and Pedro Pascal, who plays the town’s mayor, Ted Garcia, in a scene from Ari Aster’s film ‘Eddington.’ (A24)

This story contains spoilers about ‘Eddington,’ ‘Midsommar’ and ‘Hereditary.’

Ari Aster’s new film Eddington is a political satire set in a small American town, where a feud between Sheriff Joe Cross (Joaquin Phoenix) and Mayor Ted Garcia (Pedro Pascal) begins with disagreement over pandemic policy, but quickly escalates into a chaotic, paranoid power struggle. What starts as a clash of ego spirals into increasingly violent and absurd confrontations.

In the film’s closing minutes, Joe is abruptly stabbed in the head by someone described as “anti-fascist terrorist” after an extended shootout. He survives, and the next time we see him he’s a quadriplegic wheelchair user and non-verbal.

By the end, Joe has been appointed mayor, either by default or out of public pity. The last shot leaves him as the town’s figurehead alongside a giant AI data centre in the desert: a monument to the forces that now shape human life, governance and identity.

It’s brutal, cynical, mean-spirited and a tad esoteric. But Aster seems to be offering a scathing portrait of American politics and the men like Joe who populate it: stubborn, self-righteous, vaguely libertarian; men clinging to authority, even as the world moves on without them.

There’s irony in the film’s portrayal of anarchic protesters, live-streamers and the failed machismo of Joe’s character arc. But when the laughter of the climax fades, what remains is something all too familiar: disability as the poetic consequence of bad behaviour. Not as an ongoing human experience, but as karmic spectacle.

Disability as consequence, not condition

Joe’s fate follows a centuries-old narrative pattern in which disability is framed as punishment, poetic justice or moral revelation. From medieval Christian theology to contemporary satire, bodily difference has long been used to symbolize inner failing. Impairment stands in for sin, ego, corruption or spiritual deficiency.

Enlightenment rationalism and eugenics-era pseudoscience cemented this association, casting disabled bodies as physically manifest proof of social and moral inferiority.

Cinema inherited these tropes, and still leans on them. As disability studies scholars David Mitchell and Sharon Snyder argue, disability is often introduced not to explore a character’s complexity, but to serve as a metaphorical crutch or shortcut to meaning.

Cinematic trailer for Ari Aster’s 2025 film ‘Eddington.’

The disabled body often appears when something needs to be revealed, resolved or punished. In Eddington, Joe’s fate is exactly that.

Disability arrives not as an opening to new experience but as closure: a visual and symbolic transformation from dangerous man to inner object. Joe is no longer speaking, acting or choosing.

The film seems to punish him, not for a single action, but for what he represents: a particular kind of white male authority figure that is combative, self-righteous and increasingly out of touch.

While Joe does oppose the mayor’s early pandemic policies, including mask mandates, his real “crime” in the film is his stubbornness and inflated self-control. He doubles down on personal power, even as systems collapse around him. His disability, then, functions as justice: a final, ironic version of the control he fought so hard to maintain.

He has become a site for meaning-making — a silent figure whose stillness now says everything about the futility of power, the absurdity of authority, the fall of the American sheriff, the political centrist.

Familiar patterns in Aster’s work

Ari Aster’s earlier films follow similar patterns. In Hereditary (2018) Charlie’s facial difference and neurodivergence signal otherness and fragility; her death becomes the hinge on which the horror turns.

In Midsommar (2019), Ruben, a disabled oracle, is portrayed as both holy and someone without personal agency: a vessel for prophecy rather than a fully developed character.

Across Aster’s filmography, disability tends to show up not as life, but as a symbol; as curse, as mysticism, as moral sign. Eddington takes that formula and strips it down even further: Joe becomes disabled at the moment the movie decides he’s no longer needed. It’s efficient, final and familiar.

Disability as visual rhetoric

As disability studies scholar Rosemarie Garland-Thomson writes, disabled bodies are often positioned as visual rhetoric. They are something to be interpreted, not inhabited.

Joe’s silence and stillness as a result of him becoming disabled don’t invite audiences to understand his experience; they invite us to read him, as if his body were a sentence the director had written.

We do see a glimpse of care: Joe appears to be tended to by a man who is now also his mother-in-law’s partner. But even this is played for laughs.

There’s no reckoning with long-term adaptation, no real engagement with the material realities of disability. The body remains an object, not a subject, and once it serves a rhetorical function, the camera moves on.

The real pandemic disabled millions

The symbolic use of disability hits differently given the film’s setting. Eddington is a pandemic movie — a chaotic satire of COVID-era paranoia, misinformation and isolation.

But COVID-19 hasn’t just destabilized governments, upended social norms and exacerbated online political turmoil — it has disabled people.

Millions globally now live with long COVID, facing chronic pain, fatigue, cognitive impairment and dramatic changes to their work, social lives and health-care access.




Read more:
People with long COVID continue to experience medical gaslighting more than 3 years into the pandemic


The pandemic is a mass disabling event, not a metaphor. And yet, Eddington engages with disability only as a punchline. It becomes an ironic punishment for a man too arrogant to admit his limits. The joke lands harder when it treats disability as poetic justice instead of ongoing reality.

Toward fuller representation

Disability studies invites us to consider disability not as a narrative end point, but as a relational and continuous experience shaped by care networks, access barriers, evolving identities and collective adaptation.

What if Eddington had stayed with Joe after his injury? What if it explored his new position not through silence or shame, but through the messy human realities of interdependence?

What if the satire had gone further, asking not just what happens to a sheriff when he’s taken down, but what happens to power when it must rely on care? What if Joe had died and his mother-in-law had become mayor as a serious disruption of legacy politics?

Absurdism and dark comedy about the body aren’t the problem. But the symbolic shorthand of disability as justice carries weight, especially when disabled people continue to be marginalized, disbelieved and erased in the very real world that Eddington pretends to parody.

A sharper satire would engage with disability as part of the social fabric that challenges the audience to reckon with embodiment, dependence and mutual obligation. It could surprise audiences by refusing the easy exit of moral symbolism. It could be more rewarding, radical — and frankly, more intellectual — than another joke equating impairment with comeuppance.

The portrayal of disability in Eddington is not malicious, but it is predictable. It continues a long tradition of using disability to signal judgement, irony or narrative finality.

But real life doesn’t offer such clean punctuation. Disability is not the end of a story, but the start of a more complex, embodied and political one. Until cinema catches up, satires like Eddington will continue to undermine what they claim to critique.

The Conversation

Billie Anderson does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.

– ref. Eddington ends with a dark joke about disability – but its punchline is centuries old – https://theconversation.com/eddington-ends-with-a-dark-joke-about-disability-but-its-punchline-is-centuries-old-262780

‘Better Than Chocolate’ highlights lost 90s decade of lesbian Canadian cinema

Source: The Conversation – Canada – By Tamara de Szegheo Lang, Adjunct Assistant Professor, Department of Film and Media, Queen’s University, Ontario

“If coming out of the closet was really as much fun as it is for the sexually adventurous youths in Better Than Chocolate, then everybody would be doing it, even straight people.”

So wrote film critic Bruce Kirkland in his 1999 review of the lesbian romantic comedy by Canadian filmmaker Anne Wheeler.

film poster showing a woman's naked back who is being embraced by another woman
‘Better Than Chocolate’ was released in 1999.
(Wikipedia/Trimark Pictures/IMBD)

Kirkland pointed out that real life for queer and trans community members was “tougher, harsher and nastier” than portrayed in the 90-minute romp, but also wrote: “To hell with reality, at least for an hour-and-a-half. This movie is a little treasure and offers a lot of pleasure.”

The endearing rom-com won audience choice awards at a number of gay and lesbian film festivals, including in its hometown of Vancouver.

Today, more than a quarter century later, with hate crimes against queer and trans people on the rise and legal protections, especially in the United States, being threatened or ripped away, the film’s lens on romance — and the joy, safety and complications of being in community — may resonate with contemporary viewers and offer a brief reprieve from the heaviness of the political fight.

Like many Canadian lesbian-driven films from the 1990s, it also serves as an example of filmmakers working in queer communities to highlight once-censored voices, and reflects the sheer ingenuity and creative force of community collaboration in this moment — something that has been underrepresented in broader histories of queer and Canadian national cinema.

Whirlwind romance

In Better Than Chocolate, bookstore employee Maggie (Karyn Dwyer) and nomadic artist Kim (Christina Cox) start a whirlwind romance, moving in together within a matter of hours (echoing the classic U-Haul lesbian stereotype).

Their love story is complicated by the arrival of Maggie’s mother Lila (Wendy Crewson), a judgmental woman fresh off a divorce who doesn’t know her daughter is a lesbian. Comedic chaos ensues as the two young lovebirds navigate romantic, familial and community conflicts, all of which are neatly wrapped up by the end.

Though Better Than Chocolate may ultimately be a feel-good comedy, the film captures a community under attack from outside and within.

Skinheads harass Maggie and Kim, culminating in violence. Judy (Peter Outerbridge) is accosted for being transgender and is consistently misgendered by other lesbians.

The Canadian Border Services Agency purposefully targets neurotic bookstore owner, Frances (played by actor, author, playwright and Canadian lesbian icon Ann-Marie Macdonald), for selling queer literature.

Lesbian-centred 90s film

Better Than Chocolate is only one in a wave of lesbian-centred 90s films made in Canada. In this decade, creatives produced at least 12 narrative feature-length lesbian-centred films, several documentaries and over 400 short films.

Some echo Better Than Chocolate’s romantic tone, but the wave includes a diversity of genres – including erotic thrillers, family dramas and experimental dreamscapes.

Some of these films are well-recognized in the Canadian film canon, including Deepa Mehta’s Fire (1996) and Patricia Rozema’s When Night is Falling (1995), while others have been largely forgotten and prove hard to access today, like Patricia Rivera Spencer’s Dreamers of the Day (1990) and Jeanne Crépeau’s Revoir Julie (1998).

Ecosystem behind lesbian Canadian film

Canadian economic, social and artistic contexts offered a vital creative ecosystem that facilitated such a vibrant era of lesbian-driven cinema.

Feminist filmmaking collectives in the 1970s — like Women in Focus (Vancouver), intervisions/ARC (Toronto) and Reel Life (Halifax) — alongside the launch of Studio D at the National Film Board of Canada in 1974 — provided dedicated space for training talent and for producing films about women’s issues.

A woman smiling.
Anne Wheeler.
(www.annewheeler.com)

Wheeler came up through Studio D, co-directing the studio’s first film in 1975.

Canadian artists also had access to several funding sources, including federal, provincial and local arts councils. Beginning in the late 80s, such funding sources were soliciting more diverse content, a result of community activism driven by marginalized artists.

Importantly, a growing network of queer film festivals aided the development of an invested audience willing to pay to watch queer stories.

From 1985 to 2000, at least 11 annual queer festivals were founded in Canada, including Reel Pride (Winnipeg, 1985); Out on Screen (Vancouver, 1988); image+nation (Montréal, 1989); London Lesbian Film Festival (London, 1991); and Inside Out (Toronto, 1991).

With increasing venues to screen queer work and growing audiences came the demand for more films.

A Vancouver lesbian story

Alongside the broader Canadian context, local contexts also encouraged more filmmakers to tell lesbian stories.

Wheeler had long been committed to making films about lesser-represented Western Canada. While most of her films were set in Alberta, Better Than Chocolate moved her focus to Vancouver and its local queer politics.

The dramatic subplot between bookstore owner Frances and the Canada Border Services Agency (CBSA) is a clear reference to the then-ongoing Supreme Court of Canada case involving Little Sister’s Book and Art Emporium in Vancouver.

Little Sister’s, a queer bookstore, had been targeted for years by the CBSA, which would delay shipments while confiscating and sometimes damaging materials that it considered obscene.

The film publicized the homophobia of the CBSA, with Frances comedically demanding to know why books such as Little Red Riding Hood had been confiscated.

As we discovered in our archival research, Janine Fuller, the manager of Little Sister’s, provided feedback on an early draft of the screenplay. A flyer from the film’s production company was also used to raise the visibility of the court case.

Local community ties

The film’s community ties extended further. As noted in archival documents and the film’s press package, Canadian trans activist and performance artist Star Maris inspired the filmmakers when crafting the character of Judy. Her song, “I’m Not a Fucking Drag Queen,” was solicited for use within the film.

Vancouver’s lesbian community was invited to participate as extras in a bar scene, with an advertisement stating, “This is an excellent opportunity to meet new friends, party with old ones, have much fun being in a movie.”

Finally, as Anne Wheeler told Eye Weekly in 1999: “Right from the development phase on, we had a group of 12 young lesbian women whom we consulted with and they told us very specifically what they did and didn’t want to see. … So we set out very intentionally to break the mould and dispose of the old perceptions about gay women.”

In returning to Better Than Chocolate and other films, queer audiences may find entertaining gems, but may also be reminded of the power of survival of queer communities.

Better Than Chocolate is now available on CTV. Don’t stop there! In addition to films named above, check out these other Canadian lesbian-centred 90s feature films.

Tokyo Cowboy (1994)

Skin Deep (1994)

Devotion (1995)

Cat Swallows Parakeet and Speaks (1996)

High Art (1998)

2 Seconds (1998)

Emporte-Moi (1999)

The Conversation

Tamara de Szegheo Lang receives funding from the Social Sciences and Humanities Research Council of Canada (SSHRC)
for the project “Bodies on Fire: Rekindling Canada’s Decade of Lesbian-Driven Filmmaking.”

Dan Vena receives funding from the Social Sciences and Humanities Research Council of Canada (SSHRC) for the project “Bodies on Fire: Rekindling Canada’s Decade of Lesbian-Driven Filmmaking.”

– ref. ‘Better Than Chocolate’ highlights lost 90s decade of lesbian Canadian cinema – https://theconversation.com/better-than-chocolate-highlights-lost-90s-decade-of-lesbian-canadian-cinema-259494

As the status quo shifts, we’re becoming more forgiving when algorithms mess up

Source: The Conversation – Canada – By Hamza Tariq, PhD Student, Cognitive Psychology, University of Waterloo

New inventions — like the printing press, magnetic compasses, steam engines, calculators and the internet — can create radical shifts in our everyday lives. Many of these new technologies were met with some degree of skepticism by those who lived through the transition.

Over the past 30 years alone, we’ve seen our relationship with the internet transform dramatically — it’s fundamentally changed how we search for, remember and learn information; how we evaluate and trust information; and, more recently, how we encounter and interact with artificial intelligence.




Read more:
AI can be responsibly integrated into classrooms by answering the ‘why’ and ‘when’


As new technologies and ways of doing things emerge, we fixate on their flaws and errors, and judge them more harshly than what we’re already familiar with. These apprehensions are not unwarranted. Today, important debates continue around accountability, ethics, transparency and fairness in the use of AI.

But how much of our aversion is really about the technology itself, and how much is driven by the discomfort of moving away from the status quo?

Algorithm aversion

As a PhD student in cognitive psychology, I study human judgment and decision-making, with a focus on how we evaluate mistakes, and how context, like the status quo, can shape our biases.

In my research with cognitive psychologists Jonathan A. Fugelsang and Derek J. Koehler, we tested how people evaluate errors made by humans versus algorithms depending on what they saw as the norm.

Despite algorithms’ track record of consistently outperforming humans in several prediction and judgment tasks, people have been hesitant to use algorithms. This mistrust goes back as far as the 1950s, when psychologist Paul Meehl argued that simple statistical models could make more accurate predictions than trained clinicians. Yet the response from experts at the time was far from welcoming. As psychologist Daniel Kahneman would later put it, the reaction was marked by “hostility and disbelief.”

That early resistance continues to echo in more recent research, which shows that when an algorithm makes a mistake, people tend to judge and punish it more harshly than when a human makes the same error. This phenomenon is now called algorithm aversion.

Defining convention

We examined this bias by asking participants to evaluate mistakes made by either a human or by an algorithm. Before seeing the error, we told them which option was considered the conventional one — described as being historically dominant, widely used and typically relied upon in that scenario.

In half the trials, the task was said to be traditionally done by humans. In the other half, we reversed the roles, indicating that the role had traditionally been done by an algorithmic agent.

When humans were framed as the norm, people judged algorithmic errors more harshly. But when algorithms were framed as the norm, people’s evaluations shifted. They were now more forgiving of algorithmic mistakes, and harsher on humans making the same mistakes.

This suggests that people’s reactions may have less to do with algorithms versus humans, and more to do with whether something fits their mental picture of how things are supposed to be done. In other words, we’re more tolerant when the culprit is also the status quo. And we’re tougher on mistakes that come from what feels new or unfamiliar.

Intuition, nuance and skepticism

Yet, explanations for algorithm aversion continue to make intuitive sense. A human decision-maker, for instance, might be able to consider the nuances of real life like an algorithmic system never could.

But is this aversion really just about the non-human limitations of algorithmic technologies? Or is part of the resistance rooted in something broader — something about shifting from one status quo to another?

These questions, viewed through the historic lens of human relationships with past technologies, led us to revisit common assumptions about why people are often skeptical and less forgiving of algorithms.

Signs of that transition are all around us. After all, debates around AI haven’t slowed its adoption. And for a few decades now, algorithmic tech has already been helping us navigate traffic, find dates, detect fraud, recommend music and movies, and even help diagnose illnesses.

And while many studies document algorithm aversion, recent ones also show algorithm appreciation — where people actually prefer or defer to algorithmic advice in a variety of different situations.

We’re increasingly leaning on algorithms, especially when they’re faster, easier and appear just as (or more) reliable. As that reliance grows, a shift in how we view technologies like AI — and their errors — seems inevitable.

This shift from outright aversion to increasing tolerance suggests that how we judge mistakes may have less to do with who makes them and more to do with what we’re accustomed to.

The Conversation

Hamza Tariq has received funding from the Natural Sciences and Engineering Council (NSERC) of Canada.

– ref. As the status quo shifts, we’re becoming more forgiving when algorithms mess up – https://theconversation.com/as-the-status-quo-shifts-were-becoming-more-forgiving-when-algorithms-mess-up-261166

The bacteria killing sea stars in the Pacific: How our team uncovered a decade-long mystery

Source: The Conversation – Canada – By Melanie Prentice, Research Associate, University of British Columbia

A sunflower sea star in Knight Inlet on the British Columbia coast. (Grant Callegari/Hakai Institute)

In 2013, a mysterious epidemic swept across the Pacific Coast of North America, rapidly turning billions of sea stars from Mexico to Alaska into goo.

Its name, sea star wasting disease (SSWD), describes what you might have seen if you wandered the shores of the Pacific Northwest at that time: contorted sea star bodies and fragmented arms littered the sea floor, as the tissues of sea stars melted away within a matter of days.

In the more than 10 years that has followed, SSWD has been described as the largest epidemic ever recorded in a wild marine species, and one of the 10 greatest unsolved ocean mysteries. Attempts to identify the pathogen responsible have turned up more questions than answers, until now.

We have recently published the findings of our five-year research project into the cause of SSWD.

Our team included multi-national and multi-disciplinary researchers from academic, government and non-profit institutions, a collaboration that was critical for the success of this work.

Together, we conducted laboratory experiments and analyzed data from wild outbreaks of SSWD to identify the pathogen responsible: a novel strain of the bacterium Vibrio pectenicida.

The significance of sea stars

An explainer on sea star wasting disease. (Hakai Institute).

Aside from the devastating loss of these charismatic rocky-shore inhabitants, the implications of this epidemic reverberate throughout the ecosystems sea stars inhabit.

More than two dozen species appear to be afflicted by SSWD, which vary in their susceptibility to the disease. Most susceptible is the sunflower sea star, a voracious marine predator and the largest species of sea star. They can grow to the size of a bicycle tire and have as many as 24 arms.

Almost six billion sunflower sea stars have been lost to SSWD, placing them on the critically endangered species list. With them, the kelp forest ecosystems they help to regulate have disappeared.

Sunflower sea stars keep kelp forests thriving by preying on sea urchins. With the rapid disappearance of these predators, unchecked urchin populations increased rapidly, mowing down kelp forests and replacing these lush, biodiverse ecosystems with urchin barrens.

The loss of kelp has had knock-on effects for the numerous species that rely on it for food and habitat. Alongside this staggering loss of biodiversity, millions of dollars from fishing and tourism are lost to impacted communities.

Although harder to quantify, the ecosystem services of kelp forests are also impacted, including water filtration which improves water quality, sediment stabilization that protects coastlines from erosion and storms, and carbon capture that helps mitigate the effects of climate change.

Recovery of kelp forests, and the species reliant upon them, requires a deeper understanding of SSWD and the sea stars devastated by it. The first step in our research was to identify the culprit responsible, however, the task proved more difficult than we initially anticipated.

a group or large starfish with several arms
A group of sunflower sea stars with SSWD in Knight Inlet, B.C.
(Grant Callegari/Hakai Institute)

A pathogen is revealed

Among the many possible pathogens suspected of causing the SSWD epidemic, it is unsurprising that the culprit turned out to be in the Vibrio species group.

From multiple diseases in corals to cholera in humans, the abundance of harmful Vibrio species in our oceans is on the rise with climate change as these bacteria favour warmer waters.

However, narrowing in on the specific culprit was not an easy undertaking. Vibrio pectenicida has some unusual characteristics not observed in other Vibrio species, allowing it to evade detection for more than a decade.

This was originally considered a risky and potentially unanswerable question. However, years of laboratory experiments and field sampling recently culminated in successful mortality-inducing experiments using a pure culture isolated by Amy Chan, a research scientist at the University of British Columbia, of the novel Vibrio pectenicida strain.

Named FHCF-3, it stands apart from previously identified strains (different genetic variants) by less than three per cent of its genome. Such strains have been isolated from diseased scallop larvae in France, crabs in the United Kingdom and, most recently, geoduck larvae from the Pacific Northwest.

Whether or not these strains can cause SSWD remains to be answered, but the potential implications are clear; with a large host range, broad geographic distribution, and propensity for warm seawater temperatures anticipated with advancing climate change, this pathogen is one to watch.

a five armed starfish on a rock underwater, some of its arms are disintegrating.
A cookie sea star with SSWD near Calvert Island, B.C.
(Grant Callegari/Hakai Institute)

No time to waste

Like removing a blindfold, the identification of the pathogen causing SSWD unveils new opportunities for research and management of the species and ecosystems affected.

First on the list is developing a diagnostic test that can detect the genetic sequence of the pathogen. This would allow researchers to test sea star or seawater samples for its presence.

Much like the COVID-19 test eased humanity out of lockdowns, this test will help inform marine management by helping diagnose healthy versus sick sea stars, and identifying locations best suited for reintroduction efforts.

Another target of future research is to identify resilient sea stars (those that can either prevent infection entirely or fight it off once it takes hold) for conservation breeding.

Resilient individuals, particularly of highly vulnerable species like sunflower sea stars, will be vital for recovery efforts in a warming ocean where Vibrio pectenicida is already widespread.

Despite the odds, the identification of the SSWD pathogen provides a new hopeful vision for our oceans and their inhabitants; one where disease-resilient sunflower sea stars once again roam the sea floor among thriving kelp forests replete with vibrant marine life.

The Conversation

Melanie Prentice receives funding from The Tula Foundation and The Nature Conservancy of California.

Alyssa-Lois Gehman receives funding from the Tula Foundation and The Nature Conservancy. She is affiliated with the Hakai Institute

Drew Harvell receives funding from The Nature Conservancy.

Grace Crandall receives funding from The Nature Conservancy.

– ref. The bacteria killing sea stars in the Pacific: How our team uncovered a decade-long mystery – https://theconversation.com/the-bacteria-killing-sea-stars-in-the-pacific-how-our-team-uncovered-a-decade-long-mystery-259875

Photovoltaïque et réseau électrique : comment une IA fiable et transparente pourrait faciliter la décarbonation

Source: The Conversation – France in French (2) – By Gabriel Kasmi, Chercheur postdoctoral en IA appliquée au système électrique, Mines Paris – PSL

Les énergies renouvelables sont souvent accusées de déstabiliser le réseau électrique. Des outils d’intelligence artificielle permettent de localiser les installations photovoltaïques, facilitant ainsi le suivi de leur déploiement et assurant leur insertion sans heurts au réseau électrique.


Le photovoltaïque est une source d’électricité renouvelable profondément décentralisée. De petite taille mais représentant 99 % des systèmes raccordés au réseau électrique, les installations en toiture chez les particuliers ou dans le tertiaire (ombrières, supermarchés…) ont un poids croissant sur réseau électrique.

Hier centralisée, la production électrique est désormais répartie entre des grandes centrales et des centaines de milliers de petites installations. Ainsi, l’équilibrage du système – essentiel afin d’éviter les blackouts – nécessite de nouveaux moyens afin de tenir compte de cette production décentralisée.

Équilibrer la production d’électricité en intégrant la production solaire décentralisée est tout à fait possible en principe. Encore faut-il savoir où les panneaux sont installés et quelle taille ils font. L’intelligence artificielle (IA) peut être mise à contribution pour automatiser ce suivi… à condition que la fiabilité des données puisse être garantie.

L’intégration des renouvelables au réseau électrique : des défis maîtrisables

Historiquement, la production d’électricité provient des centrales reliées au réseau de transport (haute tension), avant d’être acheminée vers le réseau de distribution (basse tension) et les gros consommateurs industriels. La production était « pilotable », c’est-à-dire que l’on peut l’ajuster assez rapidement en fonction de la demande.

L’essor de l’éolien terrestre et du solaire photovoltaïque, majoritairement raccordés au réseau de distribution, a profondément modifié cette organisation. Aujourd’hui, la production pilotable s’ajuste à une « demande nette », c’est-à-dire à la différence entre la consommation et la production renouvelable.

Par exemple, la demande française en électricité est historiquement sensible à la température (chauffage, climatisation), et la production renouvelable est sensible aux variables climatiques. Cette variabilité n’est pas problématique en soi : l’important est qu’une filière soit observable, c’est-à-dire que l’on peut mesurer ou estimer avec précision sa production – et quantifier les incertitudes de production.

Or actuellement, le photovoltaïque manque d’observabilité, principalement en raison d’une caractérisation incertaine du parc existant de systèmes.

Le photovoltaïque est en effet caractérisé par une très grande variété de systèmes, allant de grandes centrales à des installations sur des toitures individuelles. Une telle diversité d’installations facilite le déploiement rapide à grande échelle et fait du photovoltaïque un vecteur indispensable de la décarbonisation du système électrique, mais cette même caractéristique rend le suivi de son déploiement difficile. Ainsi, 99 % des installations photovoltaïques, représentant un cinquième de la puissance installée, sont ainsi installées sur des toitures (résidentielles ou tertiaires) de manière totalement décentralisée (à la différence par exemple des centrales au sol pour lesquelles les appels d’offres sont centralisés, et donc les capacités mieux connues).

Comment répertorie-t-on aujourd’hui les systèmes installés ?

Le cadastre photovoltaïque, qui recense toutes les installations photovoltaïques raccordées au réseau, est obtenu à partir des déclarations de raccordement et est sujet à des incertitudes, tant en termes de recensement des systèmes que d’estimation de leur puissance installée et de leur répartition géographique.

L’intelligence artificielle, appliquée à l’imagerie aérienne, offre une opportunité unique de cartographier automatiquement et à grande échelle les systèmes photovoltaïques, en estimant leur surface, leur orientation et leur puissance, permettant ainsi d’améliorer notre connaissance sur le nombre, la répartition et les caractéristiques des systèmes photovoltaïques sur toiture. De nombreux modèles ont ainsi été développés afin de cartographier des systèmes photovoltaïques sur toiture.

Cependant, et malgré leurs performances remarquables, ces modèles ne sont que rarement utilisés par les acteurs du système électrique pour compléter et corriger leurs registres, ce qui réduit l’impact de telles méthodes pour répondre au problème dit de l’observabilité du photovoltaïque sur toiture, et plus largement pour faciliter la décarbonation du système électrique.

Quels sont les freins à l’adoption de l’IA comme outil d’aide à la cartographie des systèmes photovoltaïque ?

Au cours de ma thèse, j’ai prêté une attention particulière à l’identification des verrous méthodologiques pour le déploiement d’outils d’IA permettant d’améliorer l’observabilité du photovoltaïque sur toiture. Je suis parti d’un paradoxe apparent : les outils et les méthodes pour détecter les panneaux solaires existaient déjà, mais les projets existants n’allaient pas au-delà des expérimentations en laboratoire. Plusieurs raisons peuvent expliquer ce décalage. D’une part, des facteurs institutionnels ou humains, liés au manque de formation à ces outils ou encore à la question de la responsabilité en cas d’erreur de l’IA. D’autre part, il existe des facteurs liés aux méthodes elles-mêmes, où il est apparu en particulier que la combinaison d’un manque de transparence et de fiabilité des algorithmes était un frein puissant à leur adoption.

Notre travail de recherche récent a ainsi proposé une méthode permettant d’améliorer à la fois la transparence et la fiabilité des algorithmes de cartographie. L’amélioration de la transparence repose sur des techniques d’IA explicable et permet de mieux comprendre comment « voit » le modèle. L’amélioration de la fiabilité repose sur cette meilleure connaissance du processus de décision du modèle.

Nous avons utilisé une technique d’IA explicable inédite qui permet de décomposer entre différentes échelles ce que « voit » l’IA sur l’image. Cette décomposition montre les cas où l’IA s’appuie sur des composantes de l’image facilement altérables (les hautes fréquences de l’image) s’apparentant à des motifs quadrillés dont la taille au sol est de l’ordre d’un mètre par un mètre. On peut facilement comprendre la corrélation entre les panneaux et les grilles, mais on comprend aussi que toutes les grilles ne sont pas des panneaux solaires ; et que tous les panneaux solaires ne présentent pas de grilles.

comment l’IA identifie un panneau solaire
Un exemple de détection expliquée par notre méthode d’explicabilité. À gauche se trouve l’image de panneau, telle que vue par le modèle. Les zones de cette image importantes pour la prédiction du modèle sont surimposées avec des couleurs chaudes (B). Cependant, cette surimposition n’indique pas quels sont les éléments importants dans ces régions. Sur l’animation (C), nous utilisons notre méthode d’explicabilité pour filtrer progressivement l’image jusqu’à ne retenir que ce qui est nécessaire pour que le modèle fasse sa prédiction. Ce filtrage montre que le modèle a besoin du motif en grilles (plutôt que la couleur ou la forme générale du panneau) pour identifier un panneau.
Kasmi et collaborateurs, études de 2024 et 2025, Fourni par l’auteur

Cette méthode permet de mieux anticiper les occurrences de faux positifs mais aussi de faux négatifs.

Par exemple, dans une région où il y a de nombreuses verrières, qui présentent un motif grillé similaire, l’utilisateur anticipera une éventuelle surestimation du nombre réel de panneaux. À l’inverse, dans une région où les panneaux installés sont plus récents et donc plus susceptibles d’être sans cadre, on pourra s’attendre à une sous-estimation du nombre de panneaux. L’identification de potentiels faux négatifs est d’autant plus cruciale que s’il est simple d’éliminer les faux positifs par des post-traitements, il est plus difficile de lutter contre les faux négatifs.

Qu’est-ce que l’IA explicable ?

  • Les algorithmes de deep learning atteignent des performances remarquables dans de nombreuses applications. Ils sont néanmoins très complexes et donc difficiles à interpréter : on parle ainsi souvent de “boîtes noires”.
  • L’IA explicable (XAI) désigne l’ensemble des outils et méthodes permettant de décrire le fonctionnement d’un algorithme de deep learning de manière compréhensible pour un humain. Ces méthodes visent ainsi à améliorer la transparence des algorithmes, faciliter leur débogage, prévoir leurs cas d’échecs ou encore encourager leur adoption par des utilisateurs non spécialistes.
  • Dans notre cas, l’XAI nous a permis d’identifier un processus de décision (le modèle entraîné traduit souvent la présence d’un panneau PV avec un motif en forme de grille sur l’image), ce qui nous a permis d’identifier des cas d’échecs potentiels : la confusion de verrières pour des panneaux solaires.

Comment faire en sorte que l’IA reste fiable alors que les systèmes à détecter évoluent ?

Pour une IA, la fiabilité correspond à la faculté du système à atteindre une précision comparable à celle obtenue lors de l’entraînement initial du modèle, sur une période plus ou moins longue. En pratique, ces données d’entraînement sont figées, tandis que les données sur lesquelles le modèle est utilisé évoluent continuellement. Ainsi, la distribution statistique des données d’entraînement est de moins en moins représentative de celle des données sur lesquelles le modèle est déployé.

Il est établi que les IA sont sensibles à de tels changements de distributions (distribution shifts) et ainsi que les performances théoriques sont de moins en moins représentatives de la réalité. Pour la détection de panneaux solaires, on peut s’attendre à une baisse des détections liée au fait que le modèle n’a jamais « appris » à reconnaître les nouveaux panneaux. Les données d’entraînement sont ainsi biaisées par rapport aux conditions réelles, et les mises à jour des données d’entraînement veilleront à refléter la distribution actualisée des types de panneaux photovoltaïques.

L’IA seule ne résoudra pas l’ensemble des questions induites par la décarbonation du système électrique. Néanmoins, elle peut et doit y contribuer, compte tenu de la maturité des technologies, de la disponibilité des données et de l’appétence générale qui l’entoure. L’enjeu est double : d’une part, comprendre son fonctionnement et ses limites et, d’autre part, améliorer sa fiabilité, de sorte que l’adhésion à cette technologie soit fondée sur le discernement de l’utilisateur, plutôt que sur une foi aveugle en ces outils.

The Conversation

Gabriel Kasmi a reçu des financements de RTE et de l’Association Nationale de la Recherche et de la Technologie (thèse financée dans le cadre d’une convention Cifre). Il est membre du think tank Zenon.

– ref. Photovoltaïque et réseau électrique : comment une IA fiable et transparente pourrait faciliter la décarbonation – https://theconversation.com/photovolta-que-et-reseau-electrique-comment-une-ia-fiable-et-transparente-pourrait-faciliter-la-decarbonation-261681

Pourquoi les Hispaniques évangéliques ont-ils voté Trump ?

Source: The Conversation – France in French (3) – By Marie Gayte, Chercheuse au laboratoire Babel de l’Université de Toulon, spécialiste de la politique américaine contemporaine, Université de Toulon

Depuis son accession au pouvoir, Donald Trump a multiplié les discours hostiles au droit d’asile et au droit du sol. Cent jours après le début de son mandat, il avait déjà publié 140 décrets contre l’immigration illégale, ciblant tout particulièrement les migrants latino-américains. Ces mesures avaient déjà été annoncées durant la campagne présidentielle ; ce qui n’a pas empêché un net accroissement des votes en sa faveur au sein des communautés hispaniques lors de l’élection. Comment l’expliquer ?.


En mai 2016, le Los Angeles Times titrait : « Les évangéliques sont le type de Latinos dont le GOP pourrait remporter les voix. Mais sans doute pas avec Donald Trump. » Quelques mois plus tard, Trump était élu, obtenant au passage 26 % des suffrages latinos. En 2024, lors de sa seconde victoire, il a obtenu 46 % des voix au sein de ces communautés – 14 points de plus qu’en 2020, et un record pour un candidat républicain à une élection présidentielle. Bien que souvent utilisés de manière interchangeable, les termes « Latino » et « Hispanique » désignent deux réalités différentes, et sont remis en question par certains chercheurs car reflets d’une réalité coloniale. Dans le langage courant, « Hispanique » désigne les personnes originaires d’un pays d’Amérique latine hispanophone, tandis que « Latino » peut englober les personnes originaires de pays lusophones.

Quand on décompose ce vote selon la religion des électeurs, on constate que Trump a effectué une percée notable chez les catholiques (41 %, contre 31 % en 2020), et surtout un score remarquable chez les protestants évangéliques (64 %, contre 48 % en 2020).

Historiquement, les Hispaniques des États-Unis ont été très majoritairement catholiques mais, depuis quelques années, une recomposition religieuse s’opère : ils rejoignent soit les rangs des « nones », ces Américains sans affiliation religieuse, soit, dans une moindre mesure, ceux des églises évangéliques. Il semble que cette affiliation les entraîne vers le conservatisme politique – y compris chez les immigrés de première et deuxième génération –, ce qui peut sembler contre-intuitif, étant donné les propos virulents de Trump sur les migrants en provenance des pays situés au sud de la frontière avec le Mexique et ses politiques migratoires.

France 24, 22 octobre 2024.

Un lien net entre évangélisme et immigration

Si l’« exode latino » du catholicisme vers l’évangélisme annoncé par certains chercheurs ne s’est pas concrétisé – la part des évangéliques dans le mix religieux restant relativement stable (entre 14 et 24 % selon les études) –, ces pourcentages s’appliquent à une cohorte sans cesse grandissante, la proportion des Hispaniques étant passée de 16 % de la population en 2010 à 19 % en 2022.

Les immigrés nourrissent la croissance des rangs évangéliques latinos, puisque la majorité d’entre eux sont nés hors des États-Unis, tandis que la part des immigrés hispaniques de première génération se déclarant évangéliques est passée de 22 % à 32 % entre 2008 et 2022.

Les Hispaniques viennent de plus en plus de pays d’Amérique centrale à forte population évangélique. En outre, même quand ils sont catholiques à leur arrivée aux États-Unis, ils sont plus susceptibles de se convertir à l’évangélisme que les Hispaniques nés aux États-Unis, lesquels ont plutôt tendance à tomber dans l’irreligion.

Ces conversions sont par ailleurs très souvent le fruit de l’influence d’autres Hispaniques, souvent originaires du même pays. Les immigrés centre-américains qui ne trouvent pas dans les églises catholiques le soutien dont ils ont besoin seront particulièrement enclins à se tourner vers l’évangélisme s’ils sont issus de pays où celui-ci est déjà très implanté.

Pour certains, la conversion à l’évangélisme, perçu comme fondamentalement états-unien et vecteur de mobilité sociale, représenterait une démarche d’intégration, annonciatrice d’une nouvelle vie, voire un mécanisme offrant une « citoyenneté spirituelle » les rendant moins susceptibles d’être expulsés s’ils sont clandestins. Les églises qu’ils fréquentent diffusent souvent le message qu’ils doivent être « quelqu’un de bien » – ce qui réduira le risque qu’ils soient expulsés –, de « bons chrétiens », qui ne comptent que sur eux-mêmes et respectent la loi.

Individualisme et prospérité

La nature des églises fréquentées par les Hispaniques aux États-Unis est l’une des grilles de lecture qui permet d’expliquer leur adhésion croissante au conservatisme politique. Les deux tiers d’entre eux fréquentent des églises pentecôtistes ou charismatiques, très souvent dirigées par des Hispaniques, immigrés de première génération ou enfants d’immigrés.

Pour la plupart indépendantes, elles sont souvent influencées par l’« Évangile de la Prospérité », dont les idées dépassent les églises qui s’en réclament du fait de l’omniprésence en librairie et sur Internet de ses figures phares. Pour ses tenants, Dieu accorde santé et richesse à ceux dont la foi est suffisamment forte, et la pauvreté, loin d’être une vertu, est le signe d’un manque de foi individuelle.

Ces églises tendent à diffuser un message de responsabilité et d’action individuelle, avec l’aide de Dieu, et de rejet de toute responsabilité et réponse systémique aux problèmes économiques et sociaux. Une vision néolibérale de la société, qui fait de la famille (déjà un pilier pour cette communauté), de la foi, de la valeur travail, de la contribution de l’individu à l’économie du pays, et du rôle des églises plutôt que de l’État dans les mécanismes d’aide sociale, les éléments constitutifs de l’identité états-unienne.

Samuel Rodriguez, directeur la puissante National Hispanic Christian Leadership Conference, l’un des conseillers évangéliques de Trump lors de la campagne de 2016, et qui prononça une des prières lors de la cérémonie d’investiture de janvier 2017, illustre bien cette tendance. Né de parents originaires de Porto Rico, il est à la tête d’une megachurch à Sacramento dont les fidèles sont majoritairement des immigrés latinos de première et deuxième génération. Dans l’émission « Race and Grace », qu’il animait sur la chaîne évangélique TBN en 2018, il affirmait que la solution au racisme n’est pas politique, car il s’agit d’un pêché individuel, « le produit d’un cœur qui refuse de se repentir » et que la solution consiste à changer son cœur avec l’aide de Dieu.

Interrogé sur la Public Broadcasting Service (PBS) juste après la victoire de Trump en novembre dernier au sujet de son projet d’expulser des millions de sans-papiers, il affirmait avoir obtenu en haut lieu l’assurance que seuls les criminels seraient ciblés, mais que les « bonnes familles, travailleuses, qui craignent Dieu, qui sont ici depuis des années, qui ne dépendent pas d’allocations du gouvernement », ne seraient pas visées.

Il n’est pas étonnant dans ce contexte que des chercheurs aient démontré que l’adhésion à ce courant pousse tous les groupes ethniques vers le conservatisme politique. Que Donald Trump se présente en homme d’affaires prospère peut en outre être perçu comme un signe qu’il est béni de Dieu, et sa réussite résonner auprès d’immigrés venus en quête d’une vie meilleure.

Dominionisme et nationalisme chrétien

Au-delà du fait que les Latinos évangéliques sont également plus conservateurs sur les questions de morale sexuelle que leurs homologues catholiques, et ont donc sans doute été séduits par le discours de Trump sur le genre, ils appartiennent souvent à des Églises dites « dominionistes », pour lesquelles Jésus ordonne « de faire de toutes les nations des disciples » : il ne faut donc pas seulement convertir des citoyens, mais les nations elles-mêmes. Il incombe ainsi aux chrétiens de prendre le contrôle des « sept montagnes » que sont la politique, la famille, l’économie, la culture, l’éducation, les médias, et la religion.

Arte, 3 juillet 2025.

C’est le cas de l’église El Rey Jesús, de l’« apôtre » Guillermo Maldonado, né au Honduras, et de sa femme, née en Colombie, à la tête d’un réseau apostolique de plus de 500 églises dans 70 pays. Maldonado, ainsi que Mario Bramnick, Ramiro Pena, Pasqual Urrabazo, eux aussi pasteurs issus de cette mouvance, ont conseillé Trump lors de ses trois campagnes présidentielles. Ils sont fréquemment invités à la Maison Blanche, et c’est dans l’église de Maldonado, à Miami, que Trump a lancé sa campagne en direction des Hispaniques en 2020 – Maldonado ayant dû au préalable rassuré ses fidèles sans papier qu’aucune arrestation n’aurait lieu lors du meeting.

Ceci explique en partie le fait que les Hispaniques évangéliques (57 %) sont juste derrière les évangéliques blancs (66 %) dans leur adhésion au nationalisme chrétien. Ce nationalisme regroupe un ensemble de croyances, dont le fait de considérer les États-Unis comme une nation chrétienne qui doit retrouver ses racines, et dont les chrétiens devraient contrôler les institutions.

Or, les sociologues Samuel Perry et Andrew Whitehead ont montré que l’adhésion au nationalisme chrétien fut un indicateur à part entière d’un vote Trump en 2016. Trump avait, dès cette campagne, promis aux chrétiens qu’ils « retrouveraient le pouvoir » perdu, non seulement politique, mais aussi dans l’éducation et les médias.

Une évolution qui ne passe pas inaperçue

Ces messages conservateurs, largement relayés par les pasteurs, ne sont pas sans effet sur les fidèles, dans la mesure où les Hispaniques évangéliques sont très pratiquants – bien plus que les catholiques – et ne remettent pas en cause la parole de leur pasteur.

Quant à la position de Trump sur les immigrés clandestins, beaucoup d’Hispaniques ne se sentent pas visés, et les Latinos évangéliques ont plus que leurs homologues catholiques tendance à s’identifier aux États-Unis davantage qu’à leur pays d’origine. Ils estiment qu’il faut venir légalement, et ils privilégient leurs convictions sur les questions de morale sexuelle aux préoccupations sur le sort des clandestins.

En 2002, des stratèges démocrates annonçaient l’avènement proche d’une majorité démocrate durable, forte de la montée inexorable des minorités ethniques, forcément acquises au parti de l’âne. Au contraire, aujourd’hui, le conservatisme croissant des Hispaniques évangéliques est suivi de près par le GOP. Il peut compter sur des organisations comme la Faith and Freedom Coalition, fondée par Ralph Reed – l’un des artisans de la mobilisation des évangéliques blancs derrière le GOP dans les années 1990 – qui distribue désormais des guides de vote en espagnol et organise des campagnes d’inscription sur les listes électorales dans les églises de ces pasteurs latinos évangéliques, notamment dans des États clés comme la Pennsylvanie ou l’Arizona.

L’engouement de certains pasteurs pour Trump et son conservatisme est tel que depuis 2020, nombre d’entre eux – immigrés de première ou deuxième génération – se mobilisent d’eux-mêmes pour encourager leurs fidèles à s’inscrire sur les registres électoraux. Ils publient leurs propres guides électoraux selon lesquels il faut voter « selon la Bible », et prier à des meetings de campagne de Trump – et cela, sans même que le parti républicain ne les y incite…

The Conversation

Marie Gayte est chercheuse associée à l’IRSEM et a été titulaire d’une bourse Fulbright recherche en 2023.

– ref. Pourquoi les Hispaniques évangéliques ont-ils voté Trump ? – https://theconversation.com/pourquoi-les-hispaniques-evangeliques-ont-ils-vote-trump-262136